Raqamli Transformatsiya va Sun’iy Intellekt
DERMASKOPIK TASVIRLARDAGI O‘ZGARUVCHAN QALINLIKDAGI SOCH VA SOYA ARTEFAKTLARINI BARTARAF ETISHNING KO‘P MIQYOSLI TRANSFORMER HAMDA QIRRALARNI SAQLOVCHI QAYTA TIKLASH ALGORITMI
1 Toshkent axborot texnologiyalari universiteti
Keywords
Abstract
Dermaskopik tasvirlar yordamida teri o‘smalari, jumladan, melanomaga xos o‘zgarishlarni avtomatlashtirilgan tahlil qilishda soch tolalari va ularning mikro-soyalari pigmentlangan soha chegaralarining aniqligini pasaytiruvchi, to‘qima relyefini yashiruvchi hamda neyron tarmoqlarida soxta gradientlarni yuzaga keltiruvchi asosiy omillardan hisoblanadi. Mazkur tadqiqotda axborot modellarini yaratish, tasvirlarni tahlil qilish algoritmlari va timsollarni tanib olish metodologiyasi doirasida yangi ikki bosqichli qayta ishlash algoritmi ishlab chiqildi. Birinchi bosqichda ko‘p miqyosli Swin Transformer V2 asosida soch tolalari va ularning yorug‘lik yo‘nalishiga bog‘liq holda hosil bo‘ladigan soyalarini yuqori aniqlikda segmentatsiya qiladigan dinamik niqoblash modeli taklif qilingan. Ikkinchi bosqich lezyonning chekka belgilarini saqlovchi chekka-sezgir Furye-Gibrid inpainting (Edge-Preserving Frequency Inpainting) operatori taklif qilingan. Algoritm xalqaro ochiq HAM10000 va PH2 benchmark ma’lumotlar to‘plamlarida har tomonlama eksperimental sinovdan o‘tkazildi. Eksperimentlar natijasida taklif etilgan algoritm soch va soya segmentatsiyasida Dice koeffitsiyenti 0,958 ni, IoU ko‘rsatkichi esa 0,919 ni tashkil etdi. Qayta tiklangan tasvirlar sifatini baholash natijasida PSNR 32.45 dB, SSIM 0.931 va LPIPS 0.162 ga yetdi. Mazkur qiymatlar taqqoslangan an’anaviy morfologik va konvolyutsion neyron tarmoq usullarining maqolada keltirilgan ko‘rsatkichlariga nisbatan yuqoriroq natijalarni ifodalaydi. Shuningdek, dermatologik ABCD mezonlariga xos atipik pigment tarmog‘i va chegaraviy asimmetriya strukturasini saqlab qolishga imkon qadar e’tibor qaratildi.
References
- S. Mocellin, C. R. Rossi, “The Melanoma Molecular Map Project,” Melanoma Research, vol. 18, no. 3, pp. 163–165, June 2008, doi: 10.1097/CMR.0b013e328300c50b.
- L. Jütte, H. Patel, B. Roth, “Advancing dermoscopy through a synthetic hair benchmark dataset and deep learning-based hair removal,” Journal of Biomedical Optics, vol. 29, no. 11, pp. 1–17, Nov. 2024, doi: 10.1117/1.JBO.29.11.116003.
- G. Gulmirzaeva, R. Hudec, B. Akbaraliev, B. Samandarov, “An Optimal Deep Hybrid Framework with Selective Kernel U-Net for Skin Lesion Detection and Classification,” Bioengineering, vol. 13, no. 4, pp. 427, Apr. 2026, doi: 10.3390/bioengineering13040427.
- G. Randi et al., “The European Cancer Information System: exploring linkages between indoor radon concentrations and data on cancer burden,” Journal of the European Radon Association, Mar. 2022, doi: 10.35815/radon.v3.7607.
- J. K. Winkler et al., “Association between different scale bars in dermoscopic images and diagnostic performance of a market-approved deep learning convolutional neural network for melanoma recognition,” European Journal of Cancer, vol. 145, pp. 146–154, Mar. 2021, doi: 10.1016/j.ejca.2020.12.010.
- T. Mendonca, P. M. Ferreira, J. S. Marques, A. R. S. Marcal, J. Rozeira, “PH2 - A dermoscopic image database for research and benchmarking,” 2013 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Osaka: IEEE, July 2013, pp. 5437–5440. doi: 10.1109/EMBC.2013.6610779.
- Z. Liu et al., “Swin Transformer V2: Scaling Up Capacity and Resolution,” 2021, doi: 10.48550/ARXIV.2111.09883.
- P. Tschandl, C. Rosendahl, H. Kittler, “The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions,” Scientific Data, vol. 5, no. 1, pp. 180161, Aug. 2018, doi: 10.1038/sdata.2018.161.
- Z. Wang et al., “Influence of hair presence on dermoscopic image analysis by AI in skin lesion diagnosis,” Computers in Biology and Medicine, vol. 183, pp. 109335, Dec. 2024, doi: 10.1016/j.compbiomed.2024.109335.
- G. A. Gulmirzaeva, “Using semantic indexing and thesauri in processing dermoscopic images,” Современные инновации, системы и технологии - Modern Innovations, Systems and Technologies, vol. 5, no. 2, pp. 3071–3076, June 2025, doi: 10.47813/2782-2818-2025-5-2-3071-3076.
- G. A. qizi Gulmirzayeva, “Automatic segmentation of malignant skin tumors in dermoscopic images using U-Net architecture,” Raqamli Transformatsiya va Sun’iy Intellekt ilmiy jurnali, vol. 3, no. 5, pp. 104–110, Oct. 2025.
- A. H. Khan, D. N. F. A. Iskandar, J. F. Al-Asad, S. El-Nakla, “Classification of Skin Lesion with Hair and Artifacts Removal using Black-hat Morphology and Total Variation,” International Journal of Computing and Digital Systems, vol. 10, no. 1, pp. 597–604, May 2021, doi: 10.12785/ijcds/100157.
- E. Xie, W. Wang, Z. Yu, A. Anandkumar, J. M. Alvarez, P. Luo, “SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers,” 2021, arXiv. doi: 10.48550/ARXIV.2105.15203.
- R. Suvorov et al., “Resolution-robust Large Mask Inpainting with Fourier Convolutions,” 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Waikoloa, HI, USA: IEEE, Jan. 2022, pp. 3172–3182. doi: 10.1109/WACV51458.2022.00323.
- R. Kasmi et al., “SharpRazor: Automatic removal of hair and ruler marks from dermoscopy images,” Skin Research and Technology, vol. 29, no. 4, pp. e13203, Apr. 2023, doi: 10.1111/srt.13203.
- W. Li, A. N. Joseph Raj, T. Tjahjadi, Z. Zhuang, “Digital hair removal by deep learning for skin lesion segmentation,” Pattern Recognition, vol. 117, pp. 107994, Sept. 2021, doi: 10.1016/j.patcog.2021.107994.
- T. Lee, V. Ng, R. Gallagher, A. Coldman, D. McLean, “Dullrazor®: A software approach to hair removal from images,” Computers in Biology and Medicine, vol. 27, no. 6, pp. 533–543, Nov. 1997, doi: 10.1016/S0010-4825(97)00020-6.