Taking Electroretinography to the Next Level

In the treatment of diseases such as glaucoma, diabetic retinopathy (DR), and macular degeneration, early detection is crucial to ensuring a successful treatment plan. But early detection can prove difficult without the onset of significant symptoms. As we know, a patient’s visual acuity can remain reasonable even as some diseases progress.

The electroretinogram (ERG) test is a useful component of ophthalmic diagnosis and screening, but typical ERG test analysis focuses on only a few distinct data points, such as the amplitude of the response or the latency or delay of the peak of the waveform. This sort of discrete analysis ignores other factors of the waveform that may be useful in early detection of the diseases mentioned above.

Because ERG response waveforms are complex and include many data values, artificial intelligence (AI) is a useful tool for analyzing them. Rather than merely focusing on a few distinct points, AI algorithms can instead analyze the entirety of the patient’s response, providing more thorough classification and evaluation of the patient than is possible with conventional analysis techniques.

AI is a rapidly growing field in software development and engineering, and is particularly powerful for analyzing large data sets. When sufficiently trained with patient data, AI programs can be effective tools in the health care field for diagnostic categorization. With this idea in mind, I joined a team that set out to create a viable AI algorithm for use in ERG testing protocols.

AI AND ERG

To construct an AI algorithm, patient data with an already established doctor diagnosis is needed. These ERG test results are input into the AI model as training data, and the model then uses these examples to generate benchmarks to classify unknown patient data as either normal or abnormal. Therefore, in order to train an AI program to accurately diagnose and screen unknown patients, a sizable quantity of established patient ERG data is needed.

In our case, the AI models were based on data from approximately 780 eyes and their corresponding protocol waveforms. All data in the model were anonymous, and no patient identifying data were stored as part of the algorithm. After the model was generated, it was considered locked and did not continue to learn in real time. Eyes in the model were designated as either healthy or abnormal based on independent examination.

The AI model consisted of a support vector machine algorithm. The AI used its model data to create a boundary between healthy and abnormal ERG results. Subsequent, unknown ERG results could then be compared against this boundary and deemed either healthy or abnormal.

Assembling the data necessary to construct the AI model is a critically important step in order to generate an effective and efficient AI algorithm. A diverse set of patient data in the model permits the AI algorithm to better learn the patterns and nuances of patient responses and will ultimately result in more accurate classification. It is equally important that the data in the AI model be correctly classified as healthy or abnormal based on independent examination. Failure to correctly label the data in the model can result in warped algorithm outputs.

PUTTING IT TO THE TEST

To study the effectiveness of the AI algorithm, the AI model was retroactively tested in 235 patients at five sites. A total of 119 of the patients were deemed normal via conventional ophthalmic analysis, including visual fields and normal IOP. A total of 104 patients had glaucomatous optic neuropathy and characteristic glaucomatous visual fields defects. The other 12 patients had mild nonproliferative DR.

These patients were tested across three protocols: pattern ERG, chromatic red/blue screening, and photopic negative response tests. Each patient was tested on only one protocol. Although the general construction and algorithm of the AI model was the same for each protocol, the data used in each model were different owing to the differing nature and parameters of each of the testing protocols (Table).

The ERG waveform responses for the patients in this study were entered into the AI algorithm separately. The AI had previously never encountered these new data in either its model or in any sort of analysis. The AI categorized each ERG waveform response as either normal, borderline, or abnormal, based on the already established model. This categorization was independent of any other analysis, whether as part of the ERG test or subsequent doctor examination.

The pattern ERG, chromatic red/blue screening (Diopsys Chromatic Flash Vision Screener, Diopsys), and photopic negative response tests were recorded with the Diopsys Nova (Diopsys) (AI is available on Diopsys units). Using the independent diagnosis of each patient, we were able to judge the effectiveness of the AI categorization based on the medical information for each patient. Across the three test protocols, the AI algorithm’s accuracy proved to be significant: each of the protocols with AI categorization approached sensitivity and specificity rates of 90% or better, representing excellent performance in the ability to correctly classify patients.

In this study, each patient was given a final classification based on the AI output for his or her particular tests. A healthy patient was deemed correctly classified if each of his or her eyes was categorized as healthy. A patient with pathology was deemed correctly classified if at least one of his or her eyes was categorized as abnormal, and the other eye was categorized as either abnormal or borderline (Table).

The AI model’s accuracy represents an enormous improvement over conventional ERG analysis. As noted above, typical ERG results focus on a few discrete points, such as amplitude of the peak or peaks or their latency. The AI algorithm, on the other hand, analyzes the entirety of the patient response in order to reach a conclusion. The computation necessary for this sort of analysis is performed in more than 600 dimensions of perspective, whereby the entirety of the patient response waveform can be encapsulated into a single point in multidimensional space. Once this analysis is performed, this datum can be compared with other patient data in the model, which were likewise analyzed when the model was constructed. In effect, the AI learns to classify patients based on this approach, and subsequent, unknown patient data are compared with the subset of data that already exists in the model. Figures 1–3 demonstrate normal and abnormal clinical ERG tests, in this case the photopic negative response.

AI HAS A LOT TO OFFER

From a practical standpoint, possessing a higher sensitivity than specificity may be an advantage for a diagnostic screening tool because this makes false negatives less likely than if the situation were reversed (ie, higher specificity than sensitivity). Because patient outcomes are strongly tied to early detection of ophthalmologic conditions, it is critical that a screening tool be able to correctly classify most abnormal responses. A false positive test, on the other hand, can be cleared using other diagnostic techniques.

The AI model used in this analysis can be further expanded by adding more patient data. Additionally, AI models such as this one can be developed for all sorts of testing protocols, including visual evoked potentials and OCT imaging. In doing so, the process for diagnosing and screening patients can be further streamlined, allowing patients to more likely receive the important treatment necessary to manage their conditions.