Medical image classification using convolutional neural network

A brain tumor is a disease caused by abnormal growth in brain tissue. These cancerous cells can occur in different brain parts and vary in size and shape. The rapid diagnosis and treatment plan for individuals with this disease are crucial for their well-being. Brain tumors have been identified using various imaging techniques, with magnetic resonance imaging (MRI) being the most common. This technique provides detailed information about the tumor area. MRI can determine cancerous cells' location, size, and volume in brain tissue. However, accurate diagnosis and tumor identification can be challenging in some cases. The accuracy and timing of expert diagnosis are critical for patients. This study implemented U-Net, a deep convolutional neural network, for 3D multi-class semantic brain tumor segmentation using the BraTS 2020 dataset and 3D MRI images. The developed model detected different tumor labels with a high accuracy of 0.97. The trained model achieved an IoU score of 0.75 for the training set and 0.62 for the validation set, which indicates successful segmentation. Additionally, the model can detect unseen images, which are used for testing purposes, within seconds. I anticipate that this model will significantly reduce diagnosis times for brain tumor experts.