Retinoblastoma Detection via Image Processing and Interpretable Artificial Intelligence Techniques

Abstract Retinoblastoma (RB) is a treatable ocular melanoma that is diagnosed early and subsequently cured in the United States but has a poor prognosis in low- and middle-income countries (LMICs). This study outlines an approach to aid health-care professionals in identifying RB in LMICs. Transfer learning methods were utilized for detection from fundus imaging. One hundred and forty RB+ and 140 RB-images were acquired from a previous deep-learning study. Next, five models were tested: VGG16, VGG19, Xception, Inception v3, and ResNet50, which were trained on the two-hundred-and-eighty image dataset. To evaluate these models, the Dice Similarity Coefficient (DSC) and Intersection-over-Union (IoU) were used. Explainable AI techniques such as SHAP and LIME were implemented into the best-performing models to increase the transparency of their decision-making frameworks, which is critical for the use of AI in medicine. We present that VGG16 is the best at identifying RB, though the other models achieved great levels of prediction. Transfer learning methods were effective at identifying RB, and explainable AI increased viability in clinical settings.