A hybrid approach to segment and detect brain abnormalities from MRI scan

Abstract

The Detection of brain abnormality is a complex task. The images captured from the MRI scan machines have numerous information, and it is difficult to segment the appropriate information from the images. Earlier studies have shown various challenges in detecting brain abnormalities through image processing. And one among them is the segmentation of the appropriate region with abnormalities. Thus, the paper proposes a hybrid approach using GAN, K-means clustering, and MobileNet to detect brain abnormalities. Here, the GAN is used to generate fake images from the real images of MR scans. This fake image has been enhanced, and it supports segmenting the abnormal zone from the image using K-means clustering. The segmented region is identified and analyzed for the abnormalities using the MobileNet. Finally, the proposed model could detect abnormalities from the MR images of the brain. The performance metrics for the proposed model are measured and compared, which indicates the improved performance of the proposed model.