Modern approaches to segmentation and analysis of brain structures: problems and solutions

Currently, artificial intelligence is one of the most rapidly developing areas of human knowledge. This topic is of great importance for science and practice, in general, and for medicine, in particular. Application of artificial intelligence technologies to the segmentation of brain areas and detection of abnormal areas is especially demanded and promising in the field of neurophysiology, neurosurgery, psychiatry, clinical psychology and other medical disciplines. This paper investigates existing methods for automated segmentation and analysis of data on the structure and functional state of the brain, as well as metrics used to evaluate the effectiveness of this approach. Materials and methods. The work was performed using Systematic Mapping Study (SMS) methodology. This study is limited to the subject area related to segmentation of brain areas and identification of abnormal areas in the brain. Results. The main results of the study are presented in the form of classification tables and a mental map. It is shown that the purpose of the reviewed research is to improve accuracy in segmenting brain areas and finding abnormal areas. Such a metric as data processing time is used to evaluate the efficiency of the method for a small number of studies, and in most cases it is not considered at all. At the same time, the speed of image processing, depending on the method used, is measured in minutes, which significantly limits the possibility of using this approach in emergency situations, including life-threatening situations. Conclusion. To analyze data on the structure and functional state of the brain in real time, modification of already developed methods of encephalic segmentation is required, as well as development of new, more efficient approaches. At the same time, the speed of data processing should be commensurate with the time of making an urgent conclusion about the state of the human brain.