بهبود قطعه‌بندی تصاویر پزشکی با استفاده از یادگیری ماشین: نقش بهینه‌سازی ویژگی‌های استخراج‌شده در شبکه‌های عصبی کانولوشنی
محورهای موضوعی : فناوری های نوین در سیستم های توزیع شده و محاسبات الگوریتمی
مسلم کاویانی
1
,
ام الکلثوم شهریاری
2
*
1 - دانشجوی دکتری، دانشگاه آزاد اسلامی، واحد سنندج، سنندج،
2 - استادیار، عضو هیئت علمی دانشگاه آزاد اسلامی، واحد سنندج، سنندج،
کلید واژه: یادگیری ماشین, قطعه‌بندی تصاویر پزشکی, یادگیری فدرال, معماری قطعه‌بندی عمیق با فشردگی ویژگی‌ها,
چکیده مقاله :
در سال‌های اخیر، شبکه‌های عصبی کانولوشنی مبتنی بر یادگیری عمیق به پیشرفت‌های چشمگیری، به‌ویژه در حوزه قطعه‌بندی تصاویر پزشکی، دست یافته‌اند. با این حال، عواملی مانند طراحی یکنواخت لایه‌ها، بهره‌گیری ناکافی از اطلاعات چندمقیاسی و افزایش پیچیدگی مدل‌ها به دلیل تعداد بالای پارامترها، عملکرد این روش‌ها را در شرایط پیچیده محدود کرده است. افزون بر این، امنیت داده‌ها و حفظ حریم خصوصی از چالش‌های اساسی در پردازش داده‌های پزشکی محسوب می‌شوند.
در این پژوهش، یک مدل کانولوشنی رمزگذار-رمزگشا بهبودیافته معرفی شده است که از استراتژی‌هایی برای بهینه‌سازی ویژگی‌های استخراج‌شده و کاهش تعداد پارامترها بهره می‌برد. این مدل با به‌کارگیری سازوکارهای حفظ اطلاعات پایه و ماژول‌های توجه متراکم، توانایی استخراج اطلاعات چندسطحی را ارتقا داده و به‌عنوان یک گزینه بهینه برای استفاده در ساختار یادگیری فدرال، امنیت و کارایی بیشتری را ارائه می‌دهد. ارزیابی مدل پیشنهادی بر روی مجموعه داده ClinicDB-CVC نشان می‌دهد که این روش در مقایسه با سایر روش‌های پیشرفته، عملکرد بهتری را از نظر معیار میانگین تقاطع بر روی اتحاد ارائه می‌دهد.
In recent years, deep learning-based convolutional neural networks have made significant progress, especially in the field of medical image segmentation. However, factors such as uniform layer design, insufficient utilization of multi-scale information, and increased model complexity due to the large number of parameters have limited the performance of these methods in complex situations. In addition, data security and privacy are major challenges in medical data processing. In this study, an improved convolutional encoder-decoder model is introduced that uses strategies to optimize the extracted features and reduce the number of parameters. By employing basic information preservation mechanisms and dense attention modules, this model improves the ability to extract multi-level information and offers greater security and efficiency as an optimal option for use in a federated learning structure. Evaluation of the proposed model on the ClinicDB-CVC dataset shows that this method provides better performance in terms of intersection mean criteria on alliance and Dice coefficient compared to other state-of-the-art methods.
[1] Abdel-Nabi. Heba, Ali. Mostafa and Awajan. Arafat, “A comprehensive review of the deep learning-based tumor analysis approaches in histopathological images: segmentation, classification and multi-learning tasks,” Cluster ComputingVolume 26, Issue 5, Pages 3145 - 3185October 2023 , doi: 10.1007/s10586-022-03951-2
[2] Manju A. Arivukarasi, “AEDAMIDL: An Enhanced and Discriminant Analysis of Medical Images using Deep Learning,” Proceedings of the 3rd International Conference on Smart Technologies in Computing, Electrical and Electronics, ICSTCEE 2022Bengaluru16 December 2022through 17 December 2022 , doi: 10.1109/ICSTCEE56972.2022.10100240
[3] Yang. Mengzhu, Wang. Yongfang and Li, Guoqiang, “SU-Net: A retinal segmentation model based on improved U-Net network,” ACM International Conference Proceeding SeriesPages 946 - 95016 December 2022, doi: 10.1145/3584376.3584545
[4] Shu. Xiu, Yang. Yunyun and Liu. Jun, “ALVLS: Adaptive local variances-Based levelset framework for medical images segmentation,” Pattern RecognitionVolume 136April 2023, doi: 10.1016/j.patcog.2022.109257
[5] Hussain. Tahir and Shouno. Hayaru, “MAGRes-UNet: Improved Medical Image Segmentation Through a Deep Learning Paradigm of Multi-Attention Gated Residual U-Net,” IEEE AccessOpen AccessVolume 12, Pages 40290 – 403102024, doi: 10.1109/ACCESS.2024.3374108
[6] Choubineh. Abouzar, Chen. Jie and Coenen. Frans, “A Quantitative Insight Into the Role of Skip Connections in Deep Neural Networks of Low Complexity: A Case Study Directed at Fluid Flow Modeling,” Journal of Computing and Information Science in EngineeringVolume 23, Issue 1February 2023, doi: 10.1115/1.4054868
[7] Saoudi. Rania, Boudechiche. Djameleddine and Messali. Zoubeida, “Brain MRI Scans Super-Resolution With Wavelet and Attention Mechanisms,” 2nd International Conference on Electrical Engineering and Automatic Control, ICEEAC 2024, doi: 10.1109/ICEEAC61226.2024.10576395
[8] Caicedo. Juan et al, Cimini, Beth A, “Nucleus segmentation across imaging experiments: the 2018 Data Science Bowl,” Nature MethodsOpen AccessVolume 2019, doi: 10.1038/s41592-019-0612-7
[9] Zhang. Rong, Zhang. Rongguo and Ma. Jiechao, “Analysis of different encoder-decoder-based approaches for biomedical imaging segmentation,” ACM International Conference Proceeding SeriesPages 105 - 11320 November 2020 Article number 34493206th International Conference on Robotics and Artificial Intelligence, ICRAI 2020, doi: 10.1145/3449301.3449320
[10] Kaur Buttar. Preetpal and Sachan. Manoj Kumar, “Semantic segmentation of clouds in satellite images based on U-Net++ architecture and attention mechanism,” Expert Systems with ApplicationsVolume 20915 December 2022, doi: 10.1016/j.eswa.2022.118380
[11] Xu. Hanwen, Tang. Xinming and Yang. Fanlin, “Feature-Selection High-Resolution Network With Hypersphere Embedding for Semantic Segmentation of VHR Remote Sensing Images,” IEEE Transactions on Geoscience and Remote SensingVolume 602022, doi: 10.1109/TGRS.2022.3183144
[12] Jha. Debesh and Riegler. Michael A, “DoubleU-Net: A deep convolutional neural network for medical image segmentation,” Proceedings - IEEE Symposium on Computer-Based Medical SystemsOpen AccessVolume 2020-July, Pages 558 – 564, doi: 10.1109/CBMS49503.2020.00111
[13] Shen. Zhixi and Liu. Yong, “A novel connectivity of deep convolutional neural networks,” Proceedings - 2017 Chinese Automation Congress, CAC 2017Volume 2017-January, Pages 7779 - 778329 December 2017, doi: 10.1109/CAC.2017.8244187
[14] Jha. Debesh et al, “A Comprehensive Study on Colorectal Polyp Segmentation with ResUNet++, Conditional Random Field and Test-Time Augmentation,” IEEE Journal of Biomedical and Health InformaticsOpen AccessVolume 25, Issue 6, Pages 2029 - 2040June 2021, doi: 10.1109/JBHI.2021.3049304
[15] Daza. Laura and Gómez. Catalina, “Cerberus: A Multi-headed Network for Brain Tumor Segmentation,” Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)Volume 12659 LNCS, Pages 342 – 3512021, doi: 10.1007/978-3-030-72087-2_30
[16] Sabrowsky. Hirsch, Bertram. Thumfart, “A Content-Driven Architecture for Medical Image Segmentation,” ACM International Conference Proceeding SeriesPages 89 - 9627 November 2020, doi: 10.1145/3442555.3442570
[17] Wang. Zekun, Liu. Peter X and Zou, Yanni, “Hybrid dilation and attention residual U-Net for medical image segmentation,” Computers in Biology and MedicineVolume 134July 2021, doi: 10.1016/j.compbiomed.2021.104449
[18] Lu. Chengsong, “Performance analysis of attention mechanism and teacher forcing ratio based on machine translation,” Journal of Physics: Conference SeriesOpen AccessVolume 2580, Issue 12023, doi: 10.1088/1742-6596/2580/1/012006
[19] Basyal. Ganga et al, “Comparative study of CNN models for brain tumor classification: Computational efficiency versus accuracy,” 27th Annual Americas Conference on Information Systems, AMCIS 2021, ISBN: 978-173363258-4
[20] Li. Daihui, Ma. Chengxu and Zeng. Shangyou, “Design of efficient convolutional neural module based on an improved module,” Advances in Science, Technology and Engineering SystemsOpen AccessVolume 5, Issue 1, Pages 340 – 3452020,doi: 10.25046/aj050143
[21] Pengyu. Li et al, “Improving CNN Model for Residential Building Image Classification: Enhancing Parameter Estimation Accuracy Through Transfer Learning and Reducing Model Complexity with MobileNet,” Proceedings - 2023 3rd International Signal Processing, Communications and Engineering Management Conference, ISPCEM 2023Pages 50 – 542023, doi: 10.1109/ISPCEM60569.2023.00016
[22] Vats. Satvik et al, ” Advanced Image Classification on Intel Datasets Using Optimized EfficientNet and MobileNetV2,” 2024 IEEE 9th International Conference for Convergence in Technology, I2CT 2024, doi: 10.1109/I2CT61223.2024.10543649
[23] Zhou. Yingzi, Huang. Kun and Guo, Xiaoying, “End-to-end deep residual network for semantic segmentation,” Journal of Physics: Conference SeriesOpen AccessVolume 1684, Issue 130 November 2020, doi: 10.1088/1742-6596/1684/1/012053