Brain Tumor Detection Using Hardware Compatible CNN Architecture

Deep learning has been established as the best method for identifying brain tumors. While there are several complex architectures in deep learning for detecting brain tumors, many of them are intricate, prone to overfitting, and not compatible with hardware. In comparison to traditional algorithms for diagnosing MRI slices, convolutional neural networks (CNN) require less pre-processing. A new deep learning architecture has been developed using CNN for brain tumor detection to reduce overfitting and be compatible with hardware. The first step involves extracting the region of interest (ROI) followed by a three-layer CNN to detect Glioma, Meningioma, Pituitary, and non-tumor cases. When compared to other methods, the CNN model showed promising results with an accuracy rate of 96% in distinguishing between tumor and non-tumor cases, as well as achieving good precision and specificity.