研究生: |
魏旭廷 HSU-TING WEI |
---|---|
論文名稱: |
建立輕量化的腦瘤自動檢測模型與硬體設計 Establishing a Compact Automatic Brain Tumor Detection Model and Hardware Design |
指導教授: |
王煥宗
Huan-Chun Wang |
口試委員: |
林承鴻
Cheng-Hung Lin 賴坤財 Kuen-Tsair Lay 方文賢 Wen-Hsien Fang |
學位類別: |
碩士 Master |
系所名稱: |
電資學院 - 電子工程系 Department of Electronic and Computer Engineering |
論文出版年: | 2023 |
畢業學年度: | 111 |
語文別: | 中文 |
論文頁數: | 73 |
中文關鍵詞: | 卷積神經網路 、加速器 、腦瘤辨識 、深度學習 |
外文關鍵詞: | Convolutional Neural Network, Accelerator, Brain Tumor Detection, Deep Learning |
相關次數: | 點閱:314 下載:0 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
腦腫瘤是最致命的癌症之一,如果利用自動化腦瘤檢測系統來輔助醫生就能使流程更加快速,也能和人工檢測的結果做雙重確認。近年來卷積神經網路(Convolutional Neural Network, CNN)已在腦瘤檢測方面取得進展,但運算量複雜且參數量大仍是挑戰。因此本研究創建一個輕量化的腦瘤檢測模型,優化方式包括Fire module、全局平均池化及資料增強等。另外本論文提出一個基於硬體的演算法簡化,針對有批次正規化的CNN架構做計算及參數合併,能在同樣準確率下減少硬體複雜度。
最後本論文會探討腦瘤檢測系統的超大型積體電路(Very Large Scale Integration ,VLSI)設計與實作,並以降低面積及縮短延遲(Latency)為設計方向。實驗結果顯示,和AlexNet架構比較之下,本論文的架構準確率在浮點數運算降低了1.07%、定點數降低了1.23%,且GOP/s只有其1/65,但硬體使用量降低了許多,相對地此模型的泛化能力比AlexNet差,因為此模型僅限於腦瘤檢測。如果同時考慮面積及速度,Area efficiency相較2018年文獻提升了20.3%,能使用更小的面積嵌入MRI儀器中。
Brain tumors are among the most lethal types of cancer. Using automated brain tumor detection systems to assist doctors can expedite the process and provide dual confirmation with manual detection results. In recent years, Convolutional Neural Networks (CNNs) have made progress in brain tumor detection, but the computational complexity and large number of parameters pose challenges. This study creates a lightweight brain tumor detection model. Moreover, this paper proposes a hardware-based algorithm simplification that targets the calculation and parameter combination of CNN structures with batch normalization, reducing hardware complexity while maintaining the same accuracy level.
Lastly, this paper explores the design and implementation of the brain tumor detection system using Very Large Scale Integration (VLSI), circuits with the design direction focusing on reducing area and shortening latency. Experimental results show that compared with the AlexNet architecture, the accuracy of the proposed architecture decreased by 1.07%, but the hardware resource usage has decreased significantly. Conversely, the generalization ability of this model is inferior to AlexNet, as this model is solely designed for brain tumor detection. When considering both area and speed, the area efficiency increased by 20% compared to the literature in 2018.
[1] https://wd.vghtpe.gov.tw/hemaonco/files/Guide_BrainCA.pdf
[2] H. -Y. You et al., "An AMDOCT-NET for Automated AMD Detection under Evaluations of Different Image Size, Denoising and Cropping," 2021 IEEE 3rd Eurasia Conference on Biomedical Engineering, Healthcare and Sustainability (ECBIOS), Tainan, Taiwan, 2021, pp. 138-142, doi: 10.1109/ECBIOS51820.2021.9510570.
[3] M. Arbane, R. Benlamri, Y. Brik and M. Djerioui, "Transfer Learning for Automatic Brain Tumor Classification Using MRI Images," 2020 2nd International Workshop on Human-Centric Smart Environments for Health and Well-being (IHSH), Boumerdes, Algeria, 2021, pp. 210-214, doi: 10.1109/IHSH51661.2021.9378739.
[4] N. Çınar, B. Kaya and M. Kaya, "Comparison of deep learning models for brain tumor classification using MRI images," 2022 International Conference on Decision Aid Sciences and Applications (DASA), Chiangrai, Thailand, 2022, pp. 1382-1385, doi: 10.1109/DASA54658.2022.9765250.
[5] He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778).
[6] Iandola, F. N., Han, S., Moskewicz, M. W., Ashraf, K., Dally, W. J., & Keutzer, K. (2016). SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 MB model size. arXiv preprint arXiv:1602.07360.
[7] https://medium.com/@smallfishbigsea/notes-of-squeezenet-4137d51feef4
[8] Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., ... & Adam, H. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861
[9] https://medium.com/ai-academy-taiwan/efficient-cnn-%E4%BB%8B%E7%B4%B9-%E4%B8%80-mobilenetv1-304c96f5eb7e
[10] Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017). ImageNet classification with deep convolutional neural networks. Communications of the ACM, 60(6), 84-90.
[11] https://www.kaggle.com/datasets/masoudnickparvar/brain-tumor-mri-dataset
[12] https://chingtien.medium.com/%E5%BF%83%E7%90%86%E5%AD%B8%E5%92%8C%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92%E4%B8%AD%E7%9A%84-accuracy-precision-recall-rate-%E5%92%8C-confusion-matrix-529d18abc3a
[13] S. Arora and M. Sharma, "Deep Learning for Brain Tumor Classification from MRI Images," 2021 Sixth International Conference on Image Information Processing (ICIIP), Shimla, India, 2021, pp. 409-412, doi: 10.1109/ICIIP53038.2021.9702609.
[14] Ioffe, S., & Szegedy, C. (2015, June). Batch normalization: Accelerating deep network training by reducing internal covariate shift. In International conference on machine learning (pp. 448-456). pmlr.
[15] https://medium.com/@hupinwei/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E4%BB%80%E9%BA%BC%E6%98%AFglobal-average-pooling-a48bffb8e0e0
[16] H. Hsu, C. Lin, C. Lu, J. Wang and T. Huang, "A Lightweight CNN Net for AMD Detection Using OCT Volumes," 2022 IEEE International Conference on Consumer Electronics (ICCE), Las Vegas, NV, USA, 2022, pp. 01-04, doi: 10.1109/ICCE53296.2022.9730562.
[17] A. Mikołajczyk and M. Grochowski, "Data augmentation for improving deep learning in image classification problem," 2018 International Interdisciplinary PhD Workshop (IIPhDW), Świnouście, Poland, 2018, pp. 117-122, doi: 10.1109/IIPHDW.2018.8388338.
[18] 陳冠郡(2022)。基於軟硬體共同設計之低面積可重新配置卷積神經網路加速器。﹝碩士論文。國立臺北科技大學﹞臺灣博碩士論文知識加值系統。 https://hdl.handle.net/11296/48xgjj。
[19] Y. -H. Chen, T. Krishna, J. Emer and V. Sze, "14.5 Eyeriss: An energy-efficient reconfigurable accelerator for deep convolutional neural networks," 2016 IEEE International Solid-State Circuits Conference (ISSCC), San Francisco, CA, USA, 2016, pp. 262-263, doi: 10.1109/ISSCC.2016.7418007.
[20] Y. -J. Lin and T. S. Chang, "Data and Hardware Efficient Design for Convolutional Neural Network," in IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 65, no. 5, pp. 1642-1651, May 2018, doi: 10.1109/TCSI.2017.2759803.
[21] Zhang, C., Li, P., Sun, G., Guan, Y., Xiao, B., & Cong, J. (2015, February). Optimizing FPGA-based accelerator design for deep convolutional neural networks. In Proceedings of the 2015 ACM/SIGDA international symposium on field-programmable gate arrays (pp. 161-170).
[22] S. Mujawar, D. Kiran and H. Ramasangu, "An Efficient CNN Architecture for Image Classification on FPGA Accelerator," 2018 Second International Conference on Advances in Electronics, Computers and Communications (ICAECC), Bangalore, India, 2018, pp. 1-4, doi: 10.1109/ICAECC.2018.8479517.
[23] https://keras.io/zh/preprocessing/image
[24] Abbood, A. A., Shallal, Q. M., Fadhel, M. A., & Shallal, Q. M. (2021). Automated brain tumor classification using various deep learning models: a comparative study. Indonesian Journal of Electrical Engineering and Computer Science, 22(1), 252-259.
[25] 陳泓銘(2020)。基於CIFAR-10資料集之CNN模型推斷硬體加速器。﹝碩士論文。國立臺灣科技大學﹞臺灣博碩士論文知識加值系統。 https://hdl.handle.net/11296/4m4kbx。