MediFusionNet: A Hybrid CNN-Based Multi-Disease Diagnostic Web Platform

Abstract

Disease diagnosis using medical images has become an indispensable part of healthcare today. This necessitates the development of smart and effective models. This research work proposes a MediFusionNet, a hybrid web-based CNN diagnostic tool for multi-disease diagnosis based on feature fusion between VGG16 and MobileNetV2 to enhance the classification performance for different medical imaging techniques. The proposed framework diagnoses Brain Tumors, Breast Cancer, Chest Diseases, and Lung Abnormalities from Brain MRI, Breast Ultrasound, Chest X-ray, and Lung CT datasets respectively. We apply a preprocessing phase, which consists of images' resizing, normalization, and augmentation to make the model robust. Federated learning technique is applied to train the proposed model while keeping patients' privacy intact during collaborative model training by not sharing any patient's data. The proposed model is trained and tested on the aforementioned datasets and is deployed in a web-based clinical platform for disease diagnosis, prediction, and report generation.