MediVQA: Medical Visual Question Answering using VisualBERT

Medical Visual Question Answering (VQA) aims to enhance automated medical image interpretation by enabling AI systems to generate accurate responses to clinical queries. This research focuses on advancing Medical VQA, with a particular emphasis on interpreting medical images to support diagnostic and decision-making processes. Using VQA-Med-2019, ImageCLEFmed_MEDVQA and SLAKE 1.0 datasets, we developed a specialized system tailored to the unique demands of medical VQA tasks. Our approach combines textual feature extraction via the BERT model with visual feature extraction using ResNet-101. These multimodal features are integrated and processed through the MLP-integrated VisualBERT model, optimized for classification tasks. The proposed system demonstrates robust performance in generating accurate answers to complex medical queries, showcasing its potential to enhance medical diagnostics, improve clinical workflows, and ultimately contribute to better patient care outcomes.