Katalog Plus
Bibliothek der Frankfurt UAS
Bald neuer Katalog: sichern Sie sich schon vorab Ihre persönlichen Merklisten im Nutzerkonto: Anleitung.
Dieses Ergebnis aus BASE kann Gästen nicht angezeigt werden.  Login für vollen Zugriff.

Analysis of Thyroid Gland Cytological Images Using Computer Vision

Title: Analysis of Thyroid Gland Cytological Images Using Computer Vision
Authors: Lozhkin, Ilya; Mironov, Andrey; Zaytsev, Konstantin; Garmash, Aleksander; Shifman, Boris; Abdulkhabirova, Fatima; Urusova, Lilia; Platonova, Nadezhda
Source: International Journal of Open Information Technologies; Vol 13, No 8 (2025); 13-19 ; 2307-8162
Publisher Information: International Journal of Open Information Technologies
Publication Year: 2025
Collection: International Journal of Open Information Technologies (INJOIT)
Description: The aim of this article is to explore the capabilities of modern neural networks in analyzing cytological whole slide images in svs format in accordance with the Bethesda categorization system. The article presents the results of the proposed neural network models for solving tasks of automatic segmentation and classification of both various types of individual cells and their clusters. For the diagnosis of thyroid cancer using computer vision, the following cell types are identified: Hurthle cells, cells with pseudoinclusions, C-cells, and clusters of cells forming papillary structures, shapeless structures with ordered and unordered cell arrangements. The proposed models have demonstrated their effectiveness in solving tasks related to the intelligent analysis of both whole-slide cytological images and tiled image segmentation. The results obtained for the segmentation of individual cells are as follows: mean Dice coefficient (DC) = 90.9% for pseudoinclusions, DC = 86.2% for Hurthle cells, DC = 90.2% for C-cells. For cell cluster segmentation, the mean Intersection over Union (IoU) is 84%, and DC is 91%. The classification accuracy of cell clusters into three classes is 77.9%.
Document Type: article in journal/newspaper
File Description: application/pdf
Language: English
Relation: http://injoit.org/index.php/j1/article/view/2245/1946; http://injoit.org/index.php/j1/article/downloadSuppFile/2245/1024; http://injoit.org/index.php/j1/article/downloadSuppFile/2245/1025; http://injoit.org/index.php/j1/article/downloadSuppFile/2245/1026; http://injoit.org/index.php/j1/article/view/2245
Availability: http://injoit.org/index.php/j1/article/view/2245
Rights: Copyright (c) 2025 International Journal of Open Information Technologies
Accession Number: edsbas.2FE65230
Database: BASE