يک رويکرد نوين در تشخيص آلزايمر: تلفيق مدل‌هاي يادگيري عميق براي بهبود عملکرد تشخيص
محورهای موضوعی : پردازش چند رسانه ای، سیستمهای ارتباطی، سیستمهای هوشمند
بتول نادری
1
,
شهلا نعمتی
2
*
,
محمداحسان بصیری
3
1 - کارشناسي ارشد مهندسي کامپيوتر، دانشکده فني و مهندسي، دانشگاه شهرکرد، شهرکرد، ايران
2 - استاديار، گروه مهندسي کامپيوتر، دانشکده فني و مهندسي، دانشگاه شهرکرد، شهرکرد، ايران
3 - دانشيار، گروه مهندسي کامپيوتر، دانشکده فني و مهندسي، دانشگاه شهرکرد، شهرکرد، ايران
کلید واژه: بيماري آلزايمر, يادگيري عميق, يادگيري جمعي, شبکه عصبي کانولوشني,
چکیده مقاله :
بيماري آلزايمر به‌عنوان يکي از شايع‌ترين و پيچيده‌ترين اختلالات عصبي پيش‌رونده و غيرقابل‌برگشت در سنين بالا، تاکنون چالش‌هاي قابل‌توجهي براي تشخيص و درمان به همراه داشته است. اين بيماري معمولاً به‌تدريج و در طول زمان بروز مي‌کند و به‌تدريج بر حافظه، تفکر و توانايي‌هاي شناختي فرد تأثير مي‌گذارد؛ به‌طوري‌که تشخيص زودهنگام آن در بهبود کيفيت زندگي بيماران و خانواده‌هاي آن‌ها به‌شدت حياتي است. در اين مقاله يک روش يادگيري عميق جمعي براي دستهبندي تصاوير پزشکي بيماري آلزايمر پيشنهاد شده است. در اين مدل پيشنهادي از هفت مدل شبکه عصبي کانولوشني شامل Inception v3، AlexNet، MobileNet، EfficientNet، DenseNet، ResNetوNASNet به عنوان مدل‌هاي پايه استفاده‌شده است و سپس اين مدل‌ها با استفاده از الگوريتم CatBoost با هم تلفيق شده‌اند. نتايج اين مدل پيشنهادي با روش‌هاي يادگيري جمعي ديگر مانند XGBoost، AdaBoost و همچنين با تکنيک‌هاي تجميع نظير رأيگيري و ماکزيمم‌گيري مقايسه شده است. نتايج مدل پيشنهادي بر روي مجموعه داده ADNI که شامل تصاوير MRI اسکن‌هاي مغزي افراد هستند، نشان مي‌دهد که مدل پيشنهادي عملکرد بهتري در دستهبندي تصاوير پزشکي بيماري آلزايمر دارد. عملکرد مدل پيشنهادي، چه در هر يک از کلاس‌ها و چه درکل، از تمامي مدل‌هاي يادگيري عميق مستقل نيز بهتر بوده است.
بيماري آلزايمر به‌عنوان يکي از شايع‌ترين و پيچيده‌ترين اختلالات عصبي پيش‌رونده و غيرقابل‌برگشت در سنين بالا، تاکنون چالش‌هاي قابل‌توجهي براي تشخيص و درمان به همراه داشته است. اين بيماري معمولاً به‌تدريج و در طول زمان بروز مي‌کند و به‌تدريج بر حافظه، تفکر و توانايي‌هاي شناختي فرد تأثير مي‌گذارد؛ به‌طوري‌که تشخيص زودهنگام آن در بهبود کيفيت زندگي بيماران و خانواده‌هاي آن‌ها به‌شدت حياتي است. در اين مقاله يک روش يادگيري عميق جمعي براي دستهبندي تصاوير پزشکي بيماري آلزايمر پيشنهاد شده است. در اين مدل پيشنهادي از هفت مدل شبکه عصبي کانولوشني شامل Inception v3، AlexNet، MobileNet، EfficientNet، DenseNet، ResNetوNASNet به عنوان مدل‌هاي پايه استفاده‌شده است و سپس اين مدل‌ها با استفاده از الگوريتم CatBoost با هم تلفيق شده‌اند. نتايج اين مدل پيشنهادي با روش‌هاي يادگيري جمعي ديگر مانند XGBoost، AdaBoost و همچنين با تکنيک‌هاي تجميع نظير رأيگيري و ماکزيمم‌گيري مقايسه شده است. نتايج مدل پيشنهادي بر روي مجموعه داده ADNI که شامل تصاوير MRI اسکن‌هاي مغزي افراد هستند، نشان مي‌دهد که مدل پيشنهادي عملکرد بهتري در دستهبندي تصاوير پزشکي بيماري آلزايمر دارد. عملکرد مدل پيشنهادي، چه در هر يک از کلاس‌ها و چه درکل، از تمامي مدل‌هاي يادگيري عميق مستقل نيز بهتر بوده است.
Abstract
Introduction: Alzheimer's disease is a common and complex brain disorder that gradually gets worse and is irreversible. It often develops slowly, affecting memory, thinking, and the ability to carry out everyday activities. This disease has presented significant challenges for early diagnosis and treatment. Early detection is vital for improving the quality of life for both patients and their families.
Method: In this research, an ensemble deep learning model for classifying medical images related to Alzheimer's disease is proposed, utilizing seven convolutional neural networks as a foundation. The results of the study demonstrate the high performance of the proposed model compared to other independent deep learning models and traditional ensemble methods.
Results: The proposed model effectively improves the system's accuracy in identifying and classifying different stages of Alzheimer's disease using the CatBoost algorithm in ensemble stage. The F1 score achieved by the proposed model in different classes, especially in the early stages of the disease such as EMCI and MCI, indicates significant achievements. Specifically, these scores demonstrate the model's enhanced ability to accurately diagnose Alzheimer's patients at an early stage, which can have a significant impact on preventive treatments and clinical management.
Compared to independent models, the results obtained from the proposed model indicate superior performance in overall diagnostic accuracy as well as in differentiating between different stages of the disease. For instance, while other models like EfficientNet and DenseNet performed well in specific classes, the proposed model yielded better results in every class. The use of traditional ensemble methods such as voting and maximization also produced acceptable results, but clearly fell short of the performance of the proposed model.
The results also indicate that ensemble learning models like CatBoost can enhance the overall system performance by combining features from different neural networks. This is clearly evident when compared to other ensemble learning methods like AdaBoost and XGBoost.
Another noteworthy point is the importance of data preprocessing steps. Preprocessing steps such as normalization, registration, brain tissue extraction, intensity correction, and segmentation were designed to maximize the quality of the input data, which significantly impacted the final results.
Discussion: In conclusion, this study emphasizes that utilizing advanced deep learning and ensemble learning methods can contribute to the early and accurate identification of Alzheimer's disease. These achievements promise to improve medical diagnostic processes and can pave the way for future research in the field of diagnosing and treating neurological diseases.
[1] Rawat, R. M., Akram, M., & Pradeep, S. S. (2020, June). Dementia detection using machine learning by stacking models. In 2020 5th International Conference on Communication and Electronics Systems (ICCES) (pp. 849-854). IEEE.#
[2] Dashtipour, K., Taylor, W., Ansari, S., Zahid, A., Gogate, M., Ahmad, J., ... & Abbasi, Q. (2021, December). Detecting Alzheimer’s disease using machine learning methods. In EAI International Conference on Body Area Networks (pp. 89-100). #
[3] Singh, A., & Kumar, R. (2024). Brain MRI Image Analysis for Alzheimer’s Disease (AD) Prediction Using Deep Learning Approaches. SN Computer Science, 5(1), 160. #
[4] Taylor, K. (2020). Dementia: A very short introduction. Oxford University Press.#
[5] LeCun, Y. (2019, February). 1.1 deep learning hardware: Past, present, and future. In 2019 IEEE International Solid-State Circuits Conference-(ISSCC) (pp. 12-19). IEEE.#
[6] Garg, A., & Mago, V. (2021). Role of machine learning in medical research: A survey. Computer science review, 40, 100370.#
[7] Dong, X., Yu, Z., Cao, W., Shi, Y., & Ma, Q. (2020). A survey on ensemble learning. Frontiers of Computer Science, 14, 241-258.#
[8] Shukla, A., Tiwari, R., & Tiwari, S. (2023). Alzheimer’s disease detection from fused PET and MRI modalities using an ensemble classifier. Machine Learning and Knowledge Extraction, 5(2), 512-538.#
[9] Nykoniuka, M., Melnykovab, N., Paterehac, Y., Salad, D., & Cichońe, D. (2023). Classification of Patients with the Development of Alzheimer's Disease using an Ensemble of Machine Learning Models. Proceedings http://ceur-ws. org ISSN, 1613, 0073.#
[10] Chandralekha, E., Gokila, E., & Vasudevan, I. (2023, November). Exploratory Investigation and Alzheimer's Disease Classification Utilizing Ensemble Models in Machine Learning. In 2023 International Conference on Ambient Intelligence, Knowledge Informatics and Industrial Electronics (AIKIIE) (pp. 1-7). IEEE.#
[11] Fareed, M. M. S., Zikria, S., Ahmed, G., Mahmood, S., Aslam, M., Jillani, S. F., ... & Asad, M. (2022). ADD-Net: an effective deep learning model for early detection of Alzheimer disease in MRI scans. IEEE Access, 10, 96930-96951.#
[12] Ghaffari, H., Tavakoli, H., &Pirzad Jahromi, G. (2022). Deep transfer learning–based fully automated detection and classification of Alzheimer’s disease on brain MRI. The British journal of radiology, 95(1136), 20211253.#
[13] Lu, B., Li, H. X., Chang, Z. K., Li, L., Chen, N. X., Zhu, Z. C., ... & Yan, C. G. (2022). A practical Alzheimer’s disease classifier via brain imaging-based deep learning on 85,721 samples. Journal of Big Data, 9(1), 101.#
[14] Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., ... & Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1-9).#
[15] Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). "ImageNet classification with deep convolutional neural networks." Advances in Neural Information Processing Systems, 25 #
[16] ADNI Dataset. https://ida.loni.usc.edu/login.jsp?project=ADNI& page=HOME. Accessed 10 Mar 2022. #
[17] Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L. C. (2018). Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4510-4520). #
[18] Efficientnet: Rethinking model scaling for convolutional neural networks #
[19] MIPAV. https://mipav.cit.nih.gov/ #
[20] Alhichri, H., Alswayed, A. S., Bazi, Y., Ammour, N., & Alajlan, N. A. (2021). Classification of remote sensing images using EfficientNet-B3 CNN model with attention. IEEE access, 9, 14078-14094. #
[21] Ji, Q., Huang, J., He, W., & Sun, Y. (2019). Optimized deep convolutional neural networks for identification of macular diseases from optical coherence tomography images. Algorithms, 12(3), 51. #
[22] Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017). Densely connected convolutional networks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4700-4708). #
[23] Zoph, B., Vasudevan, V., Shlens, J., & Le, Q. V. (2018). Learning transferable architectures for scalable image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 8697-8710). #
[24] Saber, H. A., Younes, A., Osman, M., & Elkabani, I. (2024). Quran reciter identification using NASNetLarge. Neural Computing and Applications, 36(12), 6559-6573. #
[25] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun. Deep Residual Learning for Image Recognition. arXiv:1512.03385v1 [cs.CV] 10 Dec 2015. #
[26] Dorogush, A. V., Ershov, V., & Gulin, A. (2018). CatBoost: gradient boosting with categorical features support. arXiv preprint arXiv:1810.11363. #
[27] Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016). Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 2818-2826). #
[28] Chen, T., & Guestrin, C. (2016, August). Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining (pp. 785-794). #
[29] Li, X., Wang, L., & Sung, E. (2005, July). A study of AdaBoost with SVM based weak learners. In Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005. (Vol. 1, pp. 196-201). IEEE. #