DeepTumorNet: A CNN-Based Multi-Class Brain Tumor Detection Approach Using MRI Scans and Segmentation of Data

Abstract

Health is fundamental to human well-being, with brain health particularly critical for cognitive functions. Magnetic resonance imaging (MRI) serves as a cornerstone in diagnosing brain health issues, providing essential data for healthcare decisions. The Brain tumors have been noted as among the most life-threatening forms of cancer whereby accurate and timely diagnosis would be very important in terms of patient survival and proper treatment response. This paper presents an analysis and a comparison of the performance of three different models of deep learning, such as Multi-Layer Perceptron (MLP), AlexNet, and InceptionV3, on brain tumor multi-class classification using magnetic resonance imaging (MRI) scans. With the help of a publicly available data that contained labeled MRI images of four categories such as glioma, meningioma, pituitary tumors, and normal (non-tumorous) cases, we also trained all the models to evaluate the performance. The data also feature segmentation data that makes the model perform better in terms of targeting the appropriate areas in the brain scan. Among the models, the MLP serves as a fundamental baseline for performance comparison. However, the convolutional neural network (CNN)-based architectures, specifically AlexNet and InceptionV3, demonstrated significantly superior results. Notably, InceptionV3, which incorporates deep and wide convolutional layers along with transfer learning capabilities, achieved the best overall accuracy at 88.57%. These findings underscore the importance of using advanced CNN models and pre-trained architectures for improving diagnostic accuracy in medical imaging tasks. By leveraging transfer learning and efficient model design, this research contributes to the development of automated systems that can assist medical professionals in detecting and classifying brain tumors more reliably and quickly.