ارائه یک چارچوب یکپارچه مبتنی بر پیش‌پردازش‌های هدفمند و یادگیری عمیق جهت حذف نویز و افزایش مقاومت در برابر نمونه‌های متخاصم در تصاویر MR

مقالات آماده انتشار

نوع مقاله : مقاله پژوهشی

نویسنده

استادیار، گروه مهندسی کامپیوتر، دانشگاه ملی مهارت، تهران، ایران

چکیده

شبکه‌های عصبی کانولوشنی مورداستفاده در تحلیل تصاویر پزشکی نسبت به نویز و اغتشاشات متخاصم حساس‌اند. در این پژوهش، چارچوبی چندمرحله‌ای برای طبقه‌بندی دودویی برش‌های تشدید مغناطیسی مغز به دو کلاس دارای تومور قابل‌مشاهده (Tumor-positive) و فاقد تومور قابل‌مشاهده (Tumor-negative) ارائه‌شده است. این چارچوب، پیش‌پردازش غیرخطی، قطعه‌بندی مبتنی بر واترشید، استخراج ویژگی، کاهش بُعد با استفاده از تحلیل مؤلفه‌های اصلی، انتخاب ویژگی مبتنی بر الگوریتم ژنتیک و طبقه‌بندی با شبکه عصبی کانولوشنی را با یکدیگر ترکیب می‌کند. چارچوب پیشنهادی روی ۳۹۲۹ برش متعلق به ۱۱۰ بیمار از مجموعه داده TCGA-LGG/TCIA و با استفاده از اعتبارسنجی متقاطع پنج‌تایی در سطح بیمار ارزیابی شد. مدل روی تصاویر پاک به Accuracy برابر با 23/97٪، F1- score برابر با 22/96٪ و سطح زیر منحنی مشخصه عملکرد گیرنده (AUC) برابر با 984/0 دست‌یافت. پایداری مدل به‌صورت جداگانه تحت حمله روش علامت گرادیان سریع (FGSM) و در سه سطح اغتشاش ارزیابی شد. یافته‌ها مؤید کارایی این زنجیره پردازشی برای طبقه‌بندی برش‌های متعلق به بیماران دیده‌نشده هستند؛ بااین‌حال، پیش از تعمیم بالینی نتایج، انجام اعتبارسنجی خارجی ضروری است.

کلیدواژه‌ها

موضوعات

عنوان مقاله [English]

An Integrated Framework Based on Targeted Pre-processing and Deep Learning for Denoising and Enhancing Robustness Against Adversarial Attacks in MR Images

نویسنده [English]

  • Farnaz Hoseini
Assistant Professor, Department of Computer Engineering, Technical and Vocational University (TVU), Tehran, Iran
چکیده [English]

Convolutional neural networks used in medical image analysis are sensitive to noise and adversarial perturbations. This study presents a multistage framework for binary classification of brain magnetic resonance slices as tumor-positive, indicating a visible tumor, or tumor-negative, indicating no visible tumor. The framework integrates nonlinear preprocessing, watershed segmentation, feature extraction, dimensionality reduction using principal component analysis, genetic-algorithm-based feature selection, and convolutional neural network classification. It was evaluated on 3,929 slices from 110 patients in the TCGA-LGG/TCIA dataset using patient-wise five-fold cross-validation. On clean images, the model achieved 97.23% accuracy, a 96.22% F1-score, and an area under the receiver operating characteristic curve (AUC) of 0.984. Robustness was assessed separately under the fast gradient sign method (FGSM) attack at three perturbation levels. The findings support the utility of this processing pipeline for classifying slices from unseen patients; however, independent external validation remains necessary before clinical generalization to other populations or imaging environments

کلیدواژه‌ها [English]

  • Deep Neural Networks
  • Adversarial Examples
  • Magnetic Resonance Imaging (MRI)
  • Image Denoising
  • Genetic Algorithm-based Feature Optimization
  • Medical Image Classification

مراجع

[1]     Joel, M. Z., Avesta, A., Yang, D. X., Zhou, J.-G., Omuro, A., Herbst, R. S., Krumholz, H. M., & Aneja, S. (2023). Comparing detection schemes for adversarial images against deep learning models for cancer imaging. Cancers, 15(5), 1548. https://doi.org/10.3390/cancers15051548
[2]     Zhu, B., Liu, J. Z., Cauley, S. F., Rosen, B. R., & Rosen, M. S. (2018). Image reconstruction by domain-transform manifold learning. Nature, 555(7697), 487–492. https://doi.org/10.1038/nature25988
[3]     Szegedy, C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I. J., & Fergus, R. (2014). Intriguing properties of neural networks. International Conference on Learning Representations. https://arxiv.org/abs/1312.6199
[4]     Finlayson, S. G., Bowers, J. D., Ito, J., Zittrain, J. L., Beam, A. L., & Kohane, I. S. (2019). Adversarial attacks on medical machine learning. Science, 363(6433), 1287–1289. https://doi.org/10.1126/science.aaw4399
[5]     Goodfellow, I. J., Shlens, J., & Szegedy, C. (2015). Explaining and harnessing adversarial examples. International Conference on Learning Representations. https://arxiv.org/abs/1412.6572
[6]     Kurakin, A., Goodfellow, I. J., & Bengio, S. (2017). Adversarial examples in the physical world. International Conference on Learning Representations Workshop. https://arxiv.org/abs/1607.02533
[7]     Madry, A., Makelov, A., Schmidt, L., Tsipras, D., & Vladu, A. (2018). Towards deep learning models resistant to adversarial attacks. International Conference on Learning Representations. https://arxiv.org/abs/1706.06083
[8]     Papernot, N., McDaniel, P., Goodfellow, I. J., Jha, S., Celik, Z. B., & Swami, A. (2017). Practical black-box attacks against machine learning. In Proceedings of the 2017 ACM Asia Conference on Computer and Communications Security (pp. 506–519). Association for Computing Machinery. https://doi.org/10.1145/3052973.3053009
[9]     Liu, A., Liu, X., Yu, H., Zhang, C., Liu, Q., & Tao, D. (2021). Training robust deep neural networks via adversarial noise propagation. IEEE Transactions on Image Processing, 30, 5769–5781. https://doi.org/10.1109/TIP.2021.3082317
[10] Lin, J., Gan, C., & Han, S. (2019). Defensive quantization: When efficiency meets robustness. International Conference on Learning Representations. https://arxiv.org/abs/1904.08444
[11] Xie, C., Wu, Y., van der Maaten, L., Yuille, A. L., & He, K. (2019). Feature denoising for improving adversarial robustness. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 501–509). IEEE. https://doi.org/10.1109/CVPR.2019.00059
[12] Kwon, H., Kim, Y., Yoon, H., & Choi, D. (2021). Classification score approach for detecting adversarial example in deep neural network. Multimedia Tools and Applications, 80, 10339–10360. https://doi.org/10.1007/s11042-020-09167-z
[13] Bortsova, G., González-Gonzalo, C., Wetstein, S. C., Dubost, F., Katramados, I., Hogeweg, L., Liefers, B., van Ginneken, B., Pluim, J. P. W., Veta, M., & van Tulder, G. (2021). Adversarial attack vulnerability of medical image analysis systems. Medical Image Analysis, 73, 102141. https://doi.org/10.1016/j.media.2021.102141
[14] Hoseini, F., Shamlou, S., & Ahmadi-Gharehtoragh, M. (2024). Segmentation of MR images for brain tumor detection using autoencoder neural network. Discover Artificial Intelligence, 4(1), 71. https://doi.org/10.1007/s44163-024-00180-x
[15] Chakraborty, A., Alam, M., Dey, V., Chattopadhyay, A., & Mukhopadhyay, D. (2021). A survey on adversarial attacks and defences. CAAI Transactions on Intelligence Technology, 6(1), 25–45. https://doi.org/10.1049/cit2.12028
[16] Buda, M., Saha, A., & Mazurowski, M. A. (2019). Association of genomic subtypes of lower-grade gliomas with shape features automatically extracted by a deep learning algorithm. Computers in Biology and Medicine, 109, 218–225. https://doi.org/10.1016/j.compbiomed.2019.05.002
[17] Gajula, S., & Rajesh, V. (2023). Performance analysis of classification and detection of brain tumor MRI images using ResNet50 deep residual U-Net. Journal of Iranian Medical Council, 6(1), 101–110. https://doi.org/10.18502/jimc.v6i1.11848
[18] Ramakrishnan, A. B., Sridevi, M., Vasudevan, S. K., Manikandan, R., & Gandomi, A. H. (2024). Optimizing brain tumor classification with hybrid CNN architecture: Balancing accuracy and efficiency through oneAPI optimization. Informatics in Medicine Unlocked, 44, 101436. https://doi.org/10.1016/j.imu.2023.101436
[19] Alshomrani, F. (2024). A unified pipeline for simultaneous brain tumor classification and segmentation using fine-tuned CNN and residual UNet architecture. Life, 14(9), 1143. https://doi.org/10.3390/life14091143
[20] Jolliffe, I. T., & Cadima, J. (2016). Principal component analysis: A review and recent developments. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 374(2065), 20150202. https://doi.org/10.1098/rsta.2015.0202
[21] Roodaki, A., Sotoodeh, M., & Moosavi, M. R. (2025). Genetic algorithm-based feature selection from high-dimensional descriptors for improved content-based image retrieval. IEEE Access, 13, 198034–198053. https://doi.org/10.1109/ACCESS.2025.3633906
[22] Naser, M. A., & Deen, M. J. (2020). Brain tumor segmentation and grading of lower-grade glioma using deep learning in MRI images. Computers in Biology and Medicine, 121, 103758. https://doi.org/10.1016/j.compbiomed.2020.103758
[23] Asiri, A. A., Shaf, A., Ali, T., Aamir, M., Irfan, M., Alqahtani, S., Mehdar, K. M., Halawani, H. T., Alghamdi, A. H., Alshamrani, A. F. A., & Alqhtani, S. M. (2023). Brain tumor detection and classification using fine-tuned CNN with ResNet50 and U-Net model: A study on TCGA-LGG and TCIA dataset for MRI applications. Life, 13(7), 1449. https://doi.org/10.3390/life13071449
[24] Anaya-Isaza, A., Mera-Jiménez, L., Verdugo-Alejo, L., & Sarasti, L. (2023). Optimizing MRI-based brain tumor classification and detection using AI: A comparative analysis of neural networks, transfer learning, data augmentation, and the cross-transformer network. European Journal of Radiology Open, 10, 100484. https://doi.org/10.1016/j.ejro.2023.100484
[25] Sulaiman, A., Anand, V., Gupta, S., Al Reshan, M. S., Alshahrani, H., Shaikh, A., & Elmagzoub, M. A. (2024). An intelligent LinkNet-34 model with EfficientNetB7 encoder for semantic segmentation of brain tumor. Scientific Reports, 14, 1345. https://doi.org/10.1038/s41598-024-51472-2
[26] Rai, H. M., Chatterjee, K., & Dashkevich, S. (2021). Automatic and accurate abnormality detection from brain MR images using a novel hybrid UnetResNext-50 deep CNN model. Biomedical Signal Processing and Control, 66, 102477. https://doi.org/10.1016/j.bspc.2021.102477
پژوهش های نظری و کاربردی هوش ماشینی

مقالات آماده انتشار، پذیرفته شده
انتشار آنلاین از تاریخ 29 مهر 1405

فایل ها

هم رسانی

ارجاع به این مقاله

آمار