Cost-Effective early warning solution for Anisocoria Eye-Disease through Optical Sensing and Machine Learning: A Preliminary Analysis

Abstract

Anisocoria is the medical term associated when one of the pupil’s radius is not equal to the other one. This often leads to disease occurrence in the human eye when it remains undetected in its "silent" early phases. Therefore, this paper proposes a prototype of a low-cost early-warning anisocoria detection system by sensing and measuring the pupil diameter in the human eye. The unprocessed human-eye images were transformed to efficiently detect the pupil’s circumference using image binarization, leveling, and Hough transform techniques. Applying the machine learning (ML) algorithms using logistic regression, the model was trained and tested on the data set consisting of 75 random eye images. The prediction accuracy achieved was 81% when tested under red, green, blue, and ambient illumination. Furthermore, the proposed method was compared with the two other image processing methods, namely the Canny edge and Daugman algorithms, for optimum selection at the pre-ML stage. This method could prove to be a cost-effective solution for early diagnosis of anisocoria vis-a-vis database production to further accurate the proposed sensor system.