Automated Medical Image Analysis in Digital Mammography
Mohsen Karimi, Majid Harouni, Shadi Rafieipour · 2021
Labeling and categorizing pixels in medical images is to obtain meaningful regions, which is called medical image segmentation process and/or is to detect and localize target objects as medical abnormalities of the images, which is called image detection process. Both of these processes divide an image into different parts/regions with common features that are the most important stages for computer-aided medical image analysis, in particular related to the human tumors or blood vessels issues. The analysis of these issues is often technically difficult due to tumor-tissue similarity, tumor-surrounding tissues, partial volume effect, numerous variations in tumor shape, depth, size, and location of the mass; distribution of vessels within the tumor and related vessels, etc. However, because of the vitality of the problem, the medical analysis is usually done manually by experts that its disadvantages include high computational time and high cost. To meet these challenges, computer-aided methods are proposed to analyze medical images with high accuracy without the need for user intervention. In other words, the presented methods in medical image analysis have reduced the need for manpower and eliminated human error. The main purpose of this work is to review and discuss findings concerning the analysis of medical mammography images and retinal fundus images. Also, detection and segmentation methods in medical images will be reviewed. Then, technical researches in the diagnosis of breast tumors and retinal blood vessels are discussed.