Article

Detection of structural damage from glaucoma with confocal laser image analysis

November 1996
Investigative Opthalmology & Visual Science 37(12):2393-401

November 1996
37(12):2393-401

Source
PubMed

Authors:

University of California, Los Angeles

To determine which structural optic nerve head parameters measured with confocal scanning laser image analysis that best discriminate between normal persons and those with glaucoma. One randomly selected eye of 53 patients with early open-angle glaucoma (average visual field mean deviation = -4.8 dB) and of 43 age-, race-, and refractive error-matched normal subjects were studied. The performance of nine structural measures was evaluated with linear multivariate analysis and a neural network: cup area, cup to disc area ratio, rim area, height variation contour, cup volume, rim volume, cup shape measure, mean retinal nerve fiber layer thickness, and retinal nerve fiber layer cross-section area. A discriminant function was derived with two thirds of the sample and its discriminant power tested on the remaining one third. This was repeated twice so that the entire sample was used for training and testing. A neural network was trained and tested in the same way. Stereoscopic color optic nerve photographs of the same eyes were evaluated qualitatively by three experienced, masked observers. Receiver operating characteristic (ROC) curves of discriminant function, neural network results, and qualitative evaluation were plotted. Comparisons of the areas under the ROC curves were performed with nonparametric statistics. There were statistically significant differences between the normal and glaucoma groups for all measures (P < or = 0.007) except for height variation contour, mean retinal nerve fiber layer thickness, and retinal nerve fiber layer cross-section area. Cup shape measure provided the single best measure to distinguish between normal subjects and those with early glaucoma and had a diagnostic precision of 84%. Neural network diagnostic precision, when all measures were used, was 92% and decreased to 82% when cup shape measure was omitted. The area under the ROC curve when all measures were combined was 0.94; it was significantly lower (P = 0.04) when cup shape measure was omitted (area = 0.84). The area under the ROC curve for qualitative optic disc evaluation by experienced observers was 0.93. There was no statistically significant difference between qualitative evaluation and neural network performance (P = 0.80). Cup shape measure, the statistical third moment of the distribution of depth values of the optic nerve head obtained with confocal laser image analysis, can be used to discriminate between normal persons and those with early glaucomatous damage with high diagnostic precision.

Correlation of structural and functional measurements in primary open angle glaucoma (optic disc morphology and psychophysics)

Thesis

Jan 2001

David F Garway-Heath

Background: Primary open angle glaucoma (POAG) is the term given to a progressive optic neuropathy for which the major risk factors are raised intraocular pressure and older age. The presence of glaucoma is defined by functional (visual field) defects that are associated with loss of retinal ganglion cells and neuroretinal tissue at the optic nerve head (ONH). The relationship between the functional and structural changes is, therefore, of great importance to the understanding of the disease process, and to the clinician's interpretation of the state of the disease. This thesis sets out to define the relationship between retinal function, as measured by conventional white-on-white perimetry, and optic nerve head structure, as measured by scanning laser ophthalmoscopy. Plan of research: The investigations are divided into four parts. Firstly, the ONH structural measurements that best distinguish glaucomatous from normal eyes are determined. This includes an analysis of the relationship between the optical components of the eye and image magnification. Secondly, an analysis of the physiological relationship between ganglion cell numbers and retinal function. Thirdly, the establishment of the anatomical relationship between visual field locations and the ONH (a map relating the visual field to the ONH). And fourthly, the investigation of the correlation between structural and functional measurements in POAG. Results: Neuroretinal rim area in relation to optic disc size is the best parameter to distinguish glaucomatous from normal eyes. The physiological relationship of ganglion cell numbers to decibel light sensitivity (10*log[1/light intensity]) is curvilinear and to light sensitivity (1/light intensity) is linear. The visual field/ONH map allows a correlation of sectoral ONH and regional visual field sensitivity. Analyses demonstrate that the relationship of neuroretinal rim area to decibel light sensitivity is curvilinear in glaucoma. Clinical significance: The curvilinear relationship between decibel light sensitivity and neuroretinal rim area indicates that staging of glaucoma by decibel summary indices may underestimate the amount of structural damage in early disease. In addition, the analysis of disease progression by linear modelling of decibel light sensitivity over time may need re-evaluation.

Morphological Features and Important Parameters of Large Optic Discs for Diagnosing Glaucoma

Article

Full-text available

Mar 2015
PLOS ONE

Purpose: To compare the optic disc parameters of glaucomatous eyes to those of non-glaucomatous eyes with large discs. Methods: We studied 225 consecutive eyes with large optic discs (>2.82 mm2): 91 eyes with glaucoma and 134 eyes without glaucoma. An eye was diagnosed with glaucoma when visual field defects were detected by the Humphrey Field Analyzer. All of the Heidelberg Retina Tomograph II (HRT II) parameters were compared between the non-glaucomatous and glaucomatous eyes. A logistic regression analysis of the HRT II parameters was used to establish a new formula for diagnosing glaucoma, and the sensitivity and specificity of the Moorfields Regression Analysis (MRA) was compared to the findings made by our analyses. Results: The mean disc area was 3.44±0.50 mm2 in the non-glaucomatous group and 3.40±0.52 mm2 in the glaucoma group. The cup area, cup volume, cup-to-disc area ratio, linear cup/disc ratio, mean cup depth, and the maximum cup depth were significantly larger in glaucomatous eyes than in the non-glaucomatous eyes. The rim area, rim volume, cup shape measurement, mean retinal nerve fiber layer (RNFL) thickness, and RFNL cross-sectional area were significantly smaller in glaucomatous eyes than in non-glaucomatous eyes. The cup-to-disc area ratio, the height variation contour (HVC), and the RNFL cross-sectional area were important parameters for diagnosing the early stage glaucoma, and the cup-to-disc area ratio and cup volume were useful for diagnosing advanced stage glaucoma in eyes with a large optic disc. The new formula had higher sensitivity and specificity for diagnosing glaucoma than MRA. Conclusions: The cup-to-disc area ratio, HVC, RNFL cross-sectional area, and cup volume were important parameters for diagnosing glaucoma in eyes with a large optic disc. The important disc parameters to diagnose glaucoma depend on the stage of glaucoma in patients with large discs.

Discriminant analysis models for early detection of glaucomatous optic disc changes

Article

May 2000

Michele Iester

AIM To evaluate and compare four different mathematical formulas for the early detection of morphometric optic nerve head changes in chronic open angle glaucoma. METHODS The optic nerve heads of 161 patients with perimetrically defined glaucomatous optic nerve damage and of 194 normal subjects were examined by confocal laser scanning tomography. Using four formulas of linear discriminant analysis and the optic cup shape measure as the single optic disc variable, the predictive power of each of these methods was examined to differentiate between the normal eyes and the glaucoma eyes. RESULTS The highest predictive power had an optic disc sector based formula, in particular in eyes with medium and large optic discs. This optic disc sector based formula was the one with the best agreement with the other formulas examined. It achieved a better predictability than any single optic disc variable evaluated. CONCLUSIONS Combining quantitative optic disc variables by discriminant analysis functions, the predictive power of semiautomatic quantitative optic nerve head evaluation can be improved by providing the ophthalmologist with a diagnostic score for the detection of glaucomatous optic nerve damage. Because of the pattern of glaucomatous neuroretinal rim loss, an optic disc sector based discriminant formula may have a higher diagnostic precision than other formulas in detecting early glaucomatous damage.

The conversion of ocular hypertension to early glaucoma: Epidemiology, risk factors, early detection and treatment

Thesis

Jan 2001

Deborah Kamal

Background: The management of ocular hypertension (OHT) has considerable practical and financial implications for ophthalmology services in the U.K. Previous prophylactic treatment trials for OHT have so far been inconclusive (Kass, 1980, Epstein, 1989, Schulzer, 1991). New methods of detecting early glaucomatous damage are needed to identify ocular hypertensive patients at greatest risk of developing glaucoma, so that appropriate treatment may be targeted at those individuals. Objectives 1. To determine the effect of betaxolol on the conversion rate of OHT to early glaucoma. 2. To identify possible risk factors for conversion. 3. To evaluate methods of early detection of glaucoma, as compared to gold-standard methods. Methods: 356 ocular hypertensives were randomised to treatment with betaxolol drops or placebo, and followed 4 monthly for 2-6 years with visual field testing, intra-ocular pressure (IOP) measurement and optic disc and retinal nerve fibre layer (RNFL) imaging. Conversion was defined using visual field criteria. Results: 1. No overall protective effect of betaxolol against conversion was found as compared to placebo. 2. The converters had significantly higher pre-and post-treatment IOPs than the group of non-converters. Betaxolol had a smaller hypotensive effect on the mean pre-treatment IOP level of the converters. 3. Sequential HRT analysis demonstrated glaucomatous optic disc change, prior to reproducible visual field change in the converters. Some non-converters demonstrated optic disc change despite maintaining normal visual fields. Conclusions: Betaxolol did not affect the conversion rate as compared to placebo, despite having a statistically significant IOP lowering effect. Higher IOP levels are a risk factor for conversion. Betaxolol appeared to have a smaller hypotensive effect in the converting group, and it is possible that these less responsive patients are therefore at greater risk of conversion. The HRT is a useful tool for the early detection of glaucomatous optic disc damage and may identify patients at risk of developing visual field loss.

Identification of glaucomatous optic nerve progression by scanning laser tomography

Thesis

Jan 2003

James Chee-Hian Tan

Analysis of sequential scanning laser tomography of the optic nerve head must be able to tell disease-induced change from measurement variability if it is to be useful for identifying glaucoma progression. Variability in neuroretinal rim area measurement was found to differ between optic nerves and between regions within each nerve, and was influenced by glaucomatous morphology, varying test conditions and different reference planes. Up to 95% of this variability could be explained by fluctuation in the height between the nerve surface and reference plane, and of the nerve head's centre of gravity along the z-axis. Such fluctuation, whether due to image variability or progressive disease, affects the position of conventional reference planes and limits their usefulness as absolute measures of change. A novel reference plane was designed that is customised to each optic nerve head, lies at a depth compatible with least variability, and stays in position despite glaucomatous damage. Its position in any nerve is calculated from surface height at the nerve margin in multiple topography images, and kept constant throughout each image series. It was found that the reference plane's description of neuroretinal rim was more reproducible and corresponded more closely with actual rim appearance compared with conventional reference planes. An analytical approach was devised to identify change based on this new reference plane. Variability in each 30degree sector of rim area in each nerve was estimated by modelling variability within images from all time-points of any nerve's image series. Confidence limits of variability represented variability in each sector, and only change repeatedly exceeding these limits in two of three tests was attributed to disease. Assessed by 90% and 95% limits of variability, progression was identified with a sensitivity and false positive rate of 90% and 6%, and 83% and 3% respectively in ocular hypertension converter eyes with unambiguous glaucomatous visual field change and unchanging eyes of normal controls. When tested in various presentations of suspected and manifest normal-pressure and high-pressure glaucoma, progression was detected in glaucoma suspect eyes without visual field defects, eyes that progressed to develop field defects, and eyes with established and more severe glaucoma.

Effect of the Eye Tracking System on the Reproducibility of Measurements Obtained With Spectral-domain Optical Coherence Tomography in Glaucoma

Article

May 2017

Purpose: The purpose of this study was to evaluate whether the eye tracking system (ETS) improved the reproducibility of a single circle peripapillary retinal nerve fiber layer (RNFL) measurement acquired with spectral-domain optical coherence tomography (OCT). Materials and methods: The sample comprised 205 individuals divided into 2 groups according to intraocular pressure and visual field outcomes. A total of 100 healthy individuals and 105 patients with open-angle glaucoma underwent imaging of the optic nerve head with OCT 3 times during the same session and 2 additional times in subsequent sessions (30 days apart). Intraclass correlation coefficient (ICC), coefficient of variation, and test-retest variability were calculated for the RNFL thickness acquired with and without the ETS enabled, and compared. Results: The glaucoma group mainly comprised patients with moderate glaucoma (mean deviation of standard automated perimetry, -6.73±6.2 dB). The RNFL thicknesses did not differ between acquisitions with or without the ETS activated and disabled. All ICCs were >0.9 in the control and glaucoma groups with or without the ETS activated. The best parameter in the intersession analysis (with ETS activated) was global RNFL thickness (ICC, 0.95; coefficient of variation, 2.7%; and test-retest variability, 2.87 μm). The reproducibility and repeatability of RNFL measurements did not differ significantly between acquisitions with or without the ETS in either group. Conclusions: The reproducibility of peripapillary RNFL thicknesses acquired with OCT was excellent. The variability between OCT measurements did not decrease with the ETS activated.

Optic Nerve Head Image Analysis for Glaucoma Progression Detection

Article

Full-text available

May 2011

Neil O'Leary

Correlación entre daño perimétrico relacionado con glaucoma y espesor de la capa de fibras nerviosas retinianas medido por tomografía óptica coherente

Article

Full-text available

Oct 2014

Objetivo: Determinar si existe correlación entre el espesor de la capa de fibras nerviosas retiniana (CFNR) medido por tomografía óptica coherente (OCT) y el grado de daño glaucomatoso obtenido por perimetría estándar, para determinar si el OCT puede ser útil como herramienta única para medir dicho daño. Material y métodos: Este es un estudio comparativo, transversal, retrospectivo, en el que se revisaron los expedientes clínicos de pacientes con glaucoma primario de ángulo abierto a fin de establecer una relación entre sus perimetrías y el resultado de sus OCT de disco óptico. Resultados: Se examinaron 144 ojos. Cincuenta y cuatro presentaban daño glaucomatoso leve, 28 daño moderado y 62 daño severo de acuerdo con sus perimetrías; el espesor promedio de la capa de fibras nerviosas fue 97.76, 79.16 y 67.73 μm, respectivamente. Se encontró una correlación de significativa entre el espesor de la CFNR y el grado de daño, pero debido a variaciones entre grupos no se pudo establecer un rango exacto de μm para relacionarlo con el daño. Conclusión: El OCT de disco óptico es una herramienta útil que permite tener una idea del nivel de daño glaucomatoso funcional utilizando como parámetro el espesor de la CFNR, pero permite una valoración exacta del mismo.

Investigation of the Role of Convolutional Neural Network Architectures in the Diagnosis of Glaucoma using Color Fundus Photography

Article

Full-text available

Jun 2022

Objectives: To evaluate the performance of convolutional neural network (CNN) architectures to distinguish eyes with glaucoma from normal eyes. Materials and methods: A total of 9,950 fundus photographs of 5,388 patients from the database of Eskişehir Osmangazi University Faculty of Medicine Ophthalmology Clinic were labelled as glaucoma, glaucoma suspect, or normal by three different experienced ophthalmologists. The categorized fundus photographs were evaluated using a state-of-the-art two-dimensional CNN and compared with deep residual networks (ResNet) and very deep neural networks (VGG). The accuracy, sensitivity, and specificity of glaucoma detection with the different algorithms were evaluated using a dataset of 238 normal and 320 glaucomatous fundus photographs. For the detection of suspected glaucoma, ResNet-101 architectures were tested with a data set of 170 normal, 170 glaucoma, and 167 glaucoma-suspect fundus photographs. Results: Accuracy, sensitivity, and specificity in detecting glaucoma were 96.2%, 99.5%, and 93.7% with ResNet-50; 97.4%, 97.8%, and 97.1% with ResNet-101; 98.9%, 100%, and 98.1% with VGG-19, and 99.4%, 100%, and 99% with the 2D CNN, respectively. Accuracy, sensitivity, and specificity values in distinguishing glaucoma suspects from normal eyes were 62%, 68%, and 56% and those for differentiating glaucoma from suspected glaucoma were 92%, 81%, and 97%, respectively. While 55 photographs could be evaluated in 2 seconds with CNN, a clinician spent an average of 24.2 seconds to evaluate a single photograph. Conclusion: An appropriately designed and trained CNN was able to distinguish glaucoma with high accuracy even with a small number of fundus photographs.

Deep learning-based improvement for the outcomes of glaucoma clinical trials

Thesis

Apr 2022

Georgios Lazaridis

Glaucoma is the leading cause of irreversible blindness worldwide. It is a progressive optic neuropathy in which retinal ganglion cell (RGC) axon loss, probably as a consequence of damage at the optic disc, causes a loss of vision, predominantly affecting the mid-peripheral visual field (VF). Glaucoma results in a decrease in vision-related quality of life and, therefore, early detection and evaluation of disease progression rates is crucial in order to assess the risk of functional impairment and to establish sound treatment strategies. The aim of my research is to improve glaucoma diagnosis by enhancing state of the art analyses of glaucoma clinical trial outcomes using advanced analytical methods. This knowledge would also help better design and analyse clinical trials, providing evidence for re-evaluating existing medications, facilitating diagnosis and suggesting novel disease management. To facilitate my objective methodology, this thesis provides the following contributions: (i) I developed deep learning-based super-resolution (SR) techniques for optical coherence tomography (OCT) image enhancement and demonstrated that using super-resolved images improves the statistical power of clinical trials, (ii) I developed a deep learning algorithm for segmentation of retinal OCT images, showing that the methodology consistently produces more accurate segmentations than state-of-the-art networks, (iii) I developed a deep learning framework for refining the relationship between structural and functional measurements and demonstrated that the mapping is significantly improved over previous techniques, iv) I developed a probabilistic method and demonstrated that glaucomatous disc haemorrhages are influenced by a possible systemic factor that makes both eyes bleed simultaneously. v) I recalculated VF slopes, using the retinal never fiber layer thickness (RNFLT) from the super-resolved OCT as a Bayesian prior and demonstrated that use of VF rates with the Bayesian prior as the outcome measure leads to a reduction in the sample size required to distinguish treatment arms in a clinical trial.

Predicting Visual Fields From Optical Coherence Tomography via an Ensemble of Deep Representation Learners

Article

Jan 2022
AM J OPHTHALMOL

Purpose To develop and validate a deep learning (DL) method of predicting visual function from spectral domain optical coherence tomography (SDOCT) derived retinal nerve fiber layer thickness (RNFLT) measurements and corresponding SDOCT images. Design Development and evaluation of diagnostic technology. Methods Two DL ensemble models to predict pointwise VF sensitivity from SDOCT images (model 1 – RNFLT profile only; model 2 – RNFLT profile plus SDOCT image), and two reference models were developed. All models were tested in an independent test-retest dataset comprising 2181 SDOCT/VF pairs; the median of ∼10 VFs per eye was taken as the best available estimate (BAE) of the true VF. The performance of single VFs predicting the BAE VF was also evaluated. Participants Training dataset: 954 eyes of 220 healthy and 332 glaucomatous participants. Test dataset: 144 eyes of 72 glaucomatous participants. Main outcome measures Pointwise prediction mean error (ME), mean absolute error (MAE) and correlation of predictions with the BAE VF sensitivity. Results The median mean deviation was -4.17 (-14.22 - 0.88) dB. Model 2 had excellent accuracy (ME 0.5, standard deviation [SD] 0.8, dB) and overall performance (MAE 2.3, SD 3.1, dB), and significantly (paired t-test) outperformed the other methods. For single VFs predicting the BAE VF, the pointwise MAE was 1.5 (SD 0.7) dB. The association between SDOCT and single VF predictions of the BAE pointwise VF sensitivities was R² = 0.78 and R² = 0.88, respectively. Conclusions Our method outperformed standard statistical and DL approaches. Predictions of BAEs from OCT images approached the accuracy of single real VF estimates of the BAE.

Evaluation of the Optic Nerve Head Parameters with Heidelberg Retinal Tomography and the Retinal Nerve Fibers with Optical Coherence Tomography in Preperimetric Glaucoma

Article

Full-text available

Jan 2020

Haşim Uslu

Comprehensive Glaucoma Imaging

Chapter

Jan 2020

Virtual Reality Improves Clinical Assessment of the Optic Nerve

Article

Full-text available

Aug 2020

The most common approach to assessing the optic nerve head (ONH) in the detection and management of glaucoma relies on frontal stereoscopic images acquired by a fundus camera. Subjective clinical assessment of ONH parameters from these images (e.g., cup/disc ratio and cup depth) is limited by the absence of monocular perspective cues normally available in oblique viewing. This study examined whether viewing a rotatable 3D reconstruction of the ONH could improve the accuracy of subjective assessments by increasing linear perspective information. Images were reconstructed from optical coherence tomography (OCT) of the ONH. Trained optometry students assessed the cup/disc (C/D) ratio of ONHs with either a flat stereoscopic display or virtual reality (VR) head-mounted display (HMD) with or without dynamic slant control. Dynamic stereoscopic assessment of optic nerve head models in VR resulted in larger estimates of C/D ratio and cup depth compared to static stereoscopic assessments. A follow-up experiment using an external display revealed that relative to static monoscopic viewing, adding either dynamic viewing or stereoscopic viewing to the same display improved subjective estimates of C/D ratio relative to Cirrus HD-OCT defined objective values of C/D ratio. The findings suggest that simply changing the viewing orientation of ONH models improves clinical evaluation of C/D ratio by generating perspective cues to depth without the need for stereo viewing.

A deep learning approach to predict visual field using optical coherence tomography

Article

Full-text available

Jul 2020
PLOS ONE

We developed a deep learning architecture based on Inception V3 to predict visual field using optical coherence tomography (OCT) imaging and evaluated its performance. Two OCT images, macular ganglion cell-inner plexiform layer (mGCIPL) and peripapillary retinal nerve fibre layer (pRNFL) thicknesses, were acquired and combined. A convolutional neural network architecture was constructed to predict visual field using this combined OCT image. The root mean square error (RMSE) between the actual and predicted visual fields was calculated to evaluate the performance. Globally (the entire visual field area), the RMSE for all patients was 4.79 ± 2.56 dB, with 3.27 dB and 5.27 dB for the normal and glaucoma groups, respectively. The RMSE of the macular region (4.40 dB) was higher than that of the peripheral region (4.29 dB) for all subjects. In normal subjects, the RMSE of the macular region (2.45 dB) was significantly lower than that of the peripheral region (3.11 dB), whereas in glaucoma subjects, the RMSE was higher (5.62 dB versus 5.03 dB, respectively). The deep learning method effectively predicted the visual field 24–2 using the combined OCT image. This method may help clinicians determine visual fields, particularly for patients who are unable to undergo a physical visual field exam.

Quantitative analysis of the optic nerve head parameters in patients with age-related macular degeneration

Article

Full-text available

Oct 2019
TURK J MED SCI

Background/aim: To evaluate the topographic parameters of the optic disc of patients with age-related macular degeneration (AMD) by performing confocal scanning laser ophthalmoscopy. Materials and methods: This prospective study included 41 eyes of 41 patients with neovascular AMD, 56 eyes of 56 patients with nonneovascular AMD, and 48 eyes of 48 healthy control subjects. Images of the optic nerve head of all of the participants were obtained using Heidelberg retinal tomography III software 3.1. The following stereometric parameters were measured for each participant: disc area, cup area, rim area, cup volume, rim volume, cup-to-disc ratio, mean cup depth, maximum cup depth, cup shape, and mean retinal nerve fiber layer thickness. Results: The cup shape values of the patients with neovascular and nonneovascular AMD were significantly different from those of the control subjects (P = 0.002 and P < 0.001, respectively). The cup-to-disc ratio was significantly higher in the patients with nonneovascular AMD when compared with the control subjects (P = 0.013), but no difference was found between the patients with neovascular AMD and the control subjects (P > 0.05). No significant differences were observed among the 3 groups with respect to the other optic disc parameters (P > 0.05). Conclusion: These data showed that the deterioration of the cup shape was an important finding in patients with AMD. Because AMD manifests with progressive ocular damage, including the optic nerve head, examination of the cup shape may be important during the follow-up of these patients.

Glaucoma management in the era of artificial intelligence

Article

Full-text available

Oct 2019

Glaucoma is a result of irreversible damage to the retinal ganglion cells. While an early intervention could minimise the risk of vision loss in glaucoma, its asymptomatic nature makes it difficult to diagnose until a late stage. The diagnosis of glaucoma is a complicated and expensive effort that is heavily dependent on the experience and expertise of a clinician. The application of artificial intelligence (AI) algorithms in ophthalmology has improved our understanding of many retinal, macular, choroidal and corneal pathologies. With the advent of deep learning, a number of tools for the classification, segmentation and enhancement of ocular images have been developed. Over the years, several AI techniques have been proposed to help detect glaucoma by analysis of functional and/or structural evaluations of the eye. Moreover, the use of AI has also been explored to improve the reliability of ascribing disease prognosis. This review summarises the role of AI in the diagnosis and prognosis of glaucoma, discusses the advantages and challenges of using AI systems in clinics and predicts likely areas of future progress.

Artificial intelligence in glaucoma

Article

Dec 2018

PURPOSE OF REVIEW: The use of computers has become increasingly relevant to medical decision-making, and artificial intelligence methods have recently demonstrated significant advances in medicine. We therefore provide an overview of current artificial intelligence methods and their applications, to help the practicing ophthalmologist understand their potential impact on glaucoma care. RECENT FINDINGS: Techniques used in artificial intelligence can successfully analyze and categorize data from visual fields, optic nerve structure [e.g., optical coherence tomography (OCT) and fundus photography], ocular biomechanical properties, and a combination thereof to identify disease severity, determine disease progression, and/or recommend referral for specialized care. Algorithms have become increasingly complex in recent years, utilizing both supervised and unsupervised methods of artificial intelligence. Impressive performance of these algorithms on previously unseen data has been reported, often outperforming standard global indices and expert observers. However, there remains no clearly defined gold standard for determining the presence and severity of glaucoma, which undermines the training of these algorithms. To improve upon existing methodologies, future work must employ more robust definitions of disease, optimize data inputs for artificial intelligence analysis, and improve methods of extracting knowledge from learned results. SUMMARY: Artificial intelligence has the potential to revolutionize the screening, diagnosis, and classification of glaucoma, both through the automated processing of large data sets, and by earlier detection of new disease patterns. In addition, artificial intelligence holds promise for fundamentally changing research aimed at understanding the development, progression, and treatment of glaucoma, by identifying novel risk factors and by evaluating the importance of existing ones.

The Relationship Between Bruch's Membrane Opening-Minimum Rim Width and Retinal Nerve Fiber Layer Thickness and a New Index Using a Neural Network

Article

Full-text available

Aug 2018

Purpose We evaluate the relationship between Bruch's membrane opening minimum rim width (BMO-MRW) and peripapillary retinal nerve fiber layer thickness (pRNFLT) and develop a new parameter combining BMO-MRW and pRNFLT using a neural network to maximize their compensatory values. Methods A total of 402 subjects were divided into two groups: 273 (validation group) and 129 (neural net training) subjects. Linear quadratic and broken-stick regression models were used to explore the relationship between BMO-MRW and pRNFLT. A multilayer neural network was used to create a combined parameter, and diagnostic performances were compared using area under the receiver operating characteristic curves (AUROCs). Results Regression analyses between BMO-MRW and pRNFLT revealed that the broken-stick model afforded the best fit. Globally, the tipping point was a BMO-MRW of 226.5 μm. BMO-MRW and pRNFLT were correlated significantly with visual field. When differentiating normal from glaucoma subjects, the neural network exhibited the largest AUROC. When differentiating normal from early glaucoma subjects, the overall diagnostic performance decreased, but the neural network still exhibited the largest AUROC. Conclusions The optimal relationship between BMO-MRW and pRNFLT was revealed using the broken-stick model. Considerable BMO-MRW thinning preceded pRNFLT thinning. The neural network significantly improved diagnostic power by combining BMO-MRW and pRNFLT. Translational Relevance A combined index featuring BMO-MRW and pRNFLT data can aid clinical decision-making, particularly when individual parameters yield confusing results. Our neural network effectively combines information from separate parameters.

A structure for characterizing clinical support systems

Chapter

Full-text available

Nov 2017

Mohamed Ibrahim Waly

Diagnostic ability of Humphrey perimetry, Octopus perimetry, and optical coherence tomography for glaucomatous optic neuropathy

Article

Nov 2016

PurposeTo evaluate and compare the diagnostic accuracy of the Humphrey Field Analyzer (HFA), Octopus perimetry, and Cirrus OCT for glaucomatous optic neuropathy.Methods Eighty-eight healthy individuals and 150 open-angle glaucoma patients were consecutive and prospectively selected. Eligibility criteria for the glaucoma group were intraocular pressure ≥21 mm Hg and glaucomatous optic nerve head morphology. All subjects underwent a reliable standard automated perimetry with the HFA and Octopus perimeter, and were imaged with the Cirrus OCT. Receiver-operating characteristic (ROC) curves were plotted for the threshold values and main indices of the HFA and Octopus, the peripapillary retinal nerve fiber layer thicknesses, and the optic nerve head parameters. Sensitivities at 85 and 95% fixed-specificities were also calculated. The best areas under the ROC curves (AUCs) were compared using the DeLong method.ResultsIn the glaucoma group, mean deviation (MD) was -5.42±4.6 dB for HFA and 3.90±3.6 dB for Octopus. The MD of the HFA (0.966; P<0.001), mean sensitivity of the Octopus (0.941; P<0.001), and average cup-to-disc (C/D) ratio measured by the Cirrus OCT (0.958; P<0.001) had the largest AUCs for each test studied. There were no significant differences among them. Sensitivities at 95% fixed-specificity were 82% for pattern standard deviation of the HFA, 81.3% for average C/D ratio of OCT, and 80% for the MD of the Octopus.ConclusionsHFA, Octopus, and Cirrus OCT demonstrated similar diagnostic accuracies for glaucomatous optic neuropathy. Visual field and OCT provide supplementary information and thus these tests are not interchangeable.Eye advance online publication, 11 November 2016; doi:10.1038/eye.2016.251.

The Effect of Optic Disc Center Displacement on Retinal Nerve Fiber Layer Measurement Determined by Spectral Domain Optical Coherence Tomography

Article

Full-text available

Oct 2016
PLOS ONE

Purpose To investigate the effect of optic disc center displacement on retinal nerve fiber layer (RNFL) measurement determined by spectral domain optical coherence tomography (SD-OCT). Methods The optic disc center was manipulated at 1-pixel intervals in horizontal, vertical, and diagonal directions. According to the manipulated optic disc center location, the RNFL thickness data were resampled: (1) at a 3.46-mm diameter circle; and (2) between a 2.5-mm diameter circle and 5.4-mm square. Error was calculated between the original and resampled RNFL measurements. The tolerable error threshold of the optic disc center displacement was determined by considering test-retest variability of SD-OCT. The unreliable zone was defined as an area with 10% or more variability. Results The maximum tolerable error thresholds of optic disc center displacement on the RNFL thickness map were distributed from 0.042 to 0.09 mm in 8 directions. The threshold shape was vertically elongated. Clinically important unreliable zones were located: (1) at superior and inferior region in the vertical displacement; (2) at inferotemporal region in the horizontal displacement, and (3) at superotemporal or inferotemporal region in the diagonal displacement. The unreliable zone pattern and threshold limit varied according to the direction of optic disc displacement. Conclusions Optic disc center displacement had a considerable impact on whole RNFL thickness measurements. Understanding the effect of optic disc center displacement could contribute to reliable RNFL measurements.

Senzitivnost parametara konfokalne skening laser oftalmoskopije kod pacijenata sa glaukomom u odnosu na zdravu populaciju

Article

Full-text available

Dec 2014

Uvod. Lokalizovani defekti retinalnog sloja nervnih vlakana (retinal nervefibre layer, RNFL) najčešće se javljaju u ranom glaukomu sa učestalošću oko20%. Heidelberg Retina Tomograf (HRT 3) je konfokalna skening laser oftalmoskopijadizajnirana kako bi olakšala objektivnu i kvantitativnu procjenupapile i RNFL. Cilj rada je bio da ispitamo dijagnostičku mogućnost HRT 3da razdvoji zdrave od očiju sa glaukomom i sa prisutnim lokalizovanim defektimaRNFL i, da ustanovimo kako je stadijum oboljenja uticao na njegovudijagnostičku sposobnost.Metode. Statistički su analizirani podaci dobijeni pregledom HRT 3 i red-freefotografijom fundusa 12 očiju sa glaukomom sa prisutnim lokalizovanimdefektima RNFL i 14 očiju zdravih ispitanika kontrolne grupe.Rezultati. Mjera oblika ekskavacije (Cup Shape Measure, CSM) je parametarkoji je pokazao najveću dijagnostičku preciznost kako globalno (AUC:0,88) tako i u sektorima javljanja lokalizovanih defekata RNFL (AUC: 0,85),odnosno temporo-superiorno i temporo-inferiorno. U ranom i umjerenomstadiju bolesti, skor vjerovatnoće glaukoma (Glaucoma Probability Score, GPS)se pokazao kao sposobniji od Moorfields-ove analize regresije (MoorfieldsRegression Analysis, MRA) da razdvoji oči sa glaukomom i lokalizovanimdefektima RNFL od zdravih i to sa relativno visokom senzitivnošću (83% i78%) u sektorima u kojima su bili prisutni lokalizovani defekti RNFL verifikovanired-free fotografijom fundusa. Analiza RNFL prikazana na izvještajuHRT 3 kao grafikon RNFL profila se pokazala kao statistički značajna daukaže na prisustvo lokalizovanih defekata u odnosu na red-free fotografijupapile i RNFL.Zaključak. Konfokalna skening laser oftalmoskopija primijenjena kod pacijenatasa ranim i umjerenim stadijumom glaukoma i prisutnim lokalizovanimdefektima RNFL, pokazala se kao pouzdana za otkrivanje morfoloških promjena,a kao najbolji indikatori su se pokazali CSM i GPS.

Retinal Nerve Fiber Layer Thickness Measurement Comparison Using Spectral Domain and Swept Source Optical Coherence Tomography

Article

Full-text available

Apr 2016

Purpose: To investigate the retinal nerve fiber layer (RNFL) thickness concordance when measured by spectral domain (SD) and swept source (SS) optical coherence tomography (OCT), and to compare glaucoma-discriminating capability. Methods: RNFL thicknesses were measured with the scan circle, centered on the optic nerve head, in 55 healthy, 41 glaucoma suspected, and 87 glaucomatous eyes. The RNFL thickness measured by the SD-OCT (sdRNFL thickness) and SS-OCT (ssRNFL thickness) were compared using the t-test. Bland-Altman analysis was performed to examine their agreement. We compared areas under the receiver operating characteristics curve and examined sdRNFL and ssRNFL thickness for discriminating glaucomatous eyes from healthy eyes, and from glaucoma suspect eyes. Results: The average ssRNFL thickness was significantly greater than sdRNFL thickness in healthy (110.0 ± 7.9 vs. 100.1 ± 6.8 µm, p < 0.001), glaucoma suspect (96.8 ± 9.3 vs. 89.6 ± 7.9 µm, p < 0.001), and glaucomatous eyes (74.3 ± 14.2 vs. 69.1 ± 12.4 µm, p = 0.011). Bland-Altman analysis showed that there was a tendency for the difference between ssRNFL and sdRNFL to increase in eyes with thicker RNFL. The area under the curves of the average sdRNFL and ssRNFL thickness for discriminating glaucomatous eyes from healthy eyes (0.984 vs. 0.986, p = 0.491) and glaucoma suspect eyes (0.936 vs. 0.918, p = 0.132) were comparable. Conclusions: There was a tendency for ssRNFL thickness to increase, compared with sdRNFL thickness, in eyes with thicker RNFL. The ssRNFL thickness had comparable diagnostic capability compared with sdRNFL thickness for discriminating glaucomatous eyes from healthy eyes and glaucoma suspect eyes.

Association Between Alzheimer's Disease and Glaucoma: A Study Based on Heidelberg Retinal Tomography and Frequency Doubling Technology Perimetry

Article

Full-text available

Dec 2015

Aim: To assess the frequency of glaucoma-like alterations in Alzheimer's disease (AD) patients using Heidelberg Retinal Tomograph III (HRT-3) and Frequency Doubling Technology (FDT) perimetry. Methods: The study included 51 eyes of 51 AD subjects and 67 eyes of 67 age- and sex-matched controls. Subjects underwent an ophthalmological examination including measurements of intraocular pressure (IOP), Matrix FDT visual field testing, optic nerve head morphology and retinal nerve fiber layer thickness (RNFLt) assessment by slit-lamp biomicroscopy and HRT-3. Results: The frequency of alterations was significantly higher in the AD group (27.5 vs. 7.5%; p = 0.003; OR = 4.69). AD patients showed lower IOP (p = 0.000) despite not significantly different values of central corneal thickness (CCT) between the groups (p = 0.336). Of all the stereometric parameters measured by HRT-3, RNFLt was significantly lower in AD patients (p = 0.013). This group also had significantly worse results in terms of Moorfields Regression Analysis (p = 0.027). Matrix showed significantly worse Mean Deviation (MD) (p = 0.000) and Pattern Standard Deviation (PSD) (p = 0.000) values and more altered Glaucoma Hemifield Test (p = 0.006) in AD patients. Pearson's R correlation test showed that Mini Mental State Examination is directly correlated with MD (R = 0.349; p = 0.034) and inversely correlated with PSD (R = −0.357; p = 0.030). Conclusion: Patients with AD have a higher frequency of glaucoma-like alterations, as detected by the use of HRT-3. These alterations were not associated with elevated IOP or abnormal CCT values.

Evaluation of Agreement between HRT III and iVue OCT in Glaucoma and Ocular Hypertension Patients

Article

Full-text available

Jan 2015

Purpose . To determine the agreement between Moorfields Regression Analysis (MRA), Glaucoma Probability Score (GPS) of Heidelberg retinal tomograph (HRT III), and peripapillary nerve fibers thickness by iVue Optical Coherence Tomography (OCT). Methods . 72 eyes with ocular hypertension or primary open angle glaucoma (POAG) were included in the study: 54 eyes had normal visual fields (VF) and 18 had VF damage. All subjects performed achromatic 30° VF by Octopus Program G1X dynamic strategy and were imaged with HRT III and iVue OCT. Sectorial and global MRA, GPS, and OCT parameters were used for the analysis. Kappa statistic was used to assess the agreement between methods. Results . A significant agreement between iVue OCT and GPS for the inferotemporal quadrant ( κ : 0.555) was found in patients with abnormal VF. A good overall agreement between GPS and MRA was found in all the eyes tested ( κ : 0.511). A good agreement between iVue OCT and MRA was shown in the superonasal ( κ : 0.656) and nasal ( κ : 0.627) quadrants followed by the superotemporal ( κ : 0.602) and inferotemporal ( κ : 0.586) sectors in all the studied eyes. Conclusion . The highest percentages of agreement were found per quadrant of the MRA and the iVue OCT confirming that in glaucoma damage starts from the temporal hemiretina.

HRT for the Diagnosis and Detection of Glaucoma Progression

Article

Full-text available

May 2015
Open Ophthalmol J

Confocal scanning laser ophthalmoscopy through the Heidelberg Retina Tomograph (HRT) provides a rapid, safe, noncontact, and noninvasive imaging of the optic disc in three-dimensions, and provides precise detailed information about the optic disc beyond that which the clinical exam can measure. The HRT I was developed for research purposes only and was not used clinically. The HRT II was developed to be user-friendly, more rapid, and was used as an adjunct to clinical examination in the detection and progression of glaucoma. One of the main pitfalls of the HRT II was that it was operator-dependent. The HRT III was developed to be operator-independent. Initially the Moorsfield Regression Analysis provided the analysis of the stereometric optic disc parameters. The Glaucoma Probability Score, given its ease of use, operator-independence, and rapidity of use, soon gained popularity. Numerous studies have compared these two methods of analysis, with the conclusion that the Glaucoma Probability Score provides a higher sensitivity and a lower specificity than the Moorsfield Regression Analysis, which may indicate that it has potential as a screening test for glaucoma. However, there is no consensus on the use of the Glaucoma Probability Score as a screening test for glaucoma. While HRT data may be useful as a clinical adjunct in the screening and diagnosis of glaucoma, it should ultimately only be used to support clinical examination.

Automated Glaucoma Identification Using Retinal Fundus Images: A Hybrid Texture Feature Extraction Paradigm

Chapter

Oct 2014

Glaucoma is one of the most common causes of blindness. Robust mass screening may help to extend the symptom-free life for a®ected patients. To realize mass screening requires a cost-e®ective glaucoma detection method which integrates well with digital medical and administrative processes. To address these requirements, we propose a novel low cost automated glaucoma diagnosis system based on hybrid feature extraction from digital fundus images. The paper discusses a system for the automated identi¯cation of normal and glaucoma classes using higher order spectra (HOS), trace transform (TT), and discrete wavelet transform (DWT) features. The extracted features are fed to a support vector machine (SVM) classi¯er with linear, polynomial order 1, 2, 3 and radial basis function (RBF) in order to select the best kernel for automated decision making. In this work, the SVM classi¯er, with a polynomial order 2 kernel function, was able to identify glaucoma and normal images with an accuracy of 91.67%, and sensitivity and speci¯city of 90% and 93.33%, respectively. Furthermore, we propose a novel integrated index called Glaucoma Risk Index (GRI) which is composed from HOS, TT, and DWT features, to diagnose the unknown class using a single feature. We hope that this GRI will aid clinicians to make a faster glaucoma diagnosis during the mass screening of normal/glaucoma images.

New nonlinear multivariable model shows the relationship between central corneal thickness and HRTII topographic parameters in glaucoma patients

Article

Full-text available

Apr 2009
Clin Ophthalmol

Dimitrios Kourkoutas

Relationship of Optic Disc Topography to Optic Nerve Fiber Number in Glaucoma

Article

Apr 1998
ARCH OPHTHALMOL-CHIC

Yeni H. Yücel

Objective To assess the relationship between in vivo measurements of optic disc topography and histomorphometric measurements of optic nerve fiber number in glaucoma. Methods Both eyes of 10 monkeys (Macaca fascicularis) with laser-induced glaucoma in the right eye were studied. Optic disc topography was measured in vivo with a confocal scanning laser ophthalmoscope. Histomorphometry was performed on optic nerve cross sections using bright-field microscopy with camera lucida. Nerve fiber density was estimated by unbiased random sampling. Nerve fiber number was estimated for each sector by multiplying nerve fiber density with neuroglial area. Nerve fiber count was compared with each of 13 global optic disc topographic parameters. Results For neuroretinal measurements in the glaucomatous eyes, rim area, retinal nerve fiber layer (RNFL) cross-sectional area, rim volume, and RNFL thickness correlated significantly with optic nerve fiber number. Differences in nerve fiber count between control and glaucomatous optic nerves showed the strongest correlation with differences in mean height contour; this was followed by RNFL cross-sectional area, RNFL thickness, rim volume, and differences in rim area. For cup measurements in the glaucomatous eyes, cup volume below reference, cup area, mean cup depth, the ratio of cup area to disc area, and cup shape correlated significantly with nerve fiber number. Differences in nerve fiber number between control and glaucomatous optic nerves showed the strongest correlation with differences in cup shape; this was followed by mean cup depth, cup volume below reference, the ratio of cup area to disc area, cup area, and differences in cup volume below surface. No association was found between optic nerve fiber number and optic disc area in glaucomatous eyes. Conclusions In experimental glaucoma, most optic disc topography measures correlated significantly with optic nerve fiber number. The results of this histomorphometric study support the use of confocal scanning laser ophthalmoscopy to evaluate optic nerve damage in glaucoma.

335 GLAUCOMA SESSION-II Cup Shape Measure (CSM) and its Morphometric and Perimetric Correlates in Primary Open Angle Glaucoma

Article

Full-text available

T he optic disc is the most important clinically visible sight of optic nerve damage in glaucoma. Various parameters of the optic disc, including the optic disc neuroretinal rim, optic disc cup, the cup disc ratio are clinically important in the diagnosis and follow up of glaucomatous damage. (1) The size and configuration of the cup and the neuroretinal rim are assessed in every patient undergoing an ophthalmic evaluation to rule out glaucomatous optic disc damage. The optic disc rim is one of the main sites of glaucoma damage and is known to correlate with the size of the optic disc, and an indicator of the health of the optic nerve. (2,3) In normal eyes there is a strong correlation between the size of the optic cup and optic disc. (4) The Size of the optic disc cup is also important because cup disc ratio is one of the important parameters considered in the diagnosis of glaucoma by the International Society for Geographical and Epidemiological Ophthalmology. (5) Cup shape measure is an CSLO morphometric parameter which is third central moment of the frequency distribution of depth values, relative to the curved surface of all parts in the contour line the steeper the cup walls, the more positive the value (6) , The cup shape measure (CSM) is a unique indicator that the walls of the cup are changing as a result of glaucomatous damage and is considered to discriminate normal from early glaucoma with high precision. (7) It was the purpose of this study to evaluate the Cup Shape Measure and determine its morphometric and perimetric correlates in subjects with Primary Open Angle Glaucoma.

The Relationship Between Retinal Nerve Fiber Layer Thickness and Optic Nerve Head Neuroretinal Rim Tissue in Glaucoma

Article

Full-text available

Sep 2014
INVEST OPHTH VIS SCI

Purpose: The purpose of this study was to determine the relationship between optical coherence tomography (OCT) measures of retinal nerve fiber layer (RNFL) and neuroretinal rim (NRR) in a nonhuman primate experimental glaucoma model, and in a population of clinical patients. Methods: For nonhuman primates, normative data were collected from 44 healthy monkeys, and nine animals with unilateral experimental glaucoma that were followed longitudinally. Cross-sectional human subjects data were collected from 89 healthy, 74 glaucoma suspects, and 104 glaucoma patients. Individualized transverse scaling for OCT scans was calculated using a schematic eye that incorporated optical ocular biometry. Custom algorithms were used to quantify RNFL thickness with and without vessels removed, scaled minimum rim width (sMRW), and neural rim volume (NRV). Results: For the experimental glaucoma group, NRR parameters showed the first changes with increased cumulative IOP. The data for both NRR and RNFL measures were best fit by an exponential rise model (NRV, R2=0.79, P<0.01, sMRW, R2=0.74, P<0.01). The major retinal vascular thickness contribution to the RNFL decreased (0.03 μm/μm, P<0.01) with RNFL loss, but the percent vascular contribution increased (-0.1%/μm, P<0.01) with disease progression. Overall, the findings for the cross-sectional human data were similar to those of the experimental model. Conclusions: The findings illustrate a nonlinear relationship between NRR and RNFL measures and provide support for the use of multiple OCT scaled morphological measures for the diagnosis and management of primary open angle glaucoma in humans.

ANALYSIS OF RETINAL IMAGES IN GLAUCOMA

Article

Full-text available

Andrew J Patterson

Glaucoma Classification Using Brownian Motion and Discrete Wavelet Transform

Article

Aug 2014

Age Associated Changes in the RNFL and Optic Nerve Head.

Article

Full-text available

Jul 2014

Purpose: Optical coherence tomography (OCT) measures of the retinal nerve fiber layer (RNFL) thickness and neuroretinal rim (NRR) parameters are often used as a surrogate for retinal ganglion cell content. The purpose of this study was to investigate the relationship between these morphological measures and the aging effects on these structures. Methods: One hundred thirteen healthy individuals, aged 19 to 76 years, with no prior history of retinal of optic nerve head pathology were recruited. A circumpapillary and radial OCT scan centered on the optic nerve head (ONH) was used for data analysis. Transverse scaling was calculated for each subject using measures from optical biometry. Custom algorithms were used for morphological analysis of the ONH NRR and RNFL that included quantification of major retinal vascular contribution. Results: There was a significant age-related loss of RNFL thickness (-0.23 μm/y, R(2) = 0.24, P < 0.01), major retinal vascular contribution (-0.03 μm/y, R(2) = 0.07, P = 0.01, neural rim volume (NRV, -0.004 mm(3)/y, R(2) = 0.15, P < 0.01), and minimum rim width (MRW, -1.77 μm/y, R(2) = 0.23, P < 0.01) before, and after, incorporating the Bruch's membrane opening size (sMRW, -1.86 μm/y, R(2) = 0.22, P < 0.01). When normalized, the rates of change for ONH NRR parameters (NRV, 0.69%/y and sMRW, 0.50%/y) exceeded that of RNFL thickness (0.19%/y, P < 0.01). Conclusions: Although both RNFL and ONH NRR parameters contain axons of retinal ganglion cells, there are differences in age-related changes in these measures that should be considered in clinical application.

A unique color-coded visualization system with multimodal information fusion and deep learning in a longitudinal study of Alzheimer's disease

Article

Apr 2023
ARTIF INTELL MED

Purpose: Automated diagnosis and prognosis of Alzheimer's Disease remain a challenging problem that machine learning (ML) techniques have attempted to resolve in the last decade. This study introduces a first-of-its-kind color-coded visualization mechanism driven by an integrated ML model to predict disease trajectory in a 2-year longitudinal study. The main aim of this study is to help capture visually in 2D and 3D renderings the diagnosis and prognosis of AD, therefore augmenting our understanding of the processes of multiclass classification and regression analysis. Method: The proposed method, Machine Learning for Visualizing AD (ML4VisAD), is designed to predict disease progression through a visual output. This newly developed model takes baseline measurements as input to generate a color-coded visual image that reflects disease progression at different time points. The architecture of the network relies on convolutional neural networks. With 1123 subjects selected from the ADNI QT-PAD dataset, we use a 10-fold cross-validation process to evaluate the method. Multimodal inputs* include neuroimaging data (MRI, PET), neuropsychological test scores (excluding MMSE, CDR-SB, and ADAS to avoid bias), cerebrospinal fluid (CSF) biomarkers with measures of amyloid beta (ABETA), phosphorylated tau protein (PTAU), total tau protein (TAU), and risk factors that include age, gender, years of education, and ApoE4 gene. Findings/results: Based on subjective scores reached by three raters, the results showed an accuracy of 0.82 ± 0.03 for a 3-way classification and 0.68 ± 0.05 for a 5-way classification. The visual renderings were generated in 0.08 msec for a 23 × 23 output image and in 0.17 ms for a 45 × 45 output image. Through visualization, this study (1) demonstrates that the ML visual output augments the prospects for a more accurate diagnosis and (2) highlights why multiclass classification and regression analysis are incredibly challenging. An online survey was conducted to gauge this visualization platform's merits and obtain valuable feedback from users. All implementation codes are shared online on GitHub. Conclusion: This approach makes it possible to visualize the many nuances that lead to a specific classification or prediction in the disease trajectory, all in context to multimodal measurements taken at baseline. This ML model can serve as a multiclass classification and prediction model while reinforcing the diagnosis and prognosis capabilities by including a visualization platform.

Comprehensive Glaucoma Imaging

Chapter

Apr 2022

Advancements in ocular imaging over the past four decades have revolutionized glaucoma practice worldwide. As glaucomatous vision loss cannot be restored, early diagnosis, proper management, and timely intervention are essential to slow disease progression. The emergence of advanced computerized imaging modalities has shifted glaucoma assessment from being largely subjective to mostly objective. These modalities can document and accurately quantify the optic nerve head and the macula regions, which are affected by the disease. With various ocular imaging devices evolving over the years, optical coherence tomography (OCT) has become the dominant imaging technology in glaucoma practice. From the first prototype device to the newest swept-source OCT, each generation improved in image acquisition time, scan resolution, and artifact reduction, making structural assessment more accurate and sensitive. This advancement allows clinicians earlier detection of glaucoma diagnosis and increased sensitivity in monitoring of progression enabling timely modification of treatment to halt further damage. Moreover, the introduction of OCT angiography (OCTA) drew attention to the vascular component as an important element in glaucoma pathogenesis. In addition to technical advancements, the imaging field continues to evolve by the incorporation of innovative software such as image processing and artificial intelligence. Taken together, the role of ocular imaging is expected to further expand with a substantial impact on clinical management.

AI and Glaucoma

Chapter

Oct 2021

Recent artificial intelligence advances in ophthalmology, especially deep learning (DL), have shown enormous potential with vast quantities of data. Unlike diabetic retinopathy and age-related macular degeneration where early DL initiatives in ophthalmology focused on, there have been limited but expanding efforts for utilizing DL algorithms to improve diagnosis and management of glaucoma. In this chapter, we provide a summary of current AI applications, challenges, and state-of-the-art DL systems in glaucoma. Key applications involve glaucoma diagnosis, longitudinal progression analysis, structural-functional correlation investigation, scan enhancement, and new knowledge discovery. Given the multifactorial etiology of glaucoma, various input modalities such as fundus photographs, optical coherence tomography (OCT), visual field testing, and demographic data have been investigated along with DL models. The ability of extracting meaningful representations from high dimensional and complex multi-modal data enables DL system to achieve high accuracy of glaucoma diagnosis and prognosis and to discover new knowledges to improve our current understanding of glaucoma. Though DL algorithms have shown promising improvements in performing tasks related to glaucoma, lack of large dataset, diversity of data formations and evaluation criteria, and poor model interpretability still remain as challenges to the research community.

Primary open angle glaucoma case finding in a UK community

Conference Paper

May 2020

Michelle Pui Yan Chan

Glaucoma is a leading cause of irreversible blindness in the UK, and generates the sixth largest share of NHS outpatient attendances. There are inadequacies in the current system of primary open angle glaucoma (POAG) case-finding and its referral to the hospital eye service (HES). Half the cases are undiagnosed, while a large proportion of referrals for raised intraocular pressure (IOP) (>21mmHg) do not have glaucoma. No diagnostic test exists with both sufficient sensitivity and specificity suitable for population-wide glaucoma screening. The EPIC-Norfolk Eye Study is a cross-sectional ophthalmic survey. This thesis aims to explore the characteristics of POAG, the risk factors for undiagnosed POAG, to re-examine the IOP referral threshold, and to evaluate the diagnostic performances of optic disc imaging in population screening. Among the 8623 participants, 4.2% had glaucoma and 3.7% had POAG. A large number were glaucoma suspects (7%) and ocular hypertensives (10%). POAG was strongly associated with higher corneal compensated IOP (IOPcc) and lower corneal hysteresis, which could become useful metrics in disease management. 34% of POAG cases were previously undiagnosed, the risk factors were having normal tension rather than high tension glaucoma, lower pre-treatment IOP and smaller cup/disc ratio. This suggests an over-reliance on IOP in diagnosis and the need to re-focus on disc assessment. The upper limit of IOP distribution (mean+2 standard deviations) in the study was 23.6mmHg. Increasing the referral threshold to >24mmHg could cut HES referrals by up to 67%. IOP used alone is ineffective in diagnosing POAG. HRTII and GDx-VCC used in combination generated high sensitivities and specificities in glaucoma screening, but would be costly and impractical to implement. The individual normative values of HRTII and GDx-VCC generated low sensitivities but high specificities (>97.0%), and could be useful in excluding glaucoma on a population scale. These findings can help transform glaucoma care in the UK.

Artificial intelligence and deep learning in glaucoma: Current state and future prospects

Chapter

Aug 2020
Progr Brain Res

Over the past few years, there has been an unprecedented and tremendous excitement for artificial intelligence (AI) research in the field of Ophthalmology; this has naturally been translated to glaucoma—a progressive optic neuropathy characterized by retinal ganglion cell axon loss and associated visual field defects. In this review, we aim to discuss how AI may have a unique opportunity to tackle the many challenges faced in the glaucoma clinic. This is because glaucoma remains poorly understood with difficulties in providing early diagnosis and prognosis accurately and in a timely fashion. In the short term, AI could also become a game changer by paving the way for the first cost-effective glaucoma screening campaigns. While there are undeniable technical and clinical challenges ahead, and more so than for other ophthalmic disorders whereby AI is already booming, we strongly believe that glaucoma specialists should embrace AI as a companion to their practice. Finally, this review will also remind ourselves that glaucoma is a complex group of disorders with a multitude of physiological manifestations that cannot yet be observed clinically. AI in glaucoma is here to stay, but it will not be the only tool to solve glaucoma.

In vivo imaging of the lamina cribrosa

Thesis

Jan 2003

Aachal Kotecha

This thesis describes the development of techniques to image the lamina cribrosa within the optic nerve head of the living human eye using a confocal scanning laser ophthalmoscope. The axons from the retinal ganglion cells pass through the pores of the lamina cribrosa to synapse in the higher levels of the brain. The lamina cribrosa has been implicated as one of the main sites of damage to the ganglion cell axons in one of the leading causes of blindness, glaucoma, where it is thought that the raised intraocular pressure results in a backward bowing of the lamina, with consequent deformation of the ganglion cell axons. Most of our understanding of this process to date has come from investigations in post-mortem tissue, since the structure is difficult to visualise due to the optical scattering properties of the overlying neural tissue. The development of new imaging technologies has enabled the investigation of the optic nerve in vivo to quantify its structure in the living human eye. These make use of the optical sectioning effects and improved depth resolution of confocal laser scanning ophthalmoscopy. The purpose of this study was to use these advances to image the lamina cribrosa in vivo, and determine whether using these techniques its structure could be reasonable quantified. Two confocal scanning laser ophthalmoscope systems were used to obtain images of the lamina cribrosa area. These instruments used different lasers, and had different depth and spatial resolution, and were compared for image quality. A comparison was made between imaging with a near infrared laser and the helium neon laser, and it was predicted that the former might give better pore visibility due to greater tissue penetration. However, it was found that using the near infrared laser gave no benefit to lamina pore visibility. It was also found that lamina cribrosa images were best obtained with one of the two instruments- the Zeiss cSLO. Images obtained by these two instruments were digitised, and these digitised in vivo images of the lamina cribrosa underwent a series of image processing techniques to maximise the visibility of the pore-like structure. A method was developed which incorporated the use of Fourier analysis to select the optimum spatial frequency components to reduce the effects of image degradation. The result was a method that, for the first time, provided adequate image quality that allowed quantification of the morphology of the pores of the lamina cribrosa and without operator intervention. This has the advantage of removing possible sources of error due to the operator. In addition, unlike earlier studies, an unselected series of eyes were imaged without pre-selection for pore visibility. A further study was performed to determine the relationship between the lamina pore characteristics and the optic nerve head topography using three dimensional reconstruction, to determine whether any differences existed between pore characteristics of normal and glaucomatous eyes. It was found that with increasing acquired neuroretinal rim loss, the average pore area and pore number increased. This agreed with post-mortem studies.Histological studies were performed on hydrated post-mortem tissue in a range of aged eyes. The purpose of this was to establish the longitudinal thickness of the lamina cribrosa in order to evaluate the imaging depth of the cSLO. This study found that the longitudinal thickness of the lamina cribrosa was approximately two thirds larger than that previously found. This was attributed to the fact that most previous studies used dehydration and digestion techniques during tissue preparation. This work has shown that it is possible to image the lamina cribrosa in vivo and obtain quantitative information about the pore structure without operator intervention in unselected eyes. An image processing routine was developed to highlight the lamina pore structure using a technique based on Fourier analysis.

Rapid Assessment of Confocal Scan Technology

Chapter

Full-text available

Apr 2013

This is a rapid health technology assessment of confocal scan technology for glaucoma diagnosis.

Optic nerve head and retinal nerve fiber layer analysis1

Article

Jul 1999
OPHTHALMOLOGY

American Academy of Ophthalmology

The purpose of the Committee on Ophthalmic Procedures Assessment is to evaluate on a scientific basis new and existing ophthalmic tests, devices, and procedures for their safety, efficacy, clinical effectiveness, and appropriate uses. Evaluations include examination of available literature, epidemiological analyses when appropriate, and compilation of opinions from recognized experts and other interested parties. After appropriate review by all contributors, including legal counsel, assessments are submitted to the Academy’s Board of Trustees for consideration as official Academy policy.

A local excitation and measurement approach for decentralized damage detection using transmissibility functions

Article

Aug 2015
STRUCT CONTROL HLTH

In order to assess structural safety conditions, many vibration-based damage detection methods have been developed in recent years. Among these methods, transmissibility function analysis can offer advantages in easy and low-cost implementation, as well as independency to the magnitude and waveform of an excitation record. Harnessing these features, a decentralized structural damage detection procedure is proposed in this paper. The proposed procedure only requires measurements in one small area at a time, and thus, is ideal for using limited number of sensors upon a large-scale structure. This study also investigates the nature of transmissibility functions for damage detection in an analytical manner, based on a chain-like spring–mass–damper system with multiple degrees-of-freedom. The analytical derivation is validated through numerical simulation and laboratory experiments using mobile sensors. Copyright © 2015 John Wiley & Sons, Ltd.

Glaucoma Disease: A Survey

Article

Jul 2015

Glaucoma is a neurodegenerative disease and is considered as one of the most common reasons of blindness. Degeneration of nerves is an irreversible process so early diagnosis of this disease is inevitable to avoid permanent loss of vision. Diagnosis of glaucoma usually consists of inspection of the structural deterioration of nerve in conjunction with the examination of visual function. This article presents a detailed overview of glaucoma, its symptoms and the potential it has to affect the persons. The research work related to its diagnosis and treatment is also reviewed in this article.

Retinal Nerve Fiber Layer Converges More Convexly on Normal Smaller Optic Nerve Head

Article

Apr 2015

To investigate retinal nerve fiber layer (RNFL) configuration in the optic nerve head (ONH) and peripapillary area according to disc size and to determine whether it explains cup discrepancy among eyes with different disc sizes. Horizontal and vertical RNFL curvature and mean thickness were measured using confocal scanning laser ophthalmoscopy (Heidelberg Retina Tomograph) in 63 normal subjects grouped by disc size. Average and quadrant RNFL thickness, disc size, average cup-to-disc ratio (CDR), and convergence angle at the optic disc were also measured using Cirrus HD-optical coherence tomography. The relationships between disc size and RNFL curvature, thickness, angle at optic disc, and CDR were evaluated. RNFL curvature and convergence angle reflects convexity "on" and "into" the optic disc, respectively. CDR was smaller for small discs and was positively correlated with disc size (P<0.001). Horizontal and vertical RNFL curvatures were significantly more convex for small than large discs (P=0.001, 0.017). Horizontal and vertical RNFL angles at the optic disc were positively correlated with disc size (P<0.001, P=0.012). Optic disc area was negatively correlated with mean RNFL thickness at the optic disc margin measured by HRT (P=0.002), but not in the peripapillary area by optical coherence tomography. Using imaging techniques, we demonstrated that the shape of the RNFLs converging "on" and entering "into" the optic disc was more convex for small optic discs compared with large discs. A low CDR for small discs could be mediated by these RNFL profiles at the ONH, which may guide the clinical evaluation of glaucomatous ONH damage.

Recognizing Structural Damage to the Optic Nerve Head and Nerve Fiber Layer in Glaucoma

Article

Oct 1997
AM J OPHTHALMOL

Joseph Capriou

Confocal Scan Laser Ophthalmoscope for Diagnosing Glaucoma: A Systematic Review and Meta-analysis

Article

Jun 2015

This systematic review was performed to estimate the diagnostic accuracy of the confocal scanning laser ophthalmoscopy in diagnosing glaucoma. We did a sensitive electronic search to find relevant studies. Two reviewers independently screened relevant articles and extracted required data about study methods and reported results of sensitivity and specificity. A meta-analysis was conducted for pooling data to compare different editions of the Heidelberg Retina Tomograph (HRT) with one of its alternatives, scanning laser polarimetry (GDx) with the criteria of "visual field defect" and "changes of nerve fiber layer" as the reference standard. We identified 37 evaluations from 28 relevant primary studies. In these studies, 9573 eyes (4883 glaucomatous and 4689 non-glaucomatous) were assessed with regards to the reference standard using one of the HRT editions with or without GDx. Diagnostic odds ratios were 9.35 [95% confidence interval (CI): 7.58-11.53] for HRT, 11.84 (95% CI: 9.97-14.06) for HRT II, 11.86 (95% CI: 9.16-15.35) for HRT III, and 21.33 (95% CI, 13.56-33.55) for GDx. Although GDx was more accurate than HRT, all editions of HRT had acceptable performance in diagnosing glaucomatous eyes with an ophthalmologist's clinical examination as the reference standard.

Klinische Interpretation der Analyse mit dem Heidelberg Retina Tomographen (HRT II)

Article

Alfonso Antón

4 Die Glaukomerkrankung führt zu typischen strukturellen Änderungen im Sehnervenkopf und der retinalen Nervenfaserschicht (RNFS). Die Erkennung und Beschreibung dieser Befunde und die Messung von Änderungen über die Zeit sind fundamentale Bausteine der Glaukomdiagnose und der Verlaufskontrolle. Die Interpretation der Befunde der klinischen Untersuchung des Sehnervenkopfes und der RNFS ist komplex, weil sich auch normale Sehnervenköpfe in Größe und Form beträchtlich voneinander unterscheiden können, und weil auch andere Krankheiten mit diesen assoziiert sein können. Es gehört unzweifelhaft zu den schwierigsten Aufgaben eines Ophthalmologen, das Erscheinungs-bild des Sehnervenkopfes zu bewerten, um daraus die Stabilität oder die Progression der Erkrankung zu erkennen. In der klinischen Praxis wird die Beurteilung dieser Strukturänderungen gewöhnlich mit Hilfe der Fundus-Biomikroskopie und der Beurteilung von Stereofotografien des Seh-nervenkopfes vorgenommen. Derartige akribische Untersuchungen mit der Spaltlampe sind sicherlich sinnvoll, dennoch fehlen ihnen Objektivität und Quantifizierbarkeit. Fotografien des Sehnervs sind objektive Dokumente, die einen Vergleich von Befunden über die Zeit ermöglichen, die aber einer zusätzlichen subjektiven Interpretation oder einer Falschfarben-Computeranalyse bedürfen. Sie sind sehr anfällig gegen Medien-trübung und die Bedingungen während der Bildaufnahme (Exposition, Lichtquelle, digitale Bildverarbeitung), die das Erscheinungsbild der Retinastrukturen deutlich beeinflussen können. Sowohl Fotografien als auch klinische Untersuchungen gehören zur Standard-Diagnostik bei Glaukom-Patienten. Objektive und quantitative Messungen des Sehnervenkopfes und der RNFS können die Beurteilung einer glaukomatösen Schädigung ergänzen und dabei helfen, die Progression der Krankheit festzustellen. Ziel dieses Beitrags ist es, eine systematische Vorgehensweise für die klinische Interpretation der HRT II Ergebnisse aufzuzeigen und den Nutzen der HRT-Technik zum Zeitpunkt der Erstuntersuchung zu diskutieren.

Predictive Value of Heidelberg Retina Tomograph Parameters for the Development of Glaucoma in the European Glaucoma Prevention Study

Article

Nov 2014
AM J OPHTHALMOL

Purpose: To determine whether baseline Heidelberg Retina Tomograph (HRT) measurements of the optic disc are associated with the development of open-angle glaucoma (OAG) in individuals with ocular hypertension in the European Glaucoma Prevention Study (EGPS). Design: Retrospective analysis of a prospective, randomized, multicenter, double-masked, controlled clinical trial. Methods: There were 489 participants in the HRT Ancillary Study to the EGPS. Each baseline HRT parameter was assessed in univariate and multivariate proportional hazards models to determine its association with the development of OAG. Proportional hazards models were used to identify HRT variables that predicted which participants in the EGPS had developed OAG. Development of OAG was based on visual field and/or optic disc changes. Results: At a median follow-up time of about 5 years, 61 participants developed OAG. In multivariate analyses, adjusting for randomization arm, age, baseline IOP, central corneal thickness, pattern standard deviation, and HRT disc area, the following HRT parameters were associated with the development of OAG: the "outside normal limits" classification of the Frederick Mikelberg (FSM) discriminant function (hazard ratio [HR] 2.51, 95% confidence interval [CI]: 1.45-4.35), larger mean cup depth (HR 1.64, 95% CI: 1.21-2.23), cup-to-disc area ratio (HR 1.43, 95% CI: 1.14-1.80), linear cup-to-disc ratio (HR 1.43, 95% CI: 1.13-1.80), cup area (HR 1.33, 95% CI: 1.08-1.64), smaller rim area (HR 1.33, 95% CI: 1.07-1.64), larger cup volume (HR 1.30, 95% CI: 1.05-1.61), smaller rim volume (HR 1.25, 95% CI: 1.01-1.54), larger maximum cup depth (HR 1.18, 95% CI: 1.01-1.36), and cup shape measure (HR 1.18, 95% CI: 1.01-1.36). Conclusions: Several baseline HRT parameters, alone or in combination with baseline clinical and demographic factors, were significantly associated with the development of OAG among the EGPS participants.

Photogrammetry of the optic disc in glaucoma and ocular hypertension with simultaneous stereo photography

Article

Full-text available

Jan 1980
INVEST OPHTH VIS SCI

Stereophotogrammetric evaluations of the optic cup were performed for normal, ocular hypertensive, and glaucomatous eyes. Average volume, area, and depth measurements were progressively larger from normal to ocular hypertensive to glaucomatous eyes, although the distributions of individual values exhibited considerable overlap among the three groups. Similar results were obtained for volume, area, and depth asymmetry between each pair of eyes. None of these measurements was able to distinguish accurately between normal and glaucomatous optic cups. However, normal eyes showed a high correlation (r = +0.85) between area and depth of the optic cup, whereas this area/depth relationship was reduced in ocular hypertensives (r = +0.63) and completely broke down for glaucomatous eyes (r = +0.04). Approximately 89% of the glaucomatous eyes and 47% of the ocular hypertensive eyes were beyond the range of normal area/depth correlation values. These findings represent an improvement over most previous attempts to quantitatively differentiate between normal and glaucomatous eyes on the basis of optic disc measurements alone, and support the hypothesis that optic disc damage usually precedes visual field loss in glaucoma. With further technical refinements such as computer image processing, stereophotogrammetry of the optic cup may become a valuable differential diagnostic technique for glaucoma.

Optic Disc Rim Area Is Related to Disc Size in Normal Subjects

Article

Full-text available

Jul 1988
ARCH OPHTHALMOL-CHIC

In Reply. —We thank Drs Hong and Shin for their observations. The equation for the regression line that relates rim area to disc area was indeed given incorrectly and should be as follows: Rim area = 0.58 × disc area + 0.15 mm2. The data shown in Fig 3, as well as the correlation coefficient and associated P value for the linear regression, remain unchanged. We regret this error.The depth determinations made by the Rodenstock analyzer are completely objective, with no operator input. We have found that in the vast majority of cases, small changes in the placement of the disc margin by the operator have no effect on the calculated cup volume. There may be some software differences between Drs Hong and Shin's apparatus and ours that account for the apparent discrepancy.We could not agree more with the last comment made by Drs Hong and Shin.

Performance of Screening and Diagnostic Tests

Article

Jun 1987
ARCH GEN PSYCHIAT

Jane M. Murphy

• Evaluation of psychiatric screening and diagnostic tests has benefited from the application of sensitivity, specificity, the k-statistic, and predictive values. These measures derive their meaning from a single criterion threshold. Receiver operating characteristic (ROC) analysis extends assessment of test performance by providing information about all possible pairs of achievable sensitivity and specificity values. The ROC analysis offers a comprehensive means for comparing different tests or different scoring procedures for one test. As a demonstration we used the ROC analysis to evaluate three types of scoring rules for one psychiatric test, the Health Opinion Survey. The demonstration indicated that ROC analysis can profitably take a place among the standard methods for test evaluation in psychiatric research. In addition, ROC analysis can assist clinicians in selecting appropriate test procedures for particular patient populations.

The management of elevated intraocular pressure with normal optic discs and visual fields

Article

May 1977
SURV OPHTHALMOL

H. Dunbar Hoskins Jr

Variability of expert observers in evaluating the optic disc

Article

Feb 1976
Trans Am Ophthalmol Soc

Paul Lichter

A significant amount of variability exists between observers in designating cup/diac (C/D) ratios. Further, different methods are used for evaluating the size of the cup. On method uses a combination of contour and color, the other specifically measures contour and pallor separately. This study confirms that these methods yeild different numerical results. Interestingly, the "cupping/pallor" observers show no significant difference among one another in C/D determination which may support the notion that cupping observations are more consistent than "standard" C/D ratios. Though both groups feel the cup is larger on stereo compared to monocular viewing, the "cupping/pallor" group demonstrated much greater differences in this regard. However, the evaluation of pallor by the latter group corresponded closely to the C/D ratios of the "standard" group both for nonstereo and stereo. An individual observer is reasonably consistent on repeat evaluation, but at times, inconsistency is quite substantial. These inconsistencies lead to the conclusion that C/D ratios are an inexact method of recording the status of a disc. Except for considerable changes over time, this numerical method is probably not reliable in checking for small disc changes. However, the disc alone can provide clues as to whether it is physiologic or pathologic. Nevertheless, even with expert observers, significant variability exists in interpretation of an optic disc in this regard. Individuals who evaluate the disc cup by cupping and pallor tend to call discs more pathologic than other observers. The clinical implications of this study suggest that certain observers are more accurate than others and certain discs are more easily evaluated than others. Yet, no one method seems foolproof and no specific criteria can as yet distinguish a normal from an abnormal disc. Stereo color transparencies are evaluated more accurately than are nonstereo color prints. Vessel detail and pallor patterns are probably the most useful disc characteristics to observe. In this regard, a C/D ratio does nothing to indicate whether a disc is normal or not. A statement as to the observer's opinion should be made for each disc evaluation in addition to recording the appearance of the disc.

The management of elevated intraocular pressure with normal optic discs and visual fields. II. An approach to early therapy

Article

May 1977
SURV OPHTHALMOL

H D Hoskins

New techniques for the examination of the optic disc and their clinical application

Article

Mar 1976

B Schwartz

Photogrammetric Analysis of Volume Asymmetry of the Optic Nerve Head Cup in Normal, Hypertensive, and Glaucomatous Eyes

Article

Aug 1975
AM J OPHTHALMOL

Gerald L. Portney

Recent studies of optic cup volume, depth, and orifice area demonstrated a large overlap of the range of these factors for normal and glaucomatous eyes, suggesting that it is usually not possible to distinguish abnormal disks on the basis of a particualr set of geometric quantifiers. Because of the knowledge that qualitative asymmetry of a person's optic cups may be a sign of glaucoma, calculation of the quantity of cup asymmetry using the aforementioned geometric parameters was undertaken in 90 patients. Results indicated that 90% of the bilaterally normal subjects who were studied and 80% of the ocular hypertensives had less than 0.20 mm3 cup volume asymmetry. In contrast, however, all of the unilateral glaucoma patients in this sample exceeded that value. It therefore appears that measurement of the amount of, volume asymmetry between a person's optic cups may be a useful tool in the diagnosis of glaucoma.

How Neural Networks Learn From Experience

Article

Oct 1992

Geoffrey E. Hinton

Networks of artificial neurons modeled on conventional computers are helping explain the ability of the brain to process and retain information. These neural-network simulations have already ruled out many theories. They are now beginning to reveal how the brain accomplishes the remarkable feat of learning.

Relationship between intraocular pressure and primary open angle glaucoma among white and black Americans. The Baltimore Eye Survey

Article

Sep 1991
ARCH OPHTHALMOL-CHIC

A detailed ocular examination, including perimetry, was conducted on 5308 black and white subjects aged 40 years and older in a population-based prevalence survey in east Baltimore, Md. Repeated, detailed examinations were carried out on selected subjects. Roughly half of all subjects with optic nerve damage from primary open angle glaucoma, regardless of race, were unaware that they had the condition. The average intraocular pressure (IOP) among black patients with glaucoma who were receiving treatment was virtually identical to that in those black patients who were not receiving treatment (median IOP, 20 mm Hg); treated eyes of white patients had a lower IOP than those eyes of white patients who were not receiving treatment (mean [+/- SD] IOP, 18.69 +/- 3.23 mm Hg vs 24.15 +/- 5.23 mm Hg; P less than .001). The risk of glaucomatous optic nerve damage increased with the height of the screening IOP, particularly at levels of 22 to 29 and 30 mm Hg and above (relative rate compared with IOP of 15 mm Hg or lower, 12.8 and 40.1 mm Hg, respectively). More than half of all glaucomatous eyes had a screening IOP below 21 mm Hg, whether these eyes were receiving treatment or not. The IOP in glaucomatous eyes tended to rise on follow-up, in contrast with nonglaucomatous eyes in which the IOP was as likely to rise as to fall. Results confirmed that IOP is an important factor in glaucoma, but did not support the traditional distinction between "normal" and "elevated" pressure, nor its corollaries, "low-tension" glaucoma and "high-tension" glaucoma.

Clinically detectable nerve fiber atrophy precedes the onset of glaucomatous field loss

Article

Feb 1991
ARCH OPHTHALMOL-CHIC

Standardized perimetry and nerve fiber layer and color fundus photography were performed annually on 1344 eyes with elevated intraocular pressures. In 83 eyes, glaucomatous field defects developed that met rigid criteria on manual kinetic and suprathreshold static perimetry. Individual nerve fiber layer photographs were read by two masked observers. The more sensitive of the two identified nerve fiber layer defects in 88% of readable photographs at the time field loss first occurred; 60% (6/10) of eyes already had nerve fiber layer defects 6 years before field loss. In contrast, the nerve fiber layer was considered abnormal in only 11% (3/27) of normal eyes and 26% (84/327) of hypertensive eyes. The location of nerve fiber layer and field defects closely corresponded, but nerve fiber layer loss was generally more widespread. Examiner experience and severity of optic nerve damage influenced results. Mild focal defects were more readily recognized than more severe diffuse atrophy. Nerve fiber layer defects expanded with time, often by the development and coalescence of adjacent areas of damage.

Reproducibility of Topographic Measurements of the Normal and Glaucomatous Optic Nerve Head With the Laser Tomographic Scanner

Article

Mar 1991
AM J OPHTHALMOL

We acquired five independent topographic images of the optic nerve head of eight normal eyes and eight eyes with primary open-angle glaucoma with a laser tomographic scanner. Each image had a field of view of 15 x 15 degrees with a resolution of 256 x 256 pixels. The pixel size was approximately 15 x 15 microns. The value of a pixel of a topographic image represented the height at this position. The mean height and the standard deviation over the five topographic images were calculated for each of the 65,536 pixel positions. The standard deviation of a single height measurement in normal eyes was 38.7 microns (range, 23.4 to 62.2 microns) for areas in the peripapillary retina and 42.6 microns (range, 24.4 to 53.7 microns) for measurements within the optic nerve head area. In glaucomatous eyes, the standard deviation was 41.2 microns (range, 23.2 to 59.6 microns) in the peripapillary retina and 49.4 microns (range, 28.1 to 72.8 microns) within the optic nerve head. There was no significant difference between the standard deviation of a single height measurement in normal and glaucomatous eyes (P = .34 within the optic nerve head area; P = .57 on peripapillary retina). No correlation was found between standard deviation of the measurements and pupil size or age of the subject.

The Contour of the Juxtapapillary Nerve Fiber Layer in Glaucoma

Article

Apr 1990
OPHTHALMOLOGY

Joseph Caprioli

Reliable structural markers for early glaucomatous optic nerve damage would facilitate the diagnosis of glaucoma at an early stage, possibly before visual field loss occurs. Computerized image analyses were used to develop and analyze new structural parameters for glaucomatous optic nerve damage. Multiple measurements of relative juxtapapillary nerve fiber layer height were made in glaucoma patients (n = 112), patients suspected of having glaucoma (n = 87), and in age-matched normal control subjects (n = 53). The average relative nerve fiber layer (NFL) height differed in glaucoma patients and normal subjects by 70 microns, but differences exceeded 100 microns at the superior and inferior poles of the disc. Mean values for "glaucoma suspects" were intermediate between those for the normal subjects and glaucoma groups. The ability of summary statistics of relative NFL height measurements to discriminate between normal and glaucomatous eyes was superior to that of the standard disc parameters cup-disc ratio, disc rim area, and cup volume. Measurements of relative NFL height correlated with indices of visual field loss; the strongest correlations occurred for measurements at the superior and inferior poles. Measurements of juxtapapillary NFL height may prove useful to detect glaucomatous optic nerve damage at an early stage and to accurately recognize progressive nerve damage over time.

Nonparametric comparison of entire ROC curves for computerized ECG left ventricular hypertrophy algorithms using data from the Framingham Heart Study

Article

Feb 1989
J ELECTROCARDIOL

A computer program may be capable of several different statements for left ventricular hypertrophy (eg, possible LVH, probable LVH, consistent with LVH), but such statements resulting from discretized levels of sensitivity/specificity would represent only isolated points on a receiver-operating characteristic (ROC) curve, which is a plot of all levels of sensitivity versus specificity. Even if two algorithms use the same discrete scales, their performances may not readily be compared. The authors present a comparison methodology for ROC curves using ROC area as a nonparametric measure of the ability of the algorithm to separate the two populations; the ROC area ranges from 0.5 (no ability) to 1.0 (perfect separation) and is unbiased if the normal versus abnormal populations have no common values for the measurement. The methodology compares the performance of ECG algorithms on the same population of cases by testing for significant differences of ROC areas and incorporating correlation of the algorithms in a nonparametric way. To illustrate this methodology, they use ECG and echocardiographic data from the Framingham Heart Study.

Comparing the Areas under Two or More Correlated Receiver Operating Characteristic Curves: A Nonparametric Approach

Article

Oct 1988

Methods of evaluating and comparing the performance of diagnostic tests are of increasing importance as new tests are developed and marketed. When a test is based on an observed variable that lies on a continuous or graded scale, an assessment of the overall value of the test can be made through the use of a receiver operating characteristic (ROC) curve. The curve is constructed by varying the cutpoint used to determine which values of the observed variable will be considered abnormal and then plotting the resulting sensitivities against the corresponding false positive rates. When two or more empirical curves are constructed based on tests performed on the same individuals, statistical analysis on differences between curves must take into account the correlated nature of the data. This paper presents a nonparametric approach to the analysis of areas under correlated ROC curves, by using the theory on generalized U-statistics to generate an estimated covariance matrix.

Reproducibility of Topographic Measurements with the Optic Nerve Head Analyzer

Article

Jan 1988
AM J OPHTHALMOL

We evaluated the reproducibility of optic nerve head contour mapping by the Optic Nerve Head Analyzer with ten recordings and analyses of one eye each in ten normal volunteers and ten analyses of a single recording in three of the eyes. Reproducibility was good for cup-to-disk ratio and neural rim area. Reproducibility was poor for cup volume, which was inversely related to cup size. Instrument and operator error were low for cup-to-disk ratio and neural rim area, but high for cup volume. Subjective comparisons of contour lines from five recordings of six eyes implied good reproducibility of this measurement.

Reproducibility of Optic Disc Measurements With Computerized Analysis of Stereoscopic Video Images

Article

Aug 1986
ARCH OPHTHALMOL-CHIC

Computerized digital image analysis of the optic nerve head was performed using a simultaneous stereoscopic video camera system to provide rapid, quantitative measurements of optic nerve head topography. Determination of the variability of the measurements is required to estimate the magnitude of optic nerve change over time that can be reliably detected. Total variability, operator variability, and interoperator variability were studied. Total variability was assessed by analyzing ten sets of video recordings made in each of seven eyes of seven normal subjects, and in seven eyes of seven patients with glaucoma. The median coefficients of variation for measurements of total variability in glaucoma patients were as follows: vertical disc measurement, 1.4%; horizontal disc measurement, 2.1%; disc area, 1.9%; vertical cup/disc, 3.9%; horizontal cup/disc, 3.3%; disc rim area, 7.5%; and volume, 7.6%. Operator and interoperator variabilities were considerably smaller in magnitude.

Neuroretinal Rim Area in Early Glaucoma

Article

Feb 1985
AM J OPHTHALMOL

The neuroretinal rim area of the optic nerve head was measured in 33 normal individuals, 50 subjects suspected of having glaucoma, and 51 patients with glaucoma. The measurements were corrected for magnification produced by the optical system of the eye. The rim areas were statistically highly significantly different in these clinical groups. Mean neuroretinal rim area in the normal controls was 1.40 +/- 0.186 mm2. Therefore, 95% of the normal subjects had rim areas greater than 1.09 mm2; 30% of the subjects with ocular hypertension and 73% of the patients with glaucoma had rim areas less than 1.09 mm2.

Measurements of Reversibility of Optic Disc Cupping and Pallor in Ocular Hypertension and Glaucoma

Article

Nov 1985
OPHTHALMOLOGY

Four cases are presented of young and elderly glaucoma patients who had both surgical and medical therapy and showed reversal of cupping and pallor of the optic disc. The cupping was measured by photogrammetry and the pallor by computerized image analysis from photographs of the optic disc. Two patients showed regression of visual field loss. The optic disc and visual field changes corresponded to the changes in ocular pressures, generally showing worsening with an increase in ocular pressure. With a decrease in ocular pressure there was regression of visual field loss and a decrease in cupping; pallor did not decrease consistently. The changes in cupping and pallor in some patients followed similar courses but in others behaved in an independent manner. It is proposed that these new sensitive and reproducible techniques for measuring changes in the optic disc may allow the detection of disc changes early in the disease, prior to visual field loss. If treatment is begun at this time, reversal of optic disc changes may occur.

Cupping and Pallor of the Optic Disc

Article

May 1973
ARCH OPHTHALMOL-CHIC

B Schwartz

The optic disc can be described essentially by cupping and pallor. These are distinct signs; each can progress independently. Cupping has been considered by many as equivalent to the amount of paleness or color contrast estimated by subjective disc examination with a monocular direct ophthalmoscope. In normal eyes, maximum color contrast coincides generally with the degree of cupping. In glaucomatous eyes, increase in cupping occurs. Maximal color contrast no longer coincides with the degree of cupping. This is best seen by examination of the disc with a contact lens and slit lamp. In addition to change in surface contour of the disc, deviation of small vessels defines the extent of cupping. Pallor of the disc is defined by maximum color contrast as well as lack of small vessels.

Identification of the glaucomatous disc.

Article

Mar 1973
Trans Am Acad Ophthalmol Otolaryngol

The practical clinical appraisal of the optic disc in glaucoma: The natural history of cup progression and some specific disc field correlations

Article

Jan 1974
Trans Am Acad Ophthalmol Otolaryngol

Certain aspects of the natural history of cup progression and some specific disc field correlations are described. Much of the material was derived from a prospective disc field correlation study performed on 460 eyes.

The correlation between appearance of the optic cup and visual function

Article

Sep 1969
Trans Am Acad Ophthalmol Otolaryngol

M F Armaly

The epidemiology of open-angle glaucoma: A review

Article

Sep 1983
AM J EPIDEMIOL

M Cristina Leske

This paper updates and expands an earlier review of epidemiologic data on open-angle glaucoma and outlines current research needs. The areas covered include definition, the impact of glaucoma on blindness, visual impairment and use of health services, incidence and prevalence, risk factors (demographic, genetic, systemic and ocular factors) and research needs.

Reliability of Optic Disk Topographic Measurements Recorded with a Video-Ophthalmograph

Article

Aug 1984
AM J OPHTHALMOL

The video-ophthalmograph records the topography of the optic disk via simultaneous stereoscopic images which are stored and analyzed with the help of a microcomputer. This information is used to generate the vertical cup-disk ratio, the vertical optic disk diameter, the cup volume, and the neuroretinal rim area. To determine the reliability of the data, we recorded information for one eye of each of five patients ten times to determine the interphotographic error variance. We also analyzed one photograph for each of five patients ten times to determine the intraphotographic variance attributable to repeated analysis of the same photograph. The interphotographic and intraphotographic coefficients of variation were 2% to 18% and 2% to 7% respectively for these measurements.

Neuroretinal rim area in suspected glaucoma and early chronic open-angle glaucoma. Correlation with parameters of visual function

Article

Aug 1984
ARCH OPHTHALMOL-CHIC

The cup-disc ratio is an indirect measure of the amount of neural tissue present at the optic nerve head. A large cup-disc ratio in a large disc may be accompanied by a normal neuroretinal rim. The area of the neuroretinal rim is relatively constant in normal persons and should relate to visual function. In 39 eyes of patients with suspected glaucoma, normal subjects, and patients with early open-angle glaucoma, biometric findings, psychophysical functions, and optic nerve measurements were studied. The area of the neuroretinal rim was determined from stereophotographs and corrected for the magnification by the optical components of the eye. A statistically significant relation between this neuroretinal area and parameters of visual function was found. The cup-disc ratio correlated more weakly with visual function. Covariant analysis showed the neuroretinal area to be the only parameter capable in our sample of discriminating among the three clinical groups.

Morphologic Changes in the Lamina Cribrosa Correlated with Neural Loss in Open-Angle Glaucoma

Article

Jun 1983
AM J OPHTHALMOL

We divided 25 glaucomatous human eyes into three groups representing mild (Group 1, seven eyes), moderate (Group 2, 11 eyes), and severe (Group 3, six eyes) optic nerve damage, based on visual field testing or remaining number of optic nerve fibers. The optic nerve head of each eye was examined by scanning electron microscopy. Compression of the successive lamina cribrosa sheets was the earliest detected abnormality, occurring in some eyes before the detection of visual field loss. Backward bowing of the entire lamina cribrosa was a later change and involved its upper and lower poles more than the mid-nerve head. The diameter of the scleral opening at the level of Bruch's membrane did not enlarge in these adult glaucomatous eyes. Mechanical compression of the nerve head occurred early enough to be considered a primary pathogenetic event in glaucomatous damage.

Optic nerve damage in human glaucoma. III. Quantitative correlation of nerve fiber loss and visual field defect in glaucoma, ischemic neuropathy, papilledema, and toxic neuropathy

Article

Feb 1982
ARCH OPHTHALMOL-CHIC

The number and distribution of human optic nerve axons were compared with clinical measurements available the same eyes, including visual acuity, disc appearance, and visual field studies. Definite loss of axons occurs prior to reproducible visual field defects in some patients suspected of having glaucoma. In glaucoma, the superior and inferior poles of the nerve lose nerve fibers at a selectively greater rate, leading to an hourglass-shaped atrophy. Cavernous degeneration of the retrobulbar optic nerve is rarely observed in chronic glaucoma. The pattern of atrophy in examples of toxic amblyopia, ischemic optic neuropathy and chronic papilledema differ from that of glaucoma, suggesting different mechanisms of damage in these conditions.

Optic disc changes in ocular hypertension

Article

Nov 1980
SURV OPHTHALMOL

Bernard Schwartz

Measurements were made of the percent areas of optic disc cupping and pallor in ocular hypertensive eyes and normal eyes. The ocular hypertensive eyes showed significant increases in pallor but not in cupping compared to normal eyes. There was also a significant positive correlation of cupping with age in normal eyes but not in ocular hypertensive eyes. Fluorescein angiography of the optic disc showed an increase in the number and area of filling defects in ocular hypertensive compared to normal eyes. There were significant positive correlations in ocular hypertensives between percent area of filling defects with both age and systolic blood pressure which were not present with normal eyes. The optic disc changes which occur in ocular hypertensive eyes are primarily characterized by an increase in pallor and an increase in both the number and area of fluorescein filling defects.

Lessons to be learned from the Collaborative Glaucoma Study

Article

Nov 1980
SURV OPHTHALMOL

Mansour F. Armaly

The Collaborative Glaucoma Study was designed to test the hypothesis that tests of ocular pressure and its fluid dynamics before and after water drinking can predict those eyes at risk of developing field defects in the future and to describe quantitatively the predictive value of these measures. Ninety-eight of 5886 eyes (1.7%) with elevated pressures developed field damage over 1-13 years of followup. Comparing the eyes that incurred damage with those that did not, five factors were identified as having a standardized coefficient which was statistically significant; these were C-value of tonography, age, applanation pressure, cup/disc ratio and pressure change after water drinking. However, the predictive ability of all five factors--alone or considered cumulatively--was limited. A major contribution of the Collaborative Study may be in highlighting the need to continue the search for factors--perhaps hitherto unsuspected--that influence the development of glaucomatous visual field defects.

Regional Test-Retest Variability of Confocal Scanning Laser Tomography

Article

Oct 1995
AM J OPHTHALMOL

To determine the regional differences in the test-retest variability of topographic measurements of the optic nerve head obtained with a confocal scanning laser retina tomograph. Three 10-degree topographic images centered on the optic disk of 20 individuals (20 eyes), five normal subjects, seven glaucoma suspects, and eight glaucoma patients, were acquired and averaged. For each eye, the standard deviation of the mean height at each pixel was calculated from the three images and displayed as a gray scale map, which represents regional test-retest variability. The standard deviation of the mean height of the eight surrounding pixels of each pixel was calculated and plotted as a gray scale map, which represents regional steepness. The variability maps and steepness maps showed close correspondence in all eyes. Regions of high steepness had high variability (r = .31; P < .001). The average variability in relatively flat areas (sixth to 15th percentiles for steepness) was 19 microns, whereas steeper areas (81st to 95th percentiles for steepness) had an average variability of 37 microns. The overall mean variability was 28 +/- 7 microns. The regional variability of topographic measurements made with the confocal scanning laser tomograph correlates with the steepness of the corresponding region and is highest at the edge of the optic disk cup and along vessels. Because of their high test-retest variability, these steep areas may not be the best locations to follow up patients for progressive glaucomatous damage.

Correlation of Visual Field With Scanning Confocal Laser Optic Disc Measurements in Glaucoma

Article

Sep 1995
ARCH OPHTHALMOL-CHIC

To evaluate the correlation between structural optic disc measurements that were obtained with a scanning confocal laser tomograph and functional measurements that were obtained with automated static threshold perimetry. Cross-sectional study. An American school of medicine. Forty-six patients (mean +/- SD age, 63 +/- 10 years) with early to moderate glaucoma (average mean deviation +/- SD, -4.8 +/- 6.2 dB). Automated static threshold perimetry (performed with either the Octopus G1 program, Interzeag, Schlieren, Switzerland, or the Humphrey 24-2 program, Allergan Humphrey, San Leandro, Calif) and confocal optic disc tomography (performed with the Heidelberg Retina Tomograph, Heidelberg [Germany] Engineering GmbH) were performed on one randomly selected eye per patient. The correlation between the visual field indexes, namely, the mean deviation and corrected pattern SD, and the topographic measurements for the optic disc and nerve fiber layer. A statistically significant correlation was found between the third central moment of the frequency distribution of the depth values for the optic disc structures and the visual field indexes. Pearson correlation coefficients were r = -.65 (P < .0001) and r = .55 (P < .0001) for the mean deviation and corrected pattern SD, respectively. The third central moment of the depth distribution for the optic disc measurements is strongly related to the overall shape of the optic disc cup and is a robust indicator of the degree of glaucomatous optic nerve damage.

Scanning Laser Polarimetry to Measure the Nerve Fiber Layer of Normal and Glaucomatous Eyes

Article

Jun 1995
AM J OPHTHALMOL

To determine whether retardation (change in polarization) measurements of healthy subjects and glaucoma patients obtained by using a confocal scanning laser polarimeter correspond to known properties of the nerve fiber layer. A polarimeter, an optical device used to measure the change in linear polarization of light (retardation), was interfaced with a scanning laser ophthalmoscope to obtain retardation data at 65,536 locations (256 x 256 pixels) in a study of normal subjects and patients with primary open-angle glaucoma. To validate the instrument, we compared our measurements with known properties of the human retinal nerve fiber layer in 105 normal subjects. Additionally, we compared retardation measurements in eyes of 64 normal subjects and 64 age-matched glaucoma patients treated in a referral practice. In normal eyes, mean (+/- S.D.) peripapillary retardation was highest in the superior and inferior arcuate regions and lowest in the temporal and nasal regions, 12.0 +/- 1.9, 13.1 +/- 2.0, 7.0 +/- 1.8, and 7.0 +/- 1.6 degrees, respectively. Retardation decreased toward the periphery and was lower over blood vessels. In normal eyes, retardation decreased with increasing age in the superior and inferior regions. Mean retardation was statistically significantly higher among normal eyes than glaucoma eyes in the inferior and superior regions but not in the temporal or nasal areas. Scanning laser polarimetry provides quantitative measurements that correspond to known properties of the retinal nerve fiber layer in normal and glaucomatous eyes.

Neural computation in medicine

Article

May 1993
ARTIF INTELL MED

James A. Reggia

During the last decade there has been a great revival of interest in neural modelling. Powerful new computational methods have resulted from work in this area and are being applied to an increasing range of medical problems. This paper briefly explains the nature of a neural model and then reviews work in neural computation involving problems in medical informatics (e.g. expert systems) and modelling of psychiatric and neurological phenomena. The state of the art is assessed, and speculation about future developments is given.

Progression of Disc and Field Damage in Early Glaucoma

Article

Feb 1993
ARCH OPHTHALMOL-CHIC

To assess the temporal relationship between visual field progression and optic disc deterioration in early glaucoma, we studied 15 patients with unilateral visual field loss from primary open angle glaucoma. Planimetric optic disc measurements were compared with automated static threshold perimetry during a mean follow-up of 6.1 years. Eight (53%) of 15 eyes with an initially normal visual field showed progression of the disc; six of these eyes did not develop field abnormalities. The mean rates of rim-area loss were 1.7%/y in eyes with initially normal fields and 2.1%/y in eyes with initial field loss. The mean rate of visual field deterioration (change in corrected loss variance) was lower in the eyes with an initially normal field (0.3 dB2/y) than in eyes with initial field loss (3.6 dB2/y; P = .016). This longitudinal study documents progressive disc damage prior to field loss in early glaucoma.

Confocal laser scanning ophthalmoscope: Reproducibility of optic nerve head topographic measurements with the confocal laser scanning ophthalmoscope

Article

Feb 1993
OPHTHALMOLOGY

Glaucoma is an optic neuropathy in which changes in the appearance of both the optic nerve head and the surrounding tissues are important in both diagnosing its presence and progression. Accurate methods to objectively document the appearance of the optic nerve are necessary. The confocal laser scanning ophthalmoscope (Zeiss) is a new prototype instrument that may have the capability to accurately perform this function. The authors performed a prospective pilot study evaluating the ability of the confocal laser scanning ophthalmoscope to reproduce three-dimensional optic nerve images. Each retinal image contained 600,000 bytes of information. Thirty discrete images of the right optic nerves of 19 visually normal volunteers were obtained. Depth measurements were compared from the same 100 x 100 micron areas (neighborhoods). Image comparisons found the variability of depth measurements for the entire image were within 102 microns (95% confidence interval). Sixty percent of the depth measurements were reproducible within 100 microns. Variability of the depth measurements was greatest where the neuroretinal rim sloped at the edge of the optic cup and lowest in the peripapillary area. The confocal laser scanning ophthalmoscope has the potential to be a safe, rapid, and reproducible method of imaging ocular structures.

Formulas for conversion between Octopus and Humphrey threshold values and indices

Article

Nov 1995

Background: Comparing the results in a patient tested with two different automated perimeters can be difficult. The purpose of this study is to derive and test formulas for conversion between threshold values of every single test location and indices measured with Octopus and Humphrey perimeters. Methods: We tested 50 eyes of 27 patients with program G1 of the Octopus perimeter and program 24-2 of the Humphrey Field Analyzer. All pairs of tests were performed on separate days but within 1 month of each other. Twelve eyes had normal visual fields, and 38 eyes had mild to moderate glaucomatous defects. Thirty-three eyes were chosen to determine empirically prediction equations for each test location and for each of the global indices, while the 17 remaining eyes were used to test the validity of the predictions. Results: The mean absolute difference (i.e. the mean of the absolute values of the differences) between the predicted and the actual Octopus thresholds was 4.21.6 dB. The mean absolute difference between the predicted and the actual Humphrey thresholds was 5.71.6 dB. These differences were lower than the mean expected 5th to 95th percentile range of long-term fluctuation (10.32 dB for the Octopus fields and 9.22.2 dB for the Humphrey fields). Conclusion: Conversion formulas between Octopus and Humphrey formats can be used, with a known error of the estimate, to compare the perimetric results of the patients tested with the two instruments.

Neural Networks to Identify Glaucoma With Structural and Functional Measurements

Article

Jun 1996
AM J OPHTHALMOL

Neural networks can recognize patterns and classify complex variables. We assessed the ability of neural networks to discriminate between normal and glaucomatous eyes by using structural and functional measurements. Several neural network algorithms were tested with a database of 185 eyes of patients with early glaucomatous visual field loss (average mean defect, 4.5 dB) and 54 eyes of age-matched normal control subjects. The information used included automated visual field indices (mean defect, corrected loss variance, and short-term fluctuation) and structural data (cup/disk ratio, rim area, cup volume, and nerve fiber layer height) from computerized image analysis. A back propagation network with two intermediate layers assigned an estimated probability of being glaucomatous to each eye and correctly identified 88% of all eyes with 90% sensitivity and 84% specificity. The same neural network trained with only structural data correctly identified 80% of the eyes with 87% sensitivity and 56% specificity, and when trained with functional data only, it correctly identified 84% of the eyes with 84% sensitivity and 86% specificity. Analysis of several optic nerve and visual field variables by neural networks can help identify early glaucomatous damage and assign an estimated probability that early damage is present in individual patients.

Clinically detectable nerve fiber atrophy precedes the onset of glaucomatous field loss Detection of Glaucoma 2401 6 Optic nerve damage in human glaucoma. III. Quantitative correlation of nerve fiber loss and visual field defect in glaucoma , ischemic neuropathy, papilledema, and toxic neuropathy

Jan 1982
77-83135

A Sommer
J Katz
Quigley
Ha

Sommer A, Katz J, Quigley HA, et al. Clinically detectable nerve fiber atrophy precedes the onset of glaucomatous field loss. Arch Ophthalmol. 1991; 109:77-83. Detection of Glaucoma 2401 6. Quigley HA, Addicks EM, Green WR. Optic nerve damage in human glaucoma. III. Quantitative correlation of nerve fiber loss and visual field defect in glaucoma, ischemic neuropathy, papilledema, and toxic neuropathy. Arch Ophthalmol. 1982; 100:135-146.

The problem of reference plane definition for cup volume measurements

Jan 1990
197-206

A Frohn
B Jean
G Zinser
Thiel

Frohn A, Jean B, Zinser G, Thiel HJ. The problem of reference plane definition for cup volume measurements. In: Neasemann JE, Burk ROW, eds. Scanning Laser Ophthalmoscopy and Tomography. Munchen, Germany: Quintessenz 1990:197-206.

Detection of structural damage from glaucoma with confocal laser image analysis

Abstract

No full-text available

Recommended publications

Retinal nerve fiber layer thickness measured by optical coherence tomography and its correlation wit...

Hemifield association between blue-on-yellow visual field and optic nerve head topographic measureme...

Retinal nerve fiber layer in normal and glaucomatous eyes. II. Correlations

Quantification of nerve fiber layer thickness in normal and glaucomatous eyes using OCT