Brain Tumors Detection by Using Fine-Tuned Transfer Deep Learning Model

Abstract

Cancer stands as a formidable adversary in global health, claiming a significant number of lives. This insidious disease manifests when cells within our body's organs or tissues undergo uncontrolled growth, posing a threat to normal cellular function. Cancer cells demonstrate a remarkable ability to deceive the immune system, evading destruction and persisting in their harmful proliferation. Tumors, the hallmark of cancer, can be categorized into three types: cancerous, non-cancerous, and pre-cancerous, each presenting distinct challenges in diagnosis and treatment. Early detection of cancer is crucial for enhancing a patient's chances of survival. Among the diagnostic tools, magnetic resonance imaging (MRI) scans play a pivotal role in identifying tumors. However, the reliance on manual interpretation introduces the potential for human error. In the pursuit of precision and efficiency, the scientific community has shifted towards leveraging computerized techniques to aid in tumor prediction. This research work focuses on the development of an automated system for classifying brain tumors using MRI scans, employing advanced deep learning technology. The proposed model harnesses the power of a convolutional neural network (CNN) architecture, specifically MobileNetV2. Trained on a meticulously pre-processed dataset of MRI images, the model adeptly distinguishes between brain tumors and normal brain tissue. To enhance the robustness of the model and address overfitting concerns, data augmentation techniques are integrated. The results of this study demonstrate that the CNN model, based on MobileNetV2, achieves commendable accuracy, sensitivity, and specificity in classifying brain tumors. Notably, it outperforms other deep learning models, including VGG16, Xception, and ResNet50, which were included in the comparison. This advancement in automated tumor classification not only streamlines diagnostic processes but also marks a significant stride towards improving patient outcomes in the realm of cancer care.