FastColitisDetector-XAI: An efficient AI model utilizing sparse Autoencoder with explainable AI for ulcerative colitis diagnosis

We suggest AI framework for UC diagnosis using Sparse Autoencoders (SA) for feature extraction combined with Explainable AI (XAI) utilizing Grad-CAM to provide a higher degree of interpretability for the model. SA is applied for dimensionality reduction of medical images to efficiently encode the images and preserve vital information required for diagnosis. Features so extracted are passed to a machine learning classifier for classification for detection of UC presence. Visualizations from Grad-CAM are utilized to demarcate areas critical for disease, like regions of inflammation, ulcers and mucosal patterns, so as to achieve transparency and also allow the clinicians to see why the model did what it did. The proposed SA-XAI model greatly surpasses competing models in their respective performance in accuracy, precision, recall and F1 score with remarkable results of 98 %, 97.5 %, 96.4 % and 95 % respectively. Coupling of Sparse Autoencoders with XAI, achieves high accuracy in diagnosis and gains clinician's confidence in AI model's decision-making transparent. Methodology include:•Sparse Autoencoders to extract and condense the most salient features from medical images.•Grad-CAM to highlight significant regions, which maintains model's decision making process transparent.•Has 98 % accuracy, 97.5 % precision, 96.4 % recall and 95 % F1 score for detecting UC.