Diabetes Detection Based on Iridology

Iridology, a holistic practice, suggests that distinct patterns and markings in the iris can offer insights into an individual's overall health and potential areas of imbalance. Practitioners examine the iris to identify subtle variations in color, texture, and structure, which are believed to correspond to specific organs and systems in the body. This study delves into the potential of iris imaging as a non-invasive tool for predicting diabetes. Through the utilization of cutting-edge machine learning techniques, particularly Graph Neural Networks (GNNs), a pioneering approach to iris-based diagnostics is formulated. The model is meticulously trained and assessed using both publicly available datasets and a combination of public and local datasets, showcasing commendable performance across various metrics such as accuracy, precision, recall, and F1-score. Notably, the most impressive results are attained with models trained on right eye images from a public dataset, achieving an average accuracy of 92% based on 16 trials. However, the study candidly acknowledges the constraints posed by the relatively modest dataset size, underscoring the imperative for larger and more diverse datasets in future investigations. Furthermore, to facilitate accessibility and usability, a prototype Flask application is developed, furnishing a user-friendly platform for iris image submission and diabetes prediction.