Federated learning for secure medical MRI brain tumor image classification

Abstract

Magnetic resonance imaging (MRI) in medical image processing is necessary for diagnosing and classifying various types of cancer. Deep learning (DL) techniques, particularly deep Convolutional Neural Networks (CNNs), have shown remarkable success in image analysis tasks. Transfer learning (TL) employs pre-trained weights to federated learning (FL) from a pre-trained ImageNet model with thousands of image categories. Transfer learning uses the rich features of these pre-trained models to enhance the convergence speed and performance in medical MRI brain image classification. However, the sensitive nature of medical data raises concerns about data privacy and security. Federated transfer learning presents a promising approach to address these key challenges by enabling collaborative model training across multiple institutions without sharing raw data. The proposed method achieves 98.4% classification accuracy in medical MRI image analysis by leveraging federated learning and VGG-based deep CNN architecture, ensuring state-of-the-art performance and data privacy.