LIMU team’s achievements in the application of neural networks in modern medical research
LIMU PhD students Krokhmal Alisa and Tyurina Anastasia have passed the course on the use of neural networks in scientific research, sponsored by the Intellect Foundation on the basis of the Physics Faculty, MSU.
The project of Krokhmal Alisa “Segmentation of tissue histology with traces of destruction as a result of histotripsy” and the one of Tyurina Anastasia “Segmentation of histological heart images using neural networks” are aimed at using neural networks in the problems of identifying the affected areas of various biological tissues and assessing the total volume of these lesions obtained as a result of noninvasive exposure to tissue by high-intensity pulsed focused ultrasound.
During their course works, Alisa and Anastasia implemented various approaches to segmentation of histological liver and heart images obtained by layer-by-layer scanning of the destroyed area. To identify the most advantageous parameters of neural networks learning, the results of their predictions were compared with manually outlined images which consisted of three types of tissue: fully damaged, partially damaged and intact areas. The constructed models of neural networks made it possible to segment the unmarked data with sufficiently high accuracy and determine the number of destructions obtained at various doses of exposure.
The projects of Alisa and Anastasia were especially noted by the teachers and the expert committee of the course: Krokhmal Alisa won the competition of scientific publications, and Tyurina Anastasia won the competition of course works.