<?xml version='1.0' encoding='utf-8'?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.2 20190208//EN" "http://jats.nlm.nih.gov/publishing/1.2/JATS-journalpublishing1.dtd">
<article article-type="research-article" dtd-version="1.2" xml:lang="ru" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><front><journal-meta><journal-id journal-id-type="issn">2518-1092</journal-id><journal-title-group><journal-title>Научный результат. Информационные технологии</journal-title></journal-title-group><issn pub-type="epub">2518-1092</issn></journal-meta><article-meta><article-id pub-id-type="doi">10.18413/2518-1092-2025-10-3-0-5</article-id><article-id pub-id-type="publisher-id">3904</article-id><article-categories><subj-group subj-group-type="heading"><subject>ИСКУССТВЕННЫЙ ИНТЕЛЛЕКТ И ПРИНЯТИЕ РЕШЕНИЙ</subject></subj-group></article-categories><title-group><article-title>&lt;strong&gt;О СЕГМЕНТАЦИИ ПОЛИПОВ С ИСПОЛЬЗОВАНИЕМ МОДЕЛИ SEGMENT ANYTHING MODEL&lt;/strong&gt;</article-title><trans-title-group xml:lang="en"><trans-title>&lt;strong&gt;ON SEGMENTATION OF POLYPS USING SEGMENT ANYTHING MODEL&lt;/strong&gt;</trans-title></trans-title-group></title-group><contrib-group><contrib contrib-type="author"><name-alternatives><name xml:lang="ru"><surname>Щетинин</surname><given-names>Евгений Юрьевич</given-names></name><name xml:lang="en"><surname>Shchetinin</surname><given-names>Eugene Yurievich</given-names></name></name-alternatives><email>riviera-molto@mail.ru</email></contrib><contrib contrib-type="author"><name-alternatives><name xml:lang="ru"><surname>Тютюнник</surname><given-names>Анастасия Александровна</given-names></name><name xml:lang="en"><surname>Tiutiunnik</surname><given-names>Anastasiia Alexandrovna</given-names></name></name-alternatives><email>tyutyunnik_aa@pfur.ru</email></contrib></contrib-group><pub-date pub-type="epub"><year>2025</year></pub-date><volume>10</volume><issue>3</issue><fpage>0</fpage><lpage>0</lpage><self-uri content-type="pdf" xlink:href="/media/information/2025/3/ИТ_НР_10_3_5.pdf" /><abstract xml:lang="ru"><p>Полипы толстой кишки являются ключевыми предикторами развития колоректального рака, и их своевременное выявление играет решающую роль в профилактике онкологических осложнений. В работе предложена усовершенствованная модель автоматической сегментации полипов Polyps-SAM2, основанная на фундаментальной архитектуре Segment Anything Model 2 (SAM2). Модель адаптирована для медицинской визуализации путём тонкой настройки с заморозкой параметров кодировщика изображений и внедрением обучаемых слоёв для обработки текстовых инструкций. Эксперименты проведены на двух общепринятых наборах данных &amp;ndash; Kvasir-Seg и CVC-ClinicDB. Polyps-SAM2 продемонстрировала высокую точность: значения метрик Dice и IoU составили 0.94 и 0.91 на Kvasir-Seg, а также 0.938 и 0.901 на CVC-ClinicDB, что превосходит или сопоставимо с современными методами сегментации. Несмотря на ограничения при обработке изображений с множественными полипами и зависимость от подсказок (например, ограничивающих рамок), предложенная модель обладает высокой обобщающей способностью и потенциалом для интеграции в клинические системы поддержки принятия решений при проведении колоноскопии.</p></abstract><trans-abstract xml:lang="en"><p>Colorectal polyps are critical precursors to colorectal cancer, and their early detection is vital for effective prevention. This study introduces Polyps-SAM2 &amp;ndash; an enhanced polyp segmentation model built upon the Segment Anything Model 2 (SAM2) foundation. Tailored for medical imaging, Polyps-SAM2 incorporates fine-tuning with a frozen image encoder and integrates trainable layers for processing textual prompts. Evaluated on two benchmark datasets &amp;ndash; Kvasir-Seg and CVC-ClinicDB &amp;ndash; the model achieves Dice and IoU scores of 0.94/0.91 and 0.938/0.901, respectively, outperforming or matching state-of-the-art segmentation approaches. While limitations remain &amp;ndash; particularly in handling images with multiple distinct polyps and reliance on user-provided prompts such as bounding boxes &amp;ndash; the model demonstrates strong generalization capabilities and significant potential for clinical deployment in computer-aided colonoscopy systems, thereby improving diagnostic accuracy and workflow efficiency for physicians.</p></trans-abstract><kwd-group xml:lang="ru"><kwd>полипы толстой кишки</kwd><kwd>сегментация</kwd><kwd>глубокое обучение</kwd><kwd>фундаментальные модели компьютерного зрения</kwd><kwd>трансформер</kwd></kwd-group><kwd-group xml:lang="en"><kwd>colon polyps</kwd><kwd>segmentation</kwd><kwd>deep learning</kwd><kwd>fundamental computer vision models</kwd><kwd>transformer</kwd></kwd-group></article-meta></front><back><ack><p>Работа выполнена при финансовой поддержке Севастопольского государственного университета, проект 42-01-09/319/2025-1.</p></ack><ref-list><title>Список литературы</title><ref id="B1"><mixed-citation>Siegel R.L., Miller K.D., Wagle N.S., Jemal A. Cancer statistics, 2023. CA: A Cancer Journal for Clinicians. 2023. 73(1), p.17-48, doi: 10.3322/caac.21763</mixed-citation></ref><ref id="B2"><mixed-citation>Antonelli G., Badalamenti M., Hassan C., Repici A. Impact of artificial intelligence on colorectal polyp detection. Best practice &amp;amp; research. Clinical gastroenterology. 2021. 52-53, 101713, doi: 10.1016/j.bpg.2020.101713</mixed-citation></ref><ref id="B3"><mixed-citation>Shchetinin E.Yu., On the effectiveness of using visual transformers in detecting abnormalities of histopathological images. Mathematical models and computer simulations. 2025. 17(1), p.46-54, doi: 10.1134/S2070048224700716</mixed-citation></ref><ref id="B4"><mixed-citation>Ronneberger O., Fischer P., Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. Medical Image Computing and Computer-Assisted Intervention &amp;ndash; MICCAI 2015. Lecture Notes in Computer Science. 2015. V. 9351, doi: 10.1007/978-3-319-24574-4_28</mixed-citation></ref><ref id="B5"><mixed-citation>Zhou Z., Rahman Siddiquee M.M., Tajbakhsh N., Liang J. UNet++: A Nested U-Net Architecture for Medical Image Segmentation. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support. DLMIA ML-CDS 2018. Lecture Notes in Computer Science. 2018.&amp;nbsp; V. 11045, doi: 10.1007/978-3-030-00889-5_1</mixed-citation></ref><ref id="B6"><mixed-citation>Kirillov A., et al., Segment Anything. 2023. URL: https://arxiv.org/abs/2304.02643</mixed-citation></ref><ref id="B7"><mixed-citation>He S., et al., Accuracy of Segment-Anything Model (SAM) in medical image segmentation tasks. 2023.&amp;nbsp; URL: https://arxiv.org/abs/2304.09324</mixed-citation></ref><ref id="B8"><mixed-citation>Dong H., Gu H., Chen Y., Yang J., Mazurowski, M. Segment anything model 2: an application to 2d and 3d medical images. 2024. URL: https://arxiv.org/abs/2408.00756</mixed-citation></ref><ref id="B9"><mixed-citation>Jha D. et al. Kvasir-SEG: A Segmented Polyp Dataset. MultiMedia Modeling. MMM 2020. Lecture Notes in Computer Science. 2020. V. 11962, doi: 10.1007/978-3-030-37734-2_37</mixed-citation></ref><ref id="B10"><mixed-citation>De La Torre R.M., Gonzalez A.A.C., Romero-Aroca A. CVC-ClinicDB: A public database for ophthalmology research. Proceedings of the 21st International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2018). 2018. pp. 430-439, doi: 10.1007/978-3-030-00934-2_49.</mixed-citation></ref><ref id="B11"><mixed-citation>Fan D-P, et al. PraNet: Parallel Reverse Attention Network for Polyp Segmentation. Proceedings of the 2020 International Conference on Medical Image Computing and Computer-Assisted Intervention. 2020. p. 263&amp;ndash;273, doi: 10.1007/978-3-030-59725-2_26</mixed-citation></ref><ref id="B12"><mixed-citation>Patel K., Bur A.M., Wang G. Enhanced U-Net: A Feature Enhancement Network for Polyp Segmentation. IEEE International Conference on Robotics and Vision (ICRV). 2021. pp. 181-188, doi: 10.1109/CRV52889.2021.00032.</mixed-citation></ref><ref id="B13"><mixed-citation>Zhang W., et al., HSNet: A hybrid semantic network for polyp segmentation. Computers in Biology and Medicine. 2022. 150, p. 106173, doi: 10.1016/j.compbiomed.2022.106173</mixed-citation></ref><ref id="B14"><mixed-citation>Srivastava A., Jha D., Chanda, S., et al. MSRF-Net: A Multi-Scale Residual Fusion Network for Biomedical Image Segmentation. IEEE Journal of Biomedical and Health Informatics. 2021. 26(5), 2252&amp;ndash;2263, doi: 10.1109/JBHI.2021.3138024</mixed-citation></ref><ref id="B15"><mixed-citation>Dong B., et al., Polyp-PVT: Polyp segmentation with pyramid vision transformers. 2021. URL: https://arxiv.org/abs/2108.06932</mixed-citation></ref><ref id="B16"><mixed-citation>Yuheng L., Mingzhe H., Xiaofeng Ya. Polyp-SAM: Transfer SAM for Polyp Segmentation. 2023. URL: https://arxiv.org/abs/2305.00293</mixed-citation></ref><ref id="B17"><mixed-citation>Chih-Hung H., Hsin-Yu W., Yu-Lun L. HarDNet-MSEG: A Simple Encoder-Decoder Polyp Segmentation Neural Network that Achieves over 0.9 Mean Dice and 86 FPS. 2021. URL: https://arxiv.org/abs/2101.07172</mixed-citation></ref></ref-list></back></article>