Abstract:

Abstract Over the past several years, a wide variety of ophthalmological disorders have seen an increase in the number of applications of machine learning. Inherited retinal disorders (IRC) are uncommon genetic conditions that present a distinct phenotype on fundus autofluorescence (FAF) imaging. Using machine learning with Bio- inspired algorithms; our goal was to develop a system that could automatically categorize distinct IRCs based on the appearance of FAF images. FAF imaging was used to conduct a retrospective study on patients who had presented to the Ophthalmology Department at the University of Paris Est Creteil between April 2007 and April 2019. Image preprocessing with Z-score normalization and Gabor Filter Bank (GFB) were utilized in order to glean characteristics from the images. The categorization was completed with the help of an innovative method called Chaotic Vortex Search Optimized Naive Bayes (CVSONB). Regarding the classifiers for inherited retinal diseases, performance metrics such as accuracy, f1 measure, sensitivity, and specificity were utilized. In this study, hereditary retinal disorders in FAF were automatically detected and classified utilizing an algorithm based on deep learning. The algorithm was developed as part of this investigation. As a result, the classifiers that were constructed demonstrated very promising outcomes. This model has the potential to become a diagnostic tool with more research and improvement, and it could also provide useful information for upcoming therapy approaches.

Keywords: