EXPLAINABLE AI IN MEDICAL IMAGING: VISUAL EXPLANATION TECHNIQUES FOR DEEP LEARNING MODELS

Abstract

Explainable AI (XAI) is changing the way medical imaging is done by making deep learning methods easier to understand and more open. This study looks at how XAI methods, specifically visual explanation approaches, can be used in medical imaging tasks like finding tumors in MRI scans and figuring out what's wrong with the eye using OCT images. Medical datasets that have been labeled are used to train deep learning models like ResNet50 for finding tumors and OculusNet for diagnosing eye diseases. XAI systems like Grad-CAM and SHAP use pictures to show areas of the picture that affect model results. This helps clinicians trust and use the systems more. These visual cues help doctors understand model decisions, which makes them more likely to accept AI-based analysis. The study rates the success of the model based on its accuracy, clarity, memory, and F1 score, as well as how useful clinicians think graphic descriptions are. To find out how XAI changes model performance and perception, statistical methods like paired t-tests are used. The results show that XAI methods make deep learning models more clear, which means that healthcare professionals can trust and use them more. This research shows how important it is to combine XAI with medical imaging so that AI can be used safely and ethically in hospital settings.