Comparative Analysis of Machine Learning Clustering Methods for Electroretinogram

Electroretinogram (ERG) signals are commonly used in electrophysiological research to measure the retina's electrical responses to light stimuli. Using publicly available signals from the IEEE DataPort repository, 11 clustering algorithms were assessed, including traditional and modern methods, to categorize the ERG signals into distinct groups. The clustering results were evaluated using metrics such as the Silhouette Coefficient, Calinski-Harabasz score, and Davies-Bouldin score. The study found that the Affinity Propagation algorithm was the most effective approach in classifying ERG signals. This method uses a message-passing framework to update the examples and responsibilities of each data point based on a damping factor and the affinities between the data points. The study's findings can contribute to enhancing the classification of ERG signals, ultimately improving the accuracy of electrophysiological research.