A Low Complexity MobileNetV2 based CNN Model for Brain Tumor Detection in MRI Images

A brain tumor is a disease characterized by the growth of abnormal cells in the brain. Given their rarity and diverse appearances, predicting the survival rate of patients with brain tumors is challenging. Magnetic Resonance Imaging (MRI) is crucial for identifying the location of these tumors, but manual detection is labor-intensive and prone to errors. To enhance the classification and detection of brain tumors, we propose a model that builds on the deep CNN MobileNetV2. Our approach incorporates several image enhancement techniques, including Gaussian, median, and contour detection filters, to improve image quality. Additionally, we employ data augmentation techniques to increase the number of training samples. Our results show that the optimized and fine-tuned MobileNetV2 outperforms other CNN models and the basic MobileNetV2 model, achieving a classification and detection accuracy of 99% with fewer parameters.