No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Automatic analysis of auditory nerve electrically evoked compound action potential with an artificial neural network
Abstract
The auditory nerve's electrically evoked compound action potential is recorded in deaf patients equipped with the Nucleus 24 cochlear implant using a reverse telemetry system (NRT). Since the threshold of the NRT response (NRT-T) is thought to reflect the psychophysics needed for programming cochlear implants, efforts have been made by specialized management teams to develop its use. This study aimed at developing a valid tool, based on artificial neural networks (ANN) technology, for automatic estimation of NRT-T. The ANN used was a single layer perceptron, trained with 120 NRT traces. Learning traces differed from data used for the validation. A total of 550 NRT traces from 11 cochlear implant subjects were analyzed separately by the system and by a group of physicians with expertise in NRT analysis. Both worked to determine 37 NRT-T values, using the response amplitude growth function (AGF) (linear regression of response amplitudes obtained at decreasing stimulus intensity levels). The validity of the system was assessed by comparing the NRT-T values automatically determined by the system with those determined by the physicians. A strong correlation was found between automatic and physician-obtained NRT-T values (Pearson r correlation coefficient >0.9). ANOVA statistics confirmed that automatic NRT-Ts did not differ from physician-obtained values (F = 0.08999, P = 0.03). Moreover, the average error between NRT-Ts predicted by the system and NRT-Ts measured by the physicians (3.6 stimulation units) did not differ significantly from the average error between NRT-Ts measured by each of the three physicians (4.2 stimulation units). In conclusion, the automatic system developed in this study was found to be as efficient as human experts for fitting the amplitude growth function and estimating NRT-T, with the advantage of considerable time-saving.
... Systems that automatically measure ECAP thresholds (or determine them offline with a given series of NRT measurements) have been built in the past (Charasse et al. 2004b;Van Dijk et al. 2003) and continue to be built (Litvak and Emadi 2005;Arnold et al. 2007). In all cases, the chosen method has been extrapolated threshold (Section 1.3.3). ...
... As Figure 2.2 shows, extrapolated threshold is poorly defined at these levels (also Abbas et al. 2004a With these limitations, previous systems have been designed to only give ECAP threshold results in highly specific circumstances, and this has reduced their success rates, and hence their level of automation. The systems of Charasse et al. (2004b) and Van Dijk et al. (2003) only use responses with clear N1 and P1 peaks. Litvak and Emadi (2005) reject 40% of their training dataset, only including traces that are classified unanimously by five clinicians. ...
... Litvak and Emadi (2005) reject 40% of their training dataset, only including traces that are classified unanimously by five clinicians. Charasse et al. and Van Dijk et al., with small subject pools, do not provide a clear indication of their systems' success rates; a reduced rate is suggested by the sensitivities of their expert systems (68% and 80% respectively, with clear peaks required) and the requirements of obtaining an AGF (Charasse et al. 2004b require five ECAPs with both N1 and P1 peaks visible). AutoNRT, by contrast, achieves a success rate of approximately 95% by freeing itself of these limitations. ...
... Systems that automatically measure T-NRT levels (or determine them offline with a given set of NRT measurements) have been built in the past [15,16] and continue to be built [17]. In all cases, the chosen method has been extrapolated threshold . ...
... An expert system analyses NRT measurements at a range of current levels; those that are deemed to represent ECAPs are used to construct an AGF, from which a T-NRT level is extrapolated. The expert systems have taken various forms: Charasse et al. [15] used an artificial neural network (ANN) where the output neurons corresponded to one of five ECAP morphologies (both N1 and P1 visible, N1 missing, no neural response, etc.); Charasse et al. [18] also compared the ANN to a cross-correlation (CC) technique, where a given NRT measurement is compared with an array of fixed neural responses, grouped according to the five ECAP morphologies; and Nicolai et al. [19] presented an expert system that combined the ANN and CC techniques with additional rule-based criteria. The AGF linearity assumption is not valid at all current levels however. ...
... To achieve T-NRT levels with a high success rate, AutoNRT is sufficiently sensitive with all possible ECAP morphologies. Whereas the systems of Charasse et al. [15] and van Dijk et al. [16] only use responses with clear N1 and P1 peaks (a prudent precaution with Nucleus CI24M/R waveforms), AutoNRT makes no distinction in ECAP morphology. This provides AutoNRT with a greater chance of success on any given electrode. ...
Objective:
AutoNRT is an automated system that measures electrically evoked compound action potential (ECAP) thresholds from the auditory nerve with the Nucleus Freedom cochlear implant. ECAP thresholds along the electrode array are useful in objectively fitting cochlear implant systems for individual use. This paper provides the first detailed description of the AutoNRT algorithm and its expert systems, and reports the clinical success of AutoNRT to date.
Methods:
AutoNRT determines thresholds by visual detection, using two decision tree expert systems that automatically recognise ECAPs. The expert systems are guided by a dataset of 5393 neural response measurements. The algorithm approaches threshold from lower stimulus levels, ensuring recipient safety during postoperative measurements. Intraoperative measurements use the same algorithm but proceed faster by beginning at stimulus levels much closer to threshold. When searching for ECAPs, AutoNRT uses a highly specific expert system (specificity of 99% during training, 96% during testing; sensitivity of 91% during training, 89% during testing). Once ECAPs are established, AutoNRT uses an unbiased expert system to determine an accurate threshold. Throughout the execution of the algorithm, recording parameters (such as implant amplifier gain) are automatically optimised when needed.
Results:
In a study that included 29 intraoperative and 29 postoperative subjects (a total of 418 electrodes), AutoNRT determined a threshold in 93% of cases where a human expert also determined a threshold. When compared to the median threshold of multiple human observers on 77 randomly selected electrodes, AutoNRT performed as accurately as the 'average' clinician.
Conclusions:
AutoNRT has demonstrated a high success rate and a level of performance that is comparable with human experts. It has been used in many clinics worldwide throughout the clinical trial and commercial launch of Nucleus Custom Sound Suite, significantly streamlining the clinical procedures associated with cochlear implant use.
... Here, we reviewed post-2015 AI applications in the field of otorhinolaryngology. Before 2015, most AI-based technologies focused on CIs [10,[75][76][77][78][79][80][81][82][83][84][85][86]. However, AI applications have expanded greatly in recent years. ...
Objectives:
To present an up-to-date survey of the use of artificial intelligence (AI) in the field of otorhinolaryngology, with respect to opportunities, research challenges, and research directions.
Methods:
We searched PubMed, the Cochrane Central Register of Controlled Trials, Embase, and the Web of Science. We initially retrieved 458 articles; we excluded non-English publications and duplicates, which resulted in a total of 90 remaining studies. These 90 studies were divided into those analyzing medical images, voice, medical devices, and clinical diagnoses and treatments.
Results:
Most studies (42.22%, 38/90) used AI for image-based analysis, followed by clinical diagnosis and treatments (24 studies); each of the remaining two subcategories included 14 studies.
Conclusion:
Machine and deep learning have been extensively applied in the field of otorhinolaryngology. However, performance varies and research challenges remain.
... A subsequent follow-up study by Gartner et al. (23) validated the performance of ''AutoNRT'' in a new software package with performance similar to that of trained audiologists with a computation time range of 30 seconds to 1 minute. In an earlier study, Charasse et al. (24) used a different algorithm approach and developed a neural network to rapidly identify ECAP waveforms in patients with CIs. When compared with human ''expert'' analysis, the neural network performed as well as human experts for fitting the auditory nerve amplitude growth function and estimating neural response thresholds. ...
Objective:
The use of machine learning technology to automate intellectual processes and boost clinical process efficiency in medicine has exploded in the past 5 years. Machine learning excels in automating pattern recognition and in adapting learned representations to new settings. Moreover, machine learning techniques have the advantage of incorporating complexity and are free from many of the limitations of traditional deterministic approaches. Cochlear implants (CI) are a unique fit for machine learning techniques given the need for optimization of signal processing to fit complex environmental scenarios and individual patients' CI MAPping. However, there are many other opportunities where machine learning may assist in CI beyond signal processing. The objective of this review was to synthesize past applications of machine learning technologies for pediatric and adult CI and describe novel opportunities for research and development.
Data sources:
The PubMed/MEDLINE, EMBASE, Scopus, and ISI Web of Knowledge databases were mined using a directed search strategy to identify the nexus between CI and artificial intelligence/machine learning literature.
Study selection:
Non-English language articles, articles without an available abstract or full-text, and nonrelevant articles were manually appraised and excluded. Included articles were evaluated for specific machine learning methodologies, content, and application success.
Data synthesis:
The database search identified 298 articles. Two hundred fifty-nine articles (86.9%) were excluded based on the available abstract/full-text, language, and relevance. The remaining 39 articles were included in the review analysis. There was a marked increase in year-over-year publications from 2013 to 2018. Applications of machine learning technologies involved speech/signal processing optimization (17; 43.6% of articles), automated evoked potential measurement (6; 15.4%), postoperative performance/efficacy prediction (5; 12.8%), and surgical anatomy location prediction (3; 7.7%), and 2 (5.1%) in each of robotics, electrode placement performance, and biomaterials performance.
Conclusion:
The relationship between CI and artificial intelligence is strengthening with a recent increase in publications reporting successful applications. Considerable effort has been directed toward augmenting signal processing and automating postoperative MAPping using machine learning algorithms. Other promising applications include augmenting CI surgery mechanics and personalized medicine approaches for boosting CI patient performance. Future opportunities include addressing scalability and the research and clinical communities' acceptance of machine learning algorithms as effective techniques.
... There has been only limited research done on automatic recording and analysis of ECAP data. Charasse et al. (2004b) published a paper on the use of Artificial Neural Networks (ANN) in NRT data analysis. Van Dijk et al. (2003) presented an automatic measurement and analysis tool prototype based on the ANN developed by Charasse et al. ...
AutoNRT is the completely automatic electrically evoked compound action potential (ECAP) measuring algorithm in the recently released Nucleus Freedom cochlear implant system. AutoNRT allows clinicians to automatically record T-NRT profiles that in turn can be used as a guide for initial fitting. The algorithm consists of a pattern recognition part that judges if the traces contain an ECAP and an intelligent flow that optimizes the measurement parameters and finds the ECAP threshold (T-NRT). The objective of this study was to determine how accurate, reliable, and fast the automatic measurements are.
Data on more than 400 electrodes were collected as part of the multicenter clinical trial of the Nucleus Freedom cochlear implant system. T-NRT values determined by the algorithm were compared with T-NRT determinations on the same data by different human observers. Also, the time the measurements took was analyzed.
In 90% of the cases, the absolute difference between the AutoNRT and the human observer determined T-NRT was less than 9 CL; the median absolute difference was 3 CL. A second experiment, in which a group of human observers were asked to analyze NRT data, showed high variability in T-NRT; in some cases, two experienced clinicians disagreed by more than 30 current levels. Compared with the group, AutoNRT performed as well as the "average" clinician, with the advantage that the AutoNRT threshold determinations are objective. Analysis of the timing data showed an average intraoperative measurement time of less than 20 sec per electrode with a standard deviation of 5 sec, suggesting that the total array of 22 electrodes can be measured intraoperatively in about 7 minutes on average.
AutoNRT provides comparable accuracy to an average clinician but with the added benefit of significant time savings over manual recordings. This makes it a valuable tool for clinical measurement of ECAP threshold in cochlear implant recipients.
This chapter describes a common methodology for programming using examples encountered in common neuromodulation programming situations. Programming is the term given to the spatial and temporal adjustment of the stimulation signal parameters used to treat specific neurologic disorders. All neuromodulation systems rely on a continuity of stimulation current from the stimulation generator to the tissue, so it is critical that implanted hardware be functioning. Continuity assessment involves passing a known quantity of current, or generating a known voltage difference across the active and reference leads, from the pulse generator and then recording the variation in potential or current across the system. The initial programming session should include a complete therapeutic evaluation. All devices are delivered in a standard configuration from the manufacturer. Prior to starting the programming session, a complete hardware check should be performed. Once second side testing is performed and documented, the patient should have both stimulators turned "on" to the optimal and combined effects sent to the waiting area for at least 30 minutes.
The Neural Response Telemetry (NRT) system® provides a simplified tool for Electrically Evoked Compound Action Potential (ECAP) recordings. This system was successfully validated in humans and permits in-situ measurement of auditory neural activity within the cochlea in response to electrical stimulation and allows many clinical applications. One of the main difficulties in using this system is the time required by the clinicians for both, measure and data analysis. Data analysis requires the selection of valid neural traces. This selection cannot be done only based on the amplitude information because trace morphology cues add to the decision. In this study, we propose two different recognition methods to automatically recognize neural responses traces recorded with NRT system. The first method uses artificial neural network technology (ANN) while the second uses a cross correlation method (CC). Both systems were fed with 120 neural responses divided in five different categories associated to the main response morphologies observed with NRT recordings. For both methods, ROC curves method was used with a set of one thousand NRT traces to compare and evaluate performances. NRT traces were classified by two human experts who worked together to reduce the expert subjectivity. NRT traces were classified in three different categories depending on the clinical application expected with each trace. Results show that CC method led to a higher degree of precision than the ANN method. Particularly a score of 82% was obtained for the correct positives and correct negatives identification with a false positive rate fixed at only 2%. The results presented in this study suggest that CC method is accurate enough to be used for clinical applications.
To evaluate the smart algorithm in speed and reliability of threshold estimation compared with the algorithm available in the standard fitting software and to evaluate the possibility of using programs based on the smart algorithm instead of programs derived from behavioral measures.
Twenty subjects unilaterally implanted with a CII Bionic Ear or HiRes90K device.
Neural response imaging thresholds (tNRI) were measured using both the smart approach within the Research Studies Platform for Objective Measures and the SoundWave fitting software. Measurements were performed intraoperatively, at first fitting, and after 3 months of implant use. Each subject received a standard behavioral program and a SmartNRI program. Speech perception tests were conducted at 3 months, and subjective preferences were documented.
Smart tNRI and SoundWave tNRI at each session; speech test results and subject preferences at the 3-month session.
High correlations were found between smart tNRI and SoundWave tNRI. The time required to obtain NRI thresholds with the smart algorithm was a quarter of the time needed with SoundWave. Although most tested subjects preferred their behavioral programs, there were no significant differences in performance between SmartNRI and behavioral programs.
Neural response imaging thresholds were obtained more rapidly with the smart algorithm than with SoundWave. Because no differences were observed between SmartNRI and behavioral programs, SmartNRI programs may be a useful alternative to behavioral programs in difficult to fit cases, where user feedback is sometimes difficult to obtain.
During the last decade, cochlear implantation has evolved into a well-established treatment of deafness, predominantly because of many improvements in speech processing and the controlled excitation of the auditory nerve. Cochlear implants now also feature telemetry, which is highly useful to monitor the proper functioning of the implanted electronics and electrode contacts. Telemetry can also support the clinical management in young children and difficult cases where neural unresponsiveness is suspected. This article will review recent advances in the telemetry of the electrically evoked compound action potential that have made these measurements simple and routine procedures in most cases. The distribution of the electrical stimulus itself sampled by "electrical field imaging" reveals general patterns of current flow in the normal cochlea and gross abnormalities in individual patients; models have been developed to derive more subtle insights from an individual electrical field imaging. Finally, some thoughts are given to the extended application of telemetry, for example, in monitoring the neural responses or in combination with other treatments of the deaf ear.
Neural response telemetry (NRT) permits in situ intracochlear recordings of the electrically evoked compound action potential from the auditory nerves using scala tympani electrodes. The recorded NRT waveforms can generally be categorized under either single positive peak or double positive peak waveforms. This is similar to the observations from Stypulkowski and van den Honert, who suggested that the double peak complex arises from two components that could be axonal and dendritic in origin, respectively. Using a simple mathematical model which linearly combines two separate waveforms similar in shape but differing in amplitude and latencies, it was possible to simulate the various NRT waveform categories. The simulation results support the view that the two waveform components originate from dendritic or axonal processes and implies that the shape of the response waveform may provide information about the degree of neural survival in the stimulated cochlea. This information could be useful for determining optimal speech coding parameters for cochlear implant users on an individual basis.
Research in neural response telemetry (NRT) with the Cochlear 'Nucleus' CI24M Implant began a few years ago, using the first NRT software version (2.04). It has been demonstrated that NRT offers possibilities for fitting the implant speech processor. NRT sessions, however, remain lengthy for the patients, especially for children, and NRT research is seeking to reduce session time. The new version of the NRT software (3.0) allows implant stimulation up to 400 Hz, whereas the previous version stimulated only up to 80 Hz. The time gained is very significant. The goal of the present study was to observe and assess the modifications in NRT produced by these higher stimulus frequencies. Growth functions were measured in six adult patients for electrodes 5-10-15-20 at frequencies of 20, 80, 150, 250 and 365 Hz to determine NRT thresholds (NRT-T). Subjective detection thresholds (T) were also measured for the same electrodes and frequencies. The results showed that the number of valid responses, like the quality and amplitude of the NRT response, quickly decreased at these higher rates. Consequently, the number of measurement points used for the growth function decreased with frequency. Moreover, when frequency increased, the T value decreased while NRT-T increased: the gap between the two thresholds increased with frequency in a linear manner common to all patients. The growth function slopes did not change significantly with frequency.
Computer-aided interpretation of electrocardiograms (ECGs) is widespread but many physicians hesitate to rely on the computer, because the advice is presented without information about the confidence of the advice. The purpose of this work was to develop a method to validate the advice of a computer by estimating the error of an artificial neural network output. A total of 1249 ECGs, recorded with computerized electrocardiographs, on patients who had undergone diagnostic cardiac catheterization were studied. The material consisted of two groups, 414 patients with and 835 without anterior myocardial infarction. The material was randomly divided into three data sets. The first set was used to train an artificial neural network for the diagnosis of anterior infarction. The second data set was used to calculate the error of the network outputs. The last data set was used to test the network performance and to estimate the error of the network outputs. The performance of the neural network, measured as the area under the receiver operating characteristic (ROC) curve, was 0·887 (0·845–0·922). The 25% test ECGs with the lowest error estimates had an area under the ROC curve as high as 0·995 (0·982–1·000), i.e. almost all of these ECGs were correctly classified. Neural networks can therefore be trained to diagnose myocardial infarction and to signal when the advice is given with great confidence or when it should be considered more carefully. This method increases the possibility that artificial neural networks will be accepted as reliable decision support systems in clinical practice.
In the clinical application of auditory brainstem responses (ABRs), the latencies of five to seven main peaks are extremely important parameters for diagnosis. In practice, the latencies have mainly been done by manual measurement so far. In recent years, some new techniques have been developed involving automatic computer recognition. Computer recognition is difficult, however, since some peaks are complicated and vary a lot individually. In this paper, we introduce an artificial neural network method for ABR research. The detection of ABR is performed by using artificial neural networks. A proper bandpass filter is designed for peak extraction. Moreover, a new approach to estimate the latencies of the peaks by artificial neural networks is presented. The neural networks are studied in relation to the selection of model, number of layers and number of neurons in each hidden layer. Experimental results are described showing that artificial neural networks are a promising method in the study of ABR.
This study describes a method for recording the electrically evoked, whole-nerve action potential (EAP) in users of the Ineraid cochlear implant. The method is an adaptation of one originally used by Charlet de Sauvage et al. [J. Acoust. Soc. Am. 73, 615-627 (1983)] in guinea pigs. The response, recorded from 11 subjects, consists of a single negative peak that occurs with a latency of approximately 0.4 ms. EAP input/output functions are steeply sloping and monotonic. Response amplitudes ranging up to 160 micro V have been recorded. Slope of the EAP input/output function correlates modestly (approximately 0.6-0.69) with results of tests measuring word recognition skills. The refractory properties of the auditory nerve were also assessed. Differences across subjects were found in the rate of recovery from the refractory state. These findings imply that there may be difference across subjects in the accuracy with which rapid temporal cues can be coded at the level of the auditory nerve. Reasonably strong correlations (approximately 0.74-0.85) have been found between the magnitude of the slope of these recovery curves and performance on tests of word recognition.
In a companion paper [Brown et al., J. Acoust. Soc. Am. 88, 1385-1391 (1990)], a method for recording the electrically evoked whole-nerve action potential in human cochlear implant users was reported. The procedure for recording the response requires that two biphasic current pulses, a "masker" and a "probe," be presented at a rate and level sufficient to drive the auditory nerve into a refractory state. The present study was designed to assess the sensitivity of that recording technique to variations in stimulation parameters. The experiments described in this paper demonstrate that: (1) the EAP as recorded in the cat is triphasic and is defined by two negative peaks occurring at latencies of approximately 0.26 and 0.82 ms; (2) EAP amplitude is independent of the level of the masker stimulus for current levels equal to or greater than the current level of the probe stimulus; and (3) the time course of recovery of the EAP from the refractory state is stable over a range of both probe and masker current levels.
Expert systems are computer programs which incorporate artificial intelligence technology and are created to emulate the decision-making abilities of human experts. The advantage of such systems lies in their ability to capture and model expert problem solving knowledge in a domain and make it available to an unlimited number of consumers in an economic and efficient way. The purpose of this project was to develop an expert system to interpret infant auditory brainstem response data as entered by the user. The resulting system provides diagnostic conclusions regarding hearing status, type of hearing loss, and brainstem function at an accuracy level equal to that of a human expert.
This paper presents a study of the application of autonomously learning multiple neural network systems to medical pattern classification tasks. In our earlier work, a hybrid neural network architecture has been developed for on-line learning and probability estimation tasks. The network has been shown to be capable of asymptotically achieving the Bayes optimal classification rates, on-line, in a number of benchmark classification experiments. In the context of pattern classification, however, the concept of multiple classifier systems has been proposed to improve the performance of a single classifier. Thus, three decision combination algorithms have been implemented to produce a multiple neural network classifier system. Here the applicability of the system is assessed using patient records in two medical domains. The first task is the prognosis of patients admitted to coronary care units; whereas the second is the prediction of survival in trauma patients. The results are compared with those from logistic regression models, and implications of the system as a useful clinical diagnostic tool are discussed.
This study aimed to compare recordings of the electrically evoked whole nerve action potential (EAP) made using the reverse telemetry system of the Nucleus CI24M device with those recorded from individuals who use the Ineraid cochlear implant system.
Data were collected in a prospective fashion from Nucleus CI24M cochlear implant users and compared with retrospective data collected from patients who use the Ineraid device.
All data were collected at the Department of Otolaryngology-Head and Neck Surgery, University of Iowa Hospitals and Clinics.
Data are reported from 8 patients who use the Nucleus CI24M cochlear implant and 20 patients who use the Ineraid cochlear implant system.
The interventions described in this study were diagnostic in nature.
EAP growth and refractory recovery data are reported. EAP thresholds recorded from patients who use the Nucleus CI24M device also are compared with behavioral thresholds for the stimulus used to evoke the EAP as well as the stimulation levels needed to program the speech processor.
EAP morphology, growth, and refractory recovery functions recorded using the Nucleus CI24M reverse telemetry system compared favorably with similar measures recorded from Ineraid cochlear implant users.
Reasonable EAP responses can be recorded using the Nucleus CI24M device. More data are needed to determine whether the information about neural responsiveness available with this device will be clinically useful.
This study outlines a series of experiments using the neural response telemetry (NRT) system of the Nucleus CI24M cochlear implant to measure the electrically evoked compound action potential (EAP). The goal of this investigation was to develop a protocol that allows successful recording of the EAP in a majority of CI24M cochlear implant users.
Twenty-six postlingually deafened adults participated in this study. A series of experiments were conducted that allowed us to examine how manipulation of stimulation and recording parameters may affect the morphology of the EAP recorded using the Nucleus NRT system.
Results of this study show consistent responses on at least some electrodes from all subjects. Cross-subject and cross electrode variations in both the growth of the response and the temporal refractory properties of the response were observed. The range of stimulus and recording parameters that can be used to record the EAP with the Nucleus NRT system is described.
Using the protocol outlined in this study, it is possible to reliably record EAP responses from most subjects and for most electrodes in Nucleus CI24M cochlear implant users. These responses are robust and recording these responses does not require that the subject sleep or remain still. Based on these results, a specific protocol is proposed for measurement of the EAP using the NRT system of the CI24M cochlear implant. Potential clinical implications of these results are discussed.
The purpose of this part of the INTERVENT project was (1) to redefine and individually predict post-procedural complications associated with coronary interventions, including alternative/adjunctive techniques to PTCA and (2) to employ the prognostic INTERVENT computer model to clarify the structural relationship between (pre)-procedural risk factors and post-procedural outcome.
In a multicentre study, 2500 data items of 455 consecutive patients (mean age: 61.1+/-8.3 years: 33-84 years) undergoing coronary interventions at three university centres were analysed. 80.4% of the patients were male, 16.7% had unstable angina, and 5.1%/10.1% acute/subacute myocardial infarction. There were multiple or multivessel stenoses in 16.0%, vessel bending >90 degrees in 14.5%, irregular vessel contours in 65.0%, moderate calcifications in 20.9%, moderate/severe vessel tortuosity in 53.2% and a diameter stenosis of 90%-99% in 44.4% of cases. The in-lab (out-of-lab) complications were: 0.4% (0.9%) death, 1.8% (0.2%) abrupt vessel closure with myocardial infarction and 5.5% (4.0) haemodynamic disorders.
Computer algorithms derived from artificial intelligence were able to predict the individual risk of these post-procedural complications with an accuracy of >95% and to explain the structural relationship between risk factors and post-procedural complications. The most important prognostic factors were: heart failure (NYHA class), use of adjunctive/alternative techniques (rotablation, atherectomy, laser), acute coronary ischaemia, pre-existent cardiac medication, stenosis length, stenosis morphology (calcification), gender, age, amount of contrast agent and smoker status. Pre-medication with aspirin or other cardiac medication had a beneficial effect. Techniques, such as laser angioplasty or atherectomy were predictors for post-procedural complications. Single predictors alone were not able to describe the individual outcome completely.
The purpose of this study is to examine the relationship between the electrically evoked compound action potential (EAP) thresholds and the MAP thresholds (T-levels) and maximum comfort levels (C-levels) in children implanted with the Nucleus 24 device.
EAP thresholds were measured using the Neural Response Telemetry system of the Nucleus 24 device. Twenty children implanted with the Nucleus 24 cochlear implant participated in this study. EAP thresholds were compared with the behavioral measures of T- and C-level used to construct the MAP these children used on a daily basis. For these subjects, both EAP and MAP T- and C-levels were obtained the same visit, which occurred at 3 to 5 mo postconnection.
EAP thresholds were shown to fall between MAP T- and C-level for 18 of 20 subjects tested; however, considerable variability across subjects was noted. On average, EAP thresholds fell at 53% of the MAP dynamic range. Correlations between EAP threshold and MAP T- and C-level improved substantially when combined with behavioral measures obtained from one electrode in the array.
Moderate correlations were found between EAP thresholds and MAP T- and C-levels for the children participating in this study. However, a technique is described for improving the accuracy of predictions of MAP T- and C-levels based on EAP data combined with a small amount of behavioral information.
The objective of this study was to determine the relationship between electrically evoked whole nerve action potential (EAP) and electrical auditory brain stem response (EABR) thresholds and MAP threshold (T-level) and maximum comfort level (C-level) for subjects who use the Nucleus 24 cochlear implant system.
Forty-four adult Nucleus 24 cochlear implant users participated in this study. EAP thresholds were recorded using the Neural Response Telemetry System developed by Cochlear Corporation. EABR thresholds were measured for a subset of 14 subjects using standard evoked potential techniques. These physiologic thresholds were collected on a set of five electrodes spaced across the cochlea, and were then compared with behavioral measures of T-level and C-level used to program the speech processor.
EAP thresholds were correlated with MAP T- and C-levels; however, the correlation was not strong. A technique for improving the correlation by combining measures of T- and C-levels made on one electrode with the EAP thresholds was presented. Correlations between predicted and measured T- and C-levels using this technique were 0.83 and 0.77, respectively. Similar results were obtained using the EABR thresholds for a smaller set of subjects. In general, EABR thresholds were recorded at levels that were approximately 4.7 programming units lower than EAP thresholds.
Either EAP or EABR thresholds can be used in combination with a limited amount of behavioral information to predict MAP T- and C-levels with reasonable accuracy.
This paper describes the neural response telemetry (NRT) results obtained from the first 30 patients tested at this centre. One hundred per cent of patients tested intra-operatively had NRT responses on at least one electrode; this compared to 82.4% of patients tested post-operatively. Reasonable correlations existed between post-operative NRT thresholds and psychophysical threshold and comfort levels, although there was too much variability for the data to be used to set these parameters directly. Post-operative NRT thresholds were always at levels audible to patients.
Response from spiral ganglion cells to electrical stimulation via the Nucleus 24 cochlear implant can be measured using the neural response telemetry system. The purpose of this study was to assess, in children, the correlation between the neural response threshold and the behavioral levels used for cochlear implant programming process.
The neural response telemetry test was administered to 23 children (mean age at implantation: 4 years) with the Nucleus 24 cochlear implant. Four intra-cochlear electrodes (electrodes 5, 10, 15 and 20) were tested. The neural response threshold at 3, 6, 9 and 12 months post-implantation was compared with the behavioral threshold and the maximum comfort level estimated during the same periods: a Pearson's correlation test was performed for each tested electrode.
On apical electrodes, the correlation with the behavioral threshold remained significant from 3 to 12 months post-implantation (r ranging from 0.696 to 0.909, P<0.05), and the correlation with the maximum comfort level was also significant throughout the study period, except on electrode 15 at 9 months (tendency to significance). On basal and intermediate electrodes, statistical correlations were found only at some points of time; nonetheless, at 12 months post-implantation, a significant correlation with behavioral levels could be clearly demonstrated both on electrode 15 (r=0.914--0.778, P<0.05) and on electrode 10 (r=0.845--0.720, P<0.05).
This preliminary study suggests that the correlation between the neural response threshold and behavioral levels may improve from the base towards the apex of the cochlea. However, a significant correlation can be demonstrated for all tested electrodes at 12 months post-implantation. During the first months post-implantation care must be exercised when interpreting neural response telemetry measurements: a positive test does not necessarily mean that the stimulus delivered to the acoustic nerve will be centrally processed with the result of an auditory perception.
The goal of this study was to estimate psychophysical levels using the electrically evoked compound action potential (EAP), measured with the Neural Response Telemetry capabilities of Cochlear Corporation's Nucleus CI24M cochlear implant system.
Twelve postlingually deafened adults with at least 3 mo of implant experience with the CI24M were subjects in this study. EAP growth functions were successfully quantified on each active electrode of every subject. Correlation and regression analyses were performed between EAP measures and cochlear implant fitting psychophysics. Other information including performance, etiology and duration of hearing loss, and individual electrode impedance was considered.
EAP thresholds were found to be highly correlated with psychophysical thresholds. The rate of EAP growth with increasing stimulation levels was also found to be correlated with the dynamic range of loudness limits and psychophysical thresholds in some subjects. No relationship was evident between EAP measures and speech perception tests.
Information from EAP growth function measurements may be used to estimate psychophysical information used in cochlear implant fitting but not to predict performance with the device.
The electrically evoked action potential (EAP) was recorded intra-operatively by use of neural response telemetry (NRT) on the Nucleus C124M cochlear implant. The aim of the present study was to investigate the EAP in young children immediately following implant surgery and whilst the children were still anaesthetized. The effect of data collection parameters on the reliability of the EAP was assessed and the relationships of the EAP findings to the intra-operative electrical auditory brainstem response (EABR) and early behavioural threshold levels (T-levels) were also investigated. The study data comprised intra-operative recordings in 60 children. Age at implantation was less than five years in 42 (70%) of the children. Aetiology of deafness was congenital in the majority of children (55, 92%), meningitic in four children and of unknown origin in one child. Optimum test parameters for the intra-operative EAP were an amplifier gain of 40 dB and a delay of 50 micros in order to minimize the effects of amplifier saturation due to stimulus artefact and to maximize the identification of the N1 component. An intra-operative protocol was established which involved recording four stimulus levels on each of the 22 electrodes of the electrode array, the range of stimulus levels being tailored towards the expected EAP thresholds and T-levels so as to identify response threshold. There was significant correlation between the intraoperative EAP thresholds and the early T-levels (Pearson's r = 0.93 ;p<0.01) when a correction factor was introduced based on a reliable behavioural measure of the threshold of electrical stimulation on electrode 10. The intra-operative EAP threshold, when combined with a limited amount of behavioural data, may therefore be used to predict the T-level with a useful degree of accuracy. This result is also supported by the significant correlation observed between the intra-operative thresholds of the EAP and EABR.
The minimum age for cochlear implantation has been reduced to 12 months in an effort to provide auditory stimulation to children with hearing loss during early development. Because behavioral measures in such young children are limited, objective measures such as the electrically evoked compound action potential (EAP) from the auditory nerve are needed to facilitate measurement of stimulation level requirements. We assessed EAPs recorded by the Nucleus 24 neural response telemetry (NRT) system in children who underwent implantation between 12 and 24 months of age. We recorded EAPs in 37 such children (mean age at implantation, 18.1+/-3.6 months). The EAPs were of large amplitude, and thresholds fell between behavioral T and C levels. A correction factor applied to EAP thresholds provided useful predictions of T levels. The EAPs can be used to ensure that even very young children receive auditory stimulation with their cochlear implants upon device activation.
The main aim of this study was to validate a new technique, neural response telemetry (NRT), for measuring the electrically evoked compound action potential in adult cochlear implant users via their Nucleus C124M implant. Thirty-eight adults were evaluated with a variety of measurement procedures with the NRT software. Electrically evoked compound action potentials were obtained in 31 of the 38 adults (81.6%) and in 132 of the 160 electrodes (82.5%) tested. In addition to validating this technique, we also established a set of default clinical test parameters.
This paper describes a brainstem auditory evoked potentials (BAEPs) detection method based on supervised pattern recognition. A previously used pattern recognition technique relying on cross-correlation with a template was modified in order to include a priori information allowing detection accuracy. Reference is made to the patient's audiogram and to the latency-intensity (LI) curve with respect to physiological mechanisms. Flexible and adaptive constraints are introduced in the optimization procedure by means of eight rules. Several data samples were used in this study. The determination of parameters was performed through 270 BAEPs from 20 subjects with normal and high audiometric thresholds and through additional BAEPs from 123 normal ears and 14 ears showing prominent wave VI BAEPs. The evaluation of the detection performance was performed in two steps: first, the sensitivity, specificity and accuracy were estimated using 283 BAEPs from 20 subjects showing normal and high audiometric thresholds and secondly, the sensitivity, specificity and accuracy of the detection and the accuracy of the response threshold were estimated using 213 BAEPs from 18 patients in clinic. Taking into account some a priori information, the accuracy in BAEPs detection was enhanced from 76 to 90%. The patient response thresholds were determined with a mean error of 5 dB and a standard deviation error of 8.3 dB. Results were obtained using experimental data; therefore, they are promising for routine use in clinic.
A learning algorithm for multilayer feedforward networks, RPROP
(resilient propagation), is proposed. To overcome the inherent
disadvantages of pure gradient-descent, RPROP performs a local
adaptation of the weight-updates according to the behavior of the error
function. Contrary to other adaptive techniques, the effect of the RPROP
adaptation process is not blurred by the unforeseeable influence of the
size of the derivative, but only dependent on the temporal behavior of
its sign. This leads to an efficient and transparent adaptation process.
The capabilities of RPROP are shown in comparison to other adaptive
techniques
Objective detection of brainstem auditory evoked potentials with a priori informa-tion from higher presentation levels Automatic analysis of auditory nerve electrically evoked compound action potential with an ANN
- Adam E O Vannier
- Motsch
- Jf
Vannier E, Adam O, Motsch JF. Objective detection of brainstem auditory evoked potentials with a priori informa-tion from higher presentation levels. Artif Intell Med 2002;25(3):283—301. Automatic analysis of auditory nerve electrically evoked compound action potential with an ANN
Neural Response Telemetry (version 2.01). Swirtzerland: Zurick
- N Dillier
- W K Lai