Towards a Transparent and Interpretable AI Model for Medical Image Classifications