Review of Brain MRI Image Segmentation Based on Deep Learning Method

Abstract

MRI images are the best brain imaging modalities for identifying tumors because they provide detailed information about regions, size, shape, and volume differences. Nevertheless, there are some issues in the brain's MRI imaging, such as low image contrast, high noise, and the boundary of objects with unclear backgrounds. As a result, sharpening anomaly object before detection and analysis on brain MRI images is critical. Currently, the segmentation of MRI brain images uses the deep learning method. It has the potential to achieve impressive results with high accuracy in the future. The clinical application of these methods remains an exciting task and a challenge. The accuracy and quality of segmentation on brain MRI image abnormalities have not been well resolved and still require improvement in segmentation performance. Public datasets, such as BRATS, are widely used for comparing and benchmarking results. It is critical to use datasets to improve. Our review focuses on segmenting brain images using deep learning methods sourced from the Science Direct and IEEE Xplore databases from 2019 to 2022, which provide better development-related knowledge on image recognition issues.