Automated Classification of Multi-Class Brain Tumor with Lightweight CNN Model

Classification of brain tumors is crucial for accurate diagnosis because different types of tumors have different treatment requirements. Through precise classification, doctors can develop more effective treatment plans, improving patient survival rates. However, current deep-learning models utilized for brain tumor classification typically exhibit large parameter sizes and high computational complexity. Therefore, we proposed a lightweight deep learning model named PatchBT-Net to reduce computational complexity and maintain accuracy. PatchBT-Net mainly includes technologies such as patch stems and channel attention blocks, which can effectively classify brain tumors in MRI data. PatchBT-Net uses 99.98% fewer parameters than the pre-trained VGG16 model and achieves 97.55% accuracy on the CE-MRI dataset.