Fig 2 - uploaded by Wu Qiu
Content may be subject to copyright.
Source publication
Intraventricular hemorrhage (IVH) is a common disease among preterm infants with an occurrence of 12-20% in those born at less than 35 weeks gestational age. Neonates at risk of IVH are monitored by conventional 2D ultrasound (US) for hemorrhage and potential ventricular dilation. Compared to 2D US relying on linear measurements from a single slice...
Similar publications
Pre-term neonates born with a low birth weight (< 1500g) are at increased risk for developing intraventricular hemorrhage (IVH). 3D ultrasound (US) imaging has been used to quantitatively monitor the ventricular volume in IVH neonates, instead of typical 2D US used clinically, which relies on linear measurements from a single slice and visually est...
Citations
... 3,11 Semi-automated segmentation methods have been proposed, but they require user interactions, need expertise,and are time-consuming. [12][13][14][15] Two such studies 13,14 required a user to initialize the background and foreground voxels before solving a convex optimized surface evolution function using a continuous max-flow algorithm. A 2D SegNet model was developed by Gontard et al. 16 using the VGG16 encoder-decoder architecture for 3D US segmentation; however, the model interpreted the data in 2D slices only, disregarding valuable 3D information. ...
... Many previous techniques for segmenting the lateral ventricles in 3D US images were based on conventional techniques, such as convex optimization, and required some manual initializations. [12][13][14] These semiautomated approaches used less than 60 3D US images and required between 1.5 to 2.8 min to fully segment one image. [12][13][14] These methods also reported a VD of 2.5 ± 2.9 cm 3 for 25 images 13 and a DSC value that ranged from 0.72 to 0.79. ...
... [12][13][14] These semiautomated approaches used less than 60 3D US images and required between 1.5 to 2.8 min to fully segment one image. [12][13][14] These methods also reported a VD of 2.5 ± 2.9 cm 3 for 25 images 13 and a DSC value that ranged from 0.72 to 0.79. [12][13][14] Although the VDs are comparable to that of our method, ours was fully automated and used a larger two-ventricle dataset of 75 test images. ...
Background
Intraventricular hemorrhaging (IVH) within cerebral lateral ventricles affects 20–30% of very low birth weight infants (<1500 g). As the ventricles increase in size, the intracranial pressure increases, leading to post‐hemorrhagic ventricle dilatation (PHVD), an abnormal enlargement of the head. The most widely used imaging tool for measuring IVH and PHVD is cranial two‐dimensional (2D) ultrasound (US). Estimating volumetric changes over time with 2D US is unreliable due to high user variability when locating the same anatomical location at different scanning sessions. Compared to 2D US, three‐dimensional (3D) US is more sensitive to volumetric changes in the ventricles and does not suffer from variability in slice acquisition. However, 3D US images require segmentation of the ventricular surface, which is tedious and time‐consuming when done manually.
Purpose
A fast, automated ventricle segmentation method for 3D US would provide quantitative information in a timely manner when monitoring IVH and PHVD in pre‐term neonates. To this end, we developed a fast and fully automated segmentation method to segment neonatal cerebral lateral ventricles from 3D US images using deep learning.
Methods
Our method consists of a 3D U‐Net ensemble model composed of three U‐Net variants, each highlighting various aspects of the segmentation task such as the shape and boundary of the ventricles. The ensemble is made of a U‐Net++, attention U‐Net, and U‐Net with a deep learning‐based shape prior combined using a mean voting strategy. We used a dataset consisting of 190 3D US images, which was separated into two subsets, one set of 87 images contained both ventricles, and one set of 103 images contained only one ventricle (caused by limited field‐of‐view during acquisition). We conducted fivefold cross‐validation to evaluate the performance of the models on a larger amount of test data; 165 test images of which 75 have two ventricles (two‐ventricle images) and 90 have one ventricle (one‐ventricle images). We compared these results to each stand‐alone model and to previous works including, 2D multiplane U‐Net and 2D SegNet models.
Results
Using fivefold cross‐validation, the ensemble method reported a Dice similarity coefficient (DSC) of 0.720 ± 0.074, absolute volumetric difference (VD) of 3.7 ± 4.1 cm³, and a mean absolute surface distance (MAD) of 1.14 ± 0.41 mm on 75 two‐ventricle test images. Using 90 test images with a single ventricle, the model after cross‐validation reported DSC, VD, and MAD values of 0.806 ± 0.111, 3.5 ± 2.9 cm³, and 1.37 ± 1.70 mm, respectively. Compared to alternatives, the proposed ensemble yielded a higher accuracy in segmentation on both test data sets. Our method required approximately 5 s to segment one image and was substantially faster than the state‐of‐the‐art conventional methods.
Conclusions
Compared to the state‐of‐the‐art non‐deep learning methods, our method based on deep learning was more efficient in segmenting neonatal cerebral lateral ventricles from 3D US images with comparable or better DSC, VD, and MAD performance. Our dataset was the largest to date (190 images) for this segmentation problem and the first to segment images that show only one lateral cerebral ventricle.
... Irritation of the membranes (meningitis) and of the ventricles (ventriculitis) which is caused by infection or the preamble of blood causing haemorrhage can be presumed from lateral ventricle segmentation. In addition, Intraventricular hemorrhage (IVH) in preterm neonates born at less than 35 weeks gestational age and weight less than 1500 gm is a common disease with an occurrence of 12-20% [2] can be monitored by volumetric ventricle segmentation which is more sensitive to longitudinal changes in ventricular volume. It is shown that volumetric measurement offers the best manifestation of the anatomical structure of interest in medical image out of one-dimensional, two-dimensional and volumetric measurements [3] Manual segmentation done in earlier studies in order to quantify 3D ventricle volumes [4,5], is too strenuous and time consuming to make it clinically feasible. ...
Human brain is a set of four communicating network of ventricles heaving with cerebrospinal fluid (CSF) which is located inside the brain parenchyma. An efficient segmentation of cerebral lateral ventricles one in each hemisphere can support the study of efficient pathologies for successful conclusion of various diseases. In this paper, an efficient and fast energy optimised technique for volumetric segmentation of lateral ventricles from MR images of human brain is proposed which is based on geodesic active contours using level set method. The proposed approach consists of mainly four main stages: 1. Preprocessing stage, 2. Presegmentation stage, 3. Contour Evolution with Energy optimisation stage, 4. Termination stage. Experiments on multislice MRI data obtained dice coefficient of 0.955, jaccard coefficient of 0.915 and other surface distance measures demonstrate the advantages of the proposed approach in both accuracy and efficiency.
... Then, a single-phase level set image segmentation method was used to partition the image into two parts: the background and the whole ventricle region. Although low observer variability was reported in Qiu et al. (2013) , the initial landmark selection was still user dependent, labor intensive, and time consuming. In addition, we re- cently reported another user-initialized semi-automatic segmentation , which required the user to pre-label some voxels inside and outside the ventricle on a few sagittal views for initialization. ...
... Only a few previous studies have focused on the 3D US lateral ventricle segmentation problem in perterm babies . A DSC of 76.7 ± 6.2 yielded by the proposed approach is slightly higher than a DSC of 72.4 ± 2.5% reported in Qiu et al. (2013) . Although Qiu et al. (2015) reported a higher DSC of 78.4 ± 4.4%, the segmentation approach is semi-automatic, which required users to manually pre-label some voxels inside and outside the ventricles as sample voxels for learning image features for background and foreground, thereby, introducing observer variability. ...
... Convex optimization methods, especially dual optimization methods, were successfully developed to efficiently solve a wide spectrum of problems in medical image segmentation [32], [33], [34], [35], [36] and registration [37], [38], [39], [40], [34], [35], [36]. Convex optimization based methods provide both a sound theory of mathematical optimization and an efficient numerical algorithm, which is able to tackle non-smooth energy functions. ...
... Three landmarks as shown in Fig. 1, located at the anterior and inferior horn of lateral ventricles, respectively, were manually chosen from each ventricle in each 3D US image, resulting in six landmarks for each image [33]. A rigid transformation was then calculated from these six corresponding point pairs to rigidly align the follow-up images onto the base line image, spatially bringing all time-point images of each patient into the same coordinate system. ...
Preterm neonates with a very low birth weight of less than 1,500 grams are at increased risk for developing intraventricular hemorrhage (IVH), which is a major cause of brain injury in preterm neonates. Quantitative measurements of ventricular dilatation or shrinkage play an important role in monitoring patients and evaluating treatment options. 3D ultrasound (US) has been developed to monitor ventricle volume as a biomarker for ventricular changes. However, ventricle volume as a global indicator does not allow for precise analysis of local ventricular changes, which could be linked to specific neurological problems often seen in the patient population later in life. In this work, a 3D+t spatial-temporal deformable registration approach is proposed, which is applied to the analysis of the detailed local changes of preterm IVH neonatal ventricles from 3D US images. In particular, a novel sequential convex/dual optimization algorithm is introduced to extract the optimal 3D+t spatialtemporal deformable field, which simultaneously optimizes the sequence of 3D deformation fields while enjoying both efficiency and simplicity in numerics. The developed registration technique was evaluated by comparing two manually extracted ventricle surfaces from the baseline and the registered follow-up images using the metrics of Dice similarity coefficient (DSC), mean absolute surface distance (MAD), and maximum absolute surface distance (MAXD). The performed experiments using 14 patients with 5 time-point images per patient show that the proposed 3D+t registration approach accurately recovered the longitudinal deformation of ventricle surfaces from 3D US images. The proposed approach may be potentially used to analyse the change pattern of cerebral ventricles of IVH patients, their response to different treatment options, and to elucidate the deficiencies that a patient could have later in life. To the best of our knowledge, this paper reports the first study on the longitudinal analysis of neonatal ventricular system from 3D US images.
... One study, published by Ukwatta et al. [ 45 ], described the segmentation of the carotid artery in 3D MRI images. The same research group also used this technique to extract myocardial scar tissue [ 38 ], to segment the femoral artery lumen and outer wall surfaces [ 46 ], to segment lateral ventricles in preterm neonates with intraventricular hemorrhage [ 36 ], and to delineate 3D prostate boundaries for the planning and guiding of prostate biopsies [ 37 , 53 ]. ...
An innovative algorithm has been developed for the segmentation of retroperitoneal tumors in 3D radiological images. This algorithm makes it possible for radiation oncologists and surgeons semiautomatically to select tumors for possible future radiation treatment and surgery. It is based on continuous convex relaxation methodology, the main novelty being the introduction of accumulated gradient distance, with intensity and gradient information being incorporated into the segmentation process. The algorithm was used to segment 26 CT image volumes. The results were compared with manual contouring of the same tumors. The proposed algorithm achieved 90 % sensitivity, 100 % specificity and 84 % positive predictive value, obtaining a mean distance to the closest point of 3.20 pixels. The algorithm's dependence on the initial manual contour was also analyzed, with results showing that the algorithm substantially reduced the variability of the manual segmentation carried out by different specialists. The algorithm was also compared with four benchmark algorithms (thresholding, edge-based level-set, region-based level-set and continuous max-flow with two labels). To the best of our knowledge, this is the first time the segmentation of retroperitoneal tumors for radiotherapy planning has been addressed.
... Thus, an accurate, efficient and reliable automatic or semi-automatic ventricle segmentation method from 3-D US images is required to translate the use of 3-D US for neonatal ventricle volume measurement into routine clinical practice. In Qiu et al. (2013a), we reported the first study on semi-automatic segmentation of the ventricular system in preterm neonates with IVH from 3-D US images, which yielded a mean Dice similarity coefficient (DSC) of 72.4%. However, this approach required presegmented ventricle surfaces at baseline as shape priors and other six user-selected landmarks for initialization. ...
... However, these are found to be application dependent and cannot be directly applied in our 3-D US application because of differences in modality and organ or structure to be segmented. To the best of our knowledge, our previous work (Qiu et al. 2013a) is the only study on semi-automatic segmentation of ventricular system in preterm neonates with IVH from 3-D US images, and yielded a mean DSC of 72.4% for 20 patient images, lower than the 78.2% reported in this study. However, the method described in Qiu et al. (2013a) required presegmented ventricle surfaces at baseline as shape priors and another six user-selected landmarks for initialization, making it difficult to apply in a clinical setting. ...
... To the best of our knowledge, our previous work (Qiu et al. 2013a) is the only study on semi-automatic segmentation of ventricular system in preterm neonates with IVH from 3-D US images, and yielded a mean DSC of 72.4% for 20 patient images, lower than the 78.2% reported in this study. However, the method described in Qiu et al. (2013a) required presegmented ventricle surfaces at baseline as shape priors and another six user-selected landmarks for initialization, making it difficult to apply in a clinical setting. ...
A three-dimensional (3-D) ultrasound (US) system has been developed to monitor the intracranial ventricular system of preterm neonates with intraventricular hemorrhage (IVH) and the resultant dilation of the ventricles (ventriculomegaly). To measure ventricular volume from 3-D US images, a semi-automatic convex optimization-based approach is proposed for segmentation of the cerebral ventricular system in preterm neonates with IVH from 3-D US images. The proposed semi-automatic segmentation method makes use of the convex optimization technique supervised by user-initialized information. Experiments using 58 patient 3-D US images reveal that our proposed approach yielded a mean Dice similarity coefficient of 78.2% compared with the surfaces that were manually contoured, suggesting good agreement between these two segmentations. Additional metrics, the mean absolute distance of 0.65 mm and the maximum absolute distance of 3.2 mm, indicated small distance errors for a voxel spacing of 0.22 × 0.22 × 0.22 mm3. The Pearson correlation coefficient (r = 0.97, p < 0.001) indicated a significant correlation of algorithm-generated ventricular system volume (VSV) with the manually generated VSV. The calculated minimal detectable difference in ventricular volume change indicated that the proposed segmentation approach with 3-D US images is capable of detecting a VSV difference of 6.5 cm3 with 95% confidence, suggesting that this approach might be used for monitoring IVH patients' ventricular changes using 3-D US imaging. The mean segmentation times of the graphics processing unit (GPU)- and central processing unit-implemented algorithms were 50 ± 2 and 205 ± 5 s for one 3-D US image, respectively, in addition to 120 ± 10 s for initialization, less than the approximately 35 min required by manual segmentation. In addition, repeatability experiments indicated that the intra-observer variability ranges from 6.5% to 7.5%, and the inter-observer variability is 8.5% in terms of the coefficient of variation of the Dice similarity coefficient. The intra-class correlation coefficient for ventricular system volume measurements for each independent observer ranged from 0.988 to 0.996 and was 0.945 for three different observers. The coefficient of variation and intra-class correlation coefficient revealed that the intra- and inter-observer variability of the proposed approach introduced by the user initialization was small, indicating good reproducibility, independent of different users.
... While studies have quantified 3D US ventricle volumes in neonates, all have used manual contouring [3] or user-initialized semi-automatic segmentation [9] in lieu of an automatic approach. In particular, we previously proposed a semi-automatic segmentation approach [10], initialized by a single subject-specific atlas based on user-manual-selected anatomic landmarks. Then, a single-phase level set image segmentation method was used to partition the image into two parts: the background and whole ventricle region. ...
... Then, a single-phase level set image segmentation method was used to partition the image into two parts: the background and whole ventricle region. Although a low observer variability was reported in [10], the initial landmark selection was still user dependant, labor intensive and time consuming. Thus, an accurate and efficient automatic ventricle segmentation approach from 3D US images would be required in clinical practice. ...
... The experimental results using 15 IVH patient images show that the proposed method is accurate and efficient in terms of metrics of DSC, MAD, and MAXD. There were only a few previous studies focusing on 3D US lateral ventricle segmentation problem [9,10]. The DSC of 73.2 ± 3.0 yielded by the proposed approach is higher than the DSC of 72.4 ± 2.5% reported in [10]. ...
Pre-term neonates born with a low birth weight (< 1500g) are at increased risk for developing intraventricular hemorrhage (IVH). 3D ultrasound (US) imaging has been used to quantitatively monitor the ventricular volume in IVH neonates, instead of typical 2D US used clinically, which relies on linear measurements from a single slice and visually estimates to determine ventricular dilation. To translate 3D US imaging into clinical setting, an accurate segmentation algorithm would be desirable to automatically extract the ventricular system from 3D US images. In this paper, we propose an automatic multi-region segmentation approach for delineating lateral ventricles of pre-term neonates from 3D US images, which makes use of multi-phase geodesic level-sets (MP-GLS) segmentation technique via a variational region competition principle and a spatial shape prior derived from pre-segmented atlases. Experimental results using 15 IVH patient images show that the proposed GPU-implemented approach is accurate in terms of the Dice similarity coefficient (DSC), the mean absolute surface distance (MAD), and maximum absolute surface distance (MAXD). To the best of our knowledge, this paper reports the first study on automatic segmentation of ventricular system of premature neonatal brains from 3D US images.
... Ultrasound (US) or MR images of their brains are typically used to monitor the ventricular size. [2][3][4][5][6] The treatment is surgical, involving, for example, the placement of a shunt, where a tube is surgically inserted to divert cerebral spinal fluid (CSF) from the ventricles to abdomen in order to decrease pressure on the surrounding brain. Once the patient is treated, the efficiency of the treatment is evaluated by MRI, and the patients are typically followed with regular MRI over their lifetime to assess the good functioning of the shunt. ...
Intraventricular hemorrhage (IVH) or bleed within the brain is a common condition among pre-term infants that occurs in very low birth weight preterm neonates. The prognosis is further worsened by the development of progressive ventricular dilatation, i.e., post-hemorrhagic ventricle dilation (PHVD), which occurs in 10-30% of IVH patients. In practice, predicting PHVD accurately and determining if that specific patient with ventricular dilatation requires the ability to measure accurately ventricular volume. While monitoring of PHVD in infants is typically done by repeated US and not MRI, once the patient has been treated, the follow-up over the lifetime of the patient is done by MRI. While manual segmentation is still seen as a gold standard, it is extremely time consuming, and therefore not feasible in a clinical context, and it also has a large inter-and intra-observer variability. This paper proposes a segmentation algorithm to extract the cerebral ventricles from 3D T1-weighted MR images of pre-term infants with PHVD. The proposed segmentation algorithm makes use of the convex optimization technique combined with the learned priors of image intensities and label probabilistic map, which is built from a multi-atlas registration scheme. The leave-one-out cross validation using 7 PHVD patient T1 weighted MR images showed that the proposed method yielded a mean DSC of 89.7% ± 4.2%, a MAD of 2.6 ± 1.1 mm, a MAXD of 17.8 ± 6.2 mm, and a VD of 11.6% ± 5.9%, suggesting a good agreement with manual segmentations.
... Thus, accurate and efficient automatic or semi-automatic ventricle segmentation is highly desired in order to translate 3D US into clinical practice. In [7], we reported the first study on semi-automatic segmentation of lateral ventricles in neonates with IVH from 3D US images, which yielded a mean dice similarity coefficient (DSC) of 72.4%. However, it requires pre-segmented ventricles at baseline as shape priors and six user selected landmarks for initialization. ...
... The proposed semiautomatic segmentation method makes use of the latest development of convex optimization technique supervised by user interactive information. The experimental results show that our proposed method is accurate with a mean DSC of 78.9% for 25 3D US images of 5 patients (5 time points for each subject), which is higher than 72.4% obtained in our previous study [7]. In addition, the intra-observer variability experiment showed that the variability introduced by the user initialization is small in terms of DSC, which demonstrates that the proposed method has a low intra-observer variability. ...
3D Ultrasound (US) has been developed recently to image the intracranial ventricular system of pre-term neonates in order to monitor these patients for intraventricular hemorrhage (IVH) and the resultant dilatation of the ventricles. 3D US is capable of providing volumetric ventricle measurements, compared to clinically used 2D US, relying on linear measurements from a single slice, and visual and quantitative estimates to determine the severity of ventricular dilatation. In this work, we propose a convex optimization-based segmentation approach for 3D US images of the cerebral ventricles in preterm neonates with IVH. The proposed semi-automatic segmentation method makes use of the latest development in convex optimization techniques supervised by user interactive information. Experiments using 25 3D US images of 5 patients (5 time points for each subject) show that our proposed approach yielded a mean DSC of 78.9% compared to a manually contoured surface. This GPU-implemented semi-automated approach reduced the time required per segmentation by 1200% (mean times: 2.5 vs. 30 minutes). In addition, the intra-observer variability experiments showed that the variability introduced by the user initialization is small in terms of DSC, demonstrating a low intra-user variability.
... Motivated by the recent developments of convex optimization-based image segmentation ( Lellmann et al., 2009;Yuan et al., 2010b;Ukwatta et al., 2013;Qiu et al., 2013a,c), we propose to solve the challenging combinatorial optimization problem (7) by its convex relaxation: ...
Efficient and accurate segmentation of the prostate and two of its clinically meaningful sub-regions: the central gland (CG) and peripheral zone (PZ), from 3D MR images, is of great interest in image-guided prostate interventions and diagnosis of prostate cancer. In this work, a novel multi-region segmentation approach is proposed to simultaneously segment the prostate and its two major sub-regions from only a single 3D T2-weighted (T2w) MR image, which makes use of the prior spatial region consistency and incorporates a customized prostate appearance model into the segmentation task. The formulated challenging combinatorial optimization problem is solved by means of convex relaxation, for which a novel spatially continuous max-flow model is introduced as the dual optimization formulation to the studied convex relaxed optimization problem with region consistency constraints. The proposed continuous max-flow model derives an efficient duality-based algorithm that enjoys numerical advantages and can be easily implemented on GPUs. The proposed approach was validated using 18 3D prostate T2w MR images with a body-coil and 25 images with an endo-rectal coil. Experimental results demonstrate that the proposed method is capable of efficiently and accurately extracting both the prostate zones: CG and PZ, and the whole prostate gland from the input 3D prostate MR images, with a mean Dice similarity coefficient (DSC) of 89.3±3.2% for the whole gland (WG), 82.2±3.0% for the CG, and 69.1±6.9% for the PZ in 3D body-coil MR images; 89.2±3.3% for the WG, 83.0±2.4% for the CG, and 70.0±6.5% for the PZ in 3D endo-rectal coil MR images. In addition, the experiments of intra- and inter-observer variability introduced by user initialization indicate a good reproducibility of the proposed approach in terms of volume difference (VD) and coefficient-of-variation (CV) of DSC.
