Pixel-wise Relevance Propagation for Detailed Insights in Brain Tumour Classification
Year: 2025
Type: conference-paper
Abstract: Medical image processing is an integral part of the modern healthcare system, offering advanced techniques to analyse, interpret, and visualise medical images acquired from various imaging modalities.XAIstrategies aim to provide a transparent and easily understandable knowledge of the decision-making process of AI models, hence improving their usability and dependability in healthcare applications. As the healthcare industry continues to embrace AI, integrating XAI into medical image analysis is poised to revolutionise how diseases are detected, diagnosed, and treated. The current work focuses on brain tumour classification using MRI images. LRP, an XAI algorithm, explains the classified images pixel-wise. Layerwise Relevance Propagation (LRP) works on the principle of the relevance of an input for a deep neural network. The VGG 19 model accurately detects and classifies tumours using LRP, outperforming other models with a 95.11% accuracy rate. We have pooled the following datasets: Figshare, Br35H and SARTAJ, with7023 images to run over our model.
Source: Procedia Computer Science
Institution: KLE Technological University
Language: English
Subfield: Neurology
Field: Neuroscience
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W4410258114