Home // AICT 2019, The Fifteenth Advanced International Conference on Telecommunications // View article
Authors:
Denis Garri
Svetlana Saakyan
Inna Khoroshilova-Maslova
Alexander Tsygankov
Oleg Nikitin
Grigory Tarasov
Keywords: E-learning; artificial neural networks; pathology; uveal melanoma
Abstract:
Uveal melanoma is a malignant tumor originating from melanocytes of an eye vascular tract. Depending on the cellular composition, the tumor is classified as a spindle cell (A or B), epithelioid cell or mixed cell. The presence of epithelioid cells reflects an unfavorable vital prognosis. The study of the cellular composition of the tumor is subjective and results in disagreements about the type of individual cells in 13% of cases among qualified pathologists. The discrepancies in diagnoses are due to the use of different classifications, which can lead to an incorrect assessment of the vital prognosis and incorrect tactics of patient treatment. Machine learning can be used to objectify the criteria of pathomorphological study of uveal melanoma, but currently there are no published works on machine analysis of pathomorphological images of this type of tumors. Our solution is based on the use of conventional neural network for the classification of images of uveal melanoma cells. We obtained an average F-score value of 0.75 to differentiate spindle cells nuclei from epithelioid cells nuclei and developed a visualization interface to explain differences between various types of cells with color mark-up of cell nuclei, probability of belonging to a certain class and deconvolution maps.
Pages: 111 to 115
Copyright: Copyright (c) IARIA, 2019
Publication date: July 28, 2019
Published in: conference
ISSN: 2308-4030
ISBN: 978-1-61208-727-6
Location: Nice, France
Dates: from July 28, 2019 to August 2, 2019