UDC 004.67
RESEARCH OF SINGLE-STAGE DEEP LEARNING NEURAL NETWORK TO CLASSIFY ACUTE LYMPHOBLASTIC LEUKEMIA VARIANTS USING OPEN DATASET
E. V. Polyakov, Ph.D. (in technical sciences), Associate Professor, Department of Medical Physics, National Research Nuclear University «MEPhI» (Moscow Engineering Physics Institute), Moscow, Russia;
orcid.org/0000-0002-5346-6504, email: This email address is being protected from spambots. You need JavaScript enabled to view it.
I. P. Churilov, Postgraduate Student, Department of Medical Physics, NRNU MEPhI, Moscow, Russia; orcid.org/0009-0003-1368-3744, email: This email address is being protected from spambots. You need JavaScript enabled to view it.
V. V. Dmitrieva, Ph.D. (in technical sciences), Associate Professor, Department of Electrophysical Installations, National Research Nuclear University «MEPhI» (Moscow Engineering Physics Institute), Moscow, Russia; orcid.org/0000-0002-9202-6691, email: This email address is being protected from spambots. You need JavaScript enabled to view it.
The article is devoted to the task of automated diagnosis of acute lymphoblastic leukemia using artificial intelligence methods. The relevance of this study stems from the need to increase the accuracy and speed of hematological disease diagnostics since traditional blood smear analysis techniques are labor-intensive and prone to human errors. This work investigates the effectiveness of a one-stage neural network for simultane ous detection and classification of blood cells into four classes: healthy cells and three subtypes of lympho blasts (Early Pre-B, Pre-B, Pro-B). Model training and testing were conducted on an open dataset contain ing more than 3000 annotated images. As a result of experiments, high-quality classification metrics have been achieved: precision – 0.97, recall – 0.95, F1-score – 0.96, mAP50 – 0.98. These results demonstrate the potential of deep neural networks in automating the diagnosis of acute lymphoblastic leukemia and allow improving the efficiency and objectivity of laboratory analysis. The proposed approach provides a more ac curate assessment of each individual element within an image compared to classifying entire snapshots.
Key words: : neural network, artificial intelligence, YOLO, acute lymphoblastic leukemia, diagnosis, clas sification, object detection, precision, recall, mAP.