Search academic texts

Information × Registration Number 0224U000943, 0120U000437 , R & D reports Title "Evaluation, on the basis of artificial intelligence elements, technology for complex radiological diagnostics of diffuse liver patology in patients of different age groups". popup.stage_title Head Дикан І.М., Тарасюк Б.А., Registration Date 15-01-2024 Organization The State Institution "Institute Of Nuclear Medicine Diagnostic Radiology of National Academy of Medical Sciences of Ukrane" popup.description2  The object of the study is children and adults with diffuse liver pathology The purpose of the work: to develop computer algorithms for processing radiological images using elements of artificial intelligence, to highlight the most informative indicators characterizing pathological changes. The research methods are ultrasound in gray scale on the device "Aplio-500" (Toshiba), mathematical processing of images in B-mode. The object of the study is children and adults with cerebral palsy (60 pediatric patients and 140 adult patients). The comparison group consisted of 55 children and 45 adults. Research results. Selective algorithms have been developed for the formation of ensembles of informative features, intended for use in machine learning of classification algorithms for the purpose of distinguishing between images of normal and pathologically changed liver, based on ultrasound images in the "gray scale", also the highest efficiency in determining the stages of fibrosis was demonstrated in ultrasound images of the liver by seven logistic regression models. A number of analytical and network-like (forest) classification models were developed using artificial intelligence to solve the problem of differentiating normality and pathology in chronic diffuse liver diseases. It was proved that the most effective was the classification of binarized images, when using which the diagnostic accuracy ranged from the type of ultrasound sensor from 75.7 % (convex) to 98.4% ("enhanced" linear), sensitivity from 0.817 (convex) to 1 ("enhanced" linear) and specificity from 0.703 (convex) to 0.914 ("enhanced" linear). Thus, logistic regression was chosen as the main algorithm for the diagnostic decision support system. Product Description popup.authors Andrushchenko Iryna Babenko Vitalii Glazovska Iryna Gordiyenko Kyryl Gurando Andrii Dyba Maryna Korobko Victor Kruglyi Vyacheslav Omelchenko Oleksii Solodushchenko Volodymyr popup.nrat_date 2024-01-15 Close