Enhancing Brain Tumor Detection with Gradient-Weighted Class Activation Mapping and Deep Learning Techniques
Kornprom Pikulkaew
Abstract
Brain tumors are a leading source of morbidity and mortality worldwide, and early identification and precise diagnosis are essential for enhancing patient outcomes. The most frequent approach for detecting brain tumors is magnetic resonance imaging (MRI), although it can be difficult to precisely identify tumors from pictures. In this article, we present a deep-learning approach for brain tumor diagnosis using MRI data. Our method employs a Deep Convolutional Neural Network (DCNN) architecture to precisely detect and classify brain malignancies, and gradient-weighted class activation mapping (Grad-CAM) to visualize data in the brain tumor area. Our objective is to support patients with tumors or assist medical personnel with patient diagnostics. Using a dataset from Kaggle, including 2114 brain MRI images, we analyze our method that achieved a high level of accuracy (97 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> ), coupled with great precision. Our results illustrate the efficacy of our proposed method for brain tumor diagnosis, and the use of Grad-CAM enables us to observe the areas of the brain most strongly related to tumor detection, providing doctors with significant insights. Additionally, with the proper modification of parameters, our study may be utilized to identify infectious illnesses.