Review on Deep Learning Methodologies in Medical Image Restoration and Segmentation

This paper comprehensively reviews two major image processing tasks, such as restoration and segmentation in the medical field, from a deep learning perspective. These processes are essential because restoration removes noise and segmentation extracts the specific region of interest of an image, both of which are necessary for accurate diagnosis and therapy. This paper mainly focuses on deep learning techniques. It plays a prominent role over other conventional techniques in handling a large number of datasets in the medical field and provides accurate results. This paper reviewed the application of different convolutional neural network architectures in the restoration and segmentation processes. Based on the results in the case of image restoration, TLR-CNN and Stat- CNN are promising in achieving better PSNR, noise suppression, artifact suppression and improving the overall image quality. For the segmentation process, LCP net achieves the Dice score of 98.12% and sensitivity of 98.95% in the cell contour segmentation; the 3D FCNN model is found to be the best method for the segmentation of brain tumors. This review shows that deep learning methodologies can be a better alternative for medical image restoration and segmentation tasks, as data size is an important concern today.

Automated summary

This paper presents a comprehensive review of restoration and segmentation tasks in the medical field using deep learning techniques. The paper highlights the importance of these tasks, as restoration removes noise and segmentation extracts the specific regions of interest, both of which are crucial for accurate diagnosis and therapy. Various convolutional neural network architectures are explored, and promising results are achieved in terms of image quality improvement, noise and artifact suppression, as well as accurate segmentation of cell contours and brain tumors.