應用三維卷積神經網路於急性缺血型中風電腦斷層影像之自動化分割

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：139

、訪客IP：18.226.159.73

姓名

江以諾(YI-NUO JIANG) 查詢紙本館藏

畢業系所

電機工程學系

論文名稱

應用三維卷積神經網路於急性缺血型中風電腦斷層影像之自動化分割
(Application of Three-Dimensional Convolutional Neural Network for Automated Segmentation of Acute Ischemic Stroke Computed Tomography)

相關論文

★ 電子式基因序列偵測晶片之原型	★ 眼動符號表達系統之可行性研究
★ 利用網印碳電極以交流阻抗法檢測糖化血紅素	★ 電子式基因序列偵測晶片可行性之研究
★ 電腦化肺音擷取系統	★ 眼寫鍵盤和眼寫滑鼠
★ 眼寫電話控制系統	★ 氣喘肺音監測系統之可行性研究
★ 肺音聽診系統之可行性研究	★ 穿戴式腳趾彎曲角度感測裝置之可行性研究
★ 注音符號眼寫系統之可行性研究	★ 英文字母眼寫系統之可行性研究
★ 數位聽診器之原型	★ 使用角度變化率為基準之心電訊號壓縮法
★ 電子式基因微陣列晶片與應用電路研究	★ 電子聽診系統應用於左右肺部比較之臨床研究

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

至系統瀏覽論文 (2027-7-31以後開放)

摘要(中)

急性缺血型腦中風(Acute Ischemic Stroke, AIS)是由大腦動脈阻塞而引起的腦部症狀。若能在黃金時間內恢復血流，未完全成為梗塞區(Infarct)的缺血區(Ischemia)還是有機會恢復。目前的臨床治療方法包括靜脈內血栓溶解治療(Intravenous Thrombolysis, IVT)或動脈內取栓手術(Endovascular Thrombectomy, EVT)。而電腦斷層影像(Computed Tomography, CT)是判斷急性缺血型中風病灶(Lesion)的重要依據。在病灶的標註上，主要依賴醫師的判讀，但是人工標註的方式耗時耗力，且存在主觀的不一致性，尤其在CT上，取栓治療前的缺血區更是肉眼較難判讀的。此研究的目標是建立自動分割急性缺血型腦中風缺血區的模型，客觀地協助臨床醫師重複確認該腦部區域是否有沒注意到的病灶。我們使用了70筆從聯新國際醫院(Landseed International Hospital)收集的醫學影像，經由影像前處理後，透過採用雙通道的三維卷積神經網路(Convolutional Neural Network, CNN)架構的DeepMedic進行腦遮罩分割模型及缺血區分割模型的訓練。缺血區分割模型的結果與黃金標準相比，其Dice達到73.8%。此研究結果在臨床上可以搭配磁振造影 (Magnetic Resonance Imaging, MRI)一同判斷，以提供更全面及詳細的診斷。

摘要(英)

Acute Ischemic Stroke (AIS) is a brain symptom caused by the obstruction of cerebral arteries. If blood flow is restored within the golden hours, ischemic areas that have not fully become infarcted still have the potential to recover. Current clinical treatments include Intravenous Thrombolysis (IVT) and Endovascular Thrombectomy (EVT). Computed Tomography (CT) is a crucial tool for identifying the lesions in acute ischemic stroke patients. Lesion delineation primarily relies on clinicians′ judgment, but manual delineation is time-consuming and labor-intensive, with subjective inconsistencies, especially in identifying ischemic areas on pre-thrombectomy CT scans. This study aims to develop a model for automatic segmentation of ischemic areas in acute ischemic stroke, aiding clinicians in objectively confirming overlooked lesions. We used 70 medical images collected from Landseed International Hospital. After image preprocessing, we trained brain mask segmentation model and ischemic segmentation model using DeepMedic, a dual-channel 3D Convolutional Neural Network (CNN) architecture. The ischemic area segmentation model achieved a Dice coefficient of 73.8% compared to the gold standard. These results can be used clinically alongside Magnetic Resonance Imaging (MRI) to provide more comprehensive and detailed diagnoses.

關鍵字(中)

★ 缺血型中風
★ 電腦斷層影像
★ 深度學習
★ 自動分割
★ 腦遮罩提取

關鍵字(英)

★ Ischemic Stroke
★ Computed Tomography
★ Deep Learning
★ Automatic Segmentation
★ Brain Mask Extraction

論文目次

摘要 iii
Abstract iv
致謝 v
目錄 vi
圖目錄 ix
表目錄 xii
第一章研究動機與文獻回顧 1
1.1 相關知識 1
1.2 研究動機 2
1.3 文獻回顧 3
第二章研究方法 7
2.1 資料來源 7
2.2 研究架構 11
2.3 影像前處理 12
2.3.1 轉換檔案格式 12
2.3.2 窗值化及標準化 12
2.3.3 影像對位 14
2.3.4 提取腦遮罩 16
2.3.5 病灶標註 17
2.3.6 影像體素尺寸重採樣 21
2.3.7 資料增量 22
2.4 深度學習模型 23
2.5 交叉驗證 25
2.6 評估指標 26
第三章研究結果 28
3.1 腦遮罩分割結果 28
3.2 缺血區分割結果 31
第四章討論 37
4.1 研究侷限 37
4.2 腦遮罩分割使用不同模型之差異 38
4.3 缺血區分割模型效能探討 46
4.4 案例探討 48
4.4.1 缺血區病灶大小範圍不同之探討 48
4.4.2 缺血區病灶明顯程度不同之探討 48
4.4.3 老舊梗塞區之探討 49
4.4.4 白質病變之探討 52
4.5 現有文獻比較 53
4.6 研究價值與臨床應用分析 56
第五章結論及未來展望 58
5.1 結論 58
5.2 未來展望 58
參考文獻 59

參考文獻

[1] D. Barthels and H. Das, “Current advances in ischemic stroke research and therapies,” Biochimica et Biophysica Acta (BBA) - Molecular Basis of Disease, vol. 1866, no. 4, p. 165260, Apr. 2020, doi: 10.1016/j.bbadis.2018.09.012.
[2] 統計處, “歷年統計,” 統計處. Accessed: Jun. 11, 2024. [Online]. Available: https://dep.mohw.gov.tw/DOS/lp-5069-113.html
[3] “Classification of Stroke Subtypes | Cerebrovascular Diseases | Karger Publishers.” Accessed: Jun. 13, 2024. [Online]. Available: https://karger.com/ced/article/27/5/493/69218/Classification-of-Stroke-Subtypes
[4] L. Wang, X. Xiong, L. Zhang, and J. Shen, “Neurovascular Unit: A critical role in ischemic stroke,” CNS Neuroscience & Therapeutics, vol. 27, no. 1, pp. 7–16, 2021, doi: 10.1111/cns.13561.
[5] H. Kato and K. Kogure, “Biochemical and Molecular Characteristics of the Brain with Developing Cerebral Infarction,” Cell Mol Neurobiol, vol. 19, no. 1, pp. 93–108, Feb. 1999, doi: 10.1023/A:1006920725663.
[6] “Mechanisms of neuronal cell death in ischemic stroke and their therapeutic implications”, doi: 10.1002/med.21817.
[7] F. Gariel et al., “Mechanical Thrombectomy Outcomes With or Without Intravenous Thrombolysis,” Stroke, vol. 49, no. 10, pp. 2383–2390, Oct. 2018, doi: 10.1161/STROKEAHA.118.021500.
[8] M. Goyal et al., “Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials,” The Lancet, vol. 387, no. 10029, pp. 1723–1731, Apr. 2016, doi: 10.1016/S0140-6736(16)00163-X.
[9] M. U. Jang et al., “In-Hospital and Post-Discharge Recovery after Acute Ischemic Stroke: a Nationwide Multicenter Stroke Registry-base Study,” Journal of Korean Medical Science, vol. 34, no. 36, Aug. 2019, doi: 10.3346/jkms.2019.34.e240.
[10] “CT for Treatment Selection in Acute Ischemic Stroke: A Code Stroke Primer | RadioGraphics.” Accessed: Jun. 13, 2024. [Online]. Available: https://pubs.rsna.org/doi/10.1148/rg.2019190142
[11] Albers Gregory W. et al., “Thrombectomy for Stroke at 6 to 16 Hours with Selection by Perfusion Imaging,” New England Journal of Medicine, vol. 378, no. 8, pp. 708–718, Feb. 2018, doi: 10.1056/NEJMoa1713973.
[12] S.-J. Peng, Y.-W. Chen, J.-Y. Yang, K.-W. Wang, and J.-Z. Tsai, “Automated Cerebral Infarct Detection on Computed Tomography Images Based on Deep Learning,” Biomedicines, vol. 10, no. 1, p. 122, Jan. 2022, doi: 10.3390/biomedicines10010122.
[13] M. Brett, I. S. Johnsrude, and A. M. Owen, “The problem of functional localization in the human brain,” Nat Rev Neurosci, vol. 3, no. 3, pp. 243–249, Mar. 2002, doi: 10.1038/nrn756.
[14] O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” May 18, 2015, arXiv: arXiv:1505.04597. doi: 10.48550/arXiv.1505.04597.
[15] W.-C. Wang et al., “Automated delineation of acute ischemic stroke lesions on non-contrast CT using 3D deep learning: A promising step towards efficient diagnosis and treatment,” Biomedical Signal Processing and Control, vol. 93, p. 106139, Jul. 2024, doi: 10.1016/j.bspc.2024.106139.
[16] G. N. Hounsfield, “Computed medical imaging,” Science, vol. 210, no. 4465, pp. 22–28, Oct. 1980, doi: 10.1126/science.6997993.
[17] Z. Xue, S. Antani, L. R. Long, D. Demner-Fushman, and G. R. Thoma, “Window Classification of Brain CT Images in Biomedical Articles,” AMIA Annu Symp Proc, vol. 2012, pp. 1023–1029, Nov. 2012.
[18] P. J. Turner and G. Holdsworth, “CT stroke window settings: an unfortunate misleading misnomer?,” Br J Radiol, vol. 84, no. 1008, pp. 1061–1066, Dec. 2011, doi: 10.1259/bjr/99730184.
[19] “SPM12 Software - Statistical Parametric Mapping.” Accessed: Jun. 13, 2024. [Online]. Available: https://www.fil.ion.ucl.ac.uk/spm/software/spm12/
[20] T. P. Naidich, D. L. Daniels, P. Pech, V. M. Haughton, A. Williams, and K. Pojunas, “Anterior commissure: anatomic-MR correlation and use as a landmark in three orthogonal planes.,” Radiology, vol. 158, no. 2, pp. 421–429, Feb. 1986, doi: 10.1148/radiology.158.2.3941867.
[21] S.-J. Peng, “以磁振造影探究有病灶及無病灶神經疾病的自動偵測方法之開發,” 中央大學電機工程學系學位論文, vol. 2014, pp. 1–137, Jan. 2014.
[22] “Body MRI artifacts in clinical practice: A physicist’s and radiologist’s perspective - Graves - 2013 - Journal of Magnetic Resonance Imaging - Wiley Online Library.” Accessed: Jun. 13, 2024. [Online]. Available: https://onlinelibrary.wiley.com/doi/10.1002/jmri.24288
[23] “ITK-SNAP Home.” Accessed: Jun. 13, 2024. [Online]. Available: http://www.itksnap.org/pmwiki/pmwiki.php
[24] K. He, G. Gkioxari, P. Dollár, and R. Girshick, “Mask R-CNN,” in 2017 IEEE International Conference on Computer Vision (ICCV), Oct. 2017, pp. 2980–2988. doi: 10.1109/ICCV.2017.322.
[25] K. Kamnitsas et al., “Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation,” Medical Image Analysis, vol. 36, pp. 61–78, Feb. 2017, doi: 10.1016/j.media.2016.10.004.
[26] S. Li, J. Zheng, and D. Li, “Precise segmentation of non-enhanced computed tomography in patients with ischemic stroke based on multi-scale U-Net deep network model,” Computer Methods and Programs in Biomedicine, vol. 208, p. 106278, Sep. 2021, doi: 10.1016/j.cmpb.2021.106278.
[27] H. Kuang, B. K. Menon, and W. Qiu, “Semi-automated infarct segmentation from follow-up noncontrast CT scans in patients with acute ischemic stroke,” Medical Physics, vol. 46, no. 9, pp. 4037–4045, 2019, doi: 10.1002/mp.13703.
[28] M. Soltanpour, R. Greiner, P. Boulanger, and B. Buck, “Improvement of automatic ischemic stroke lesion segmentation in CT perfusion maps using a learned deep neural network,” Computers in Biology and Medicine, vol. 137, p. 104849, Oct. 2021, doi: 10.1016/j.compbiomed.2021.104849.
[29] H. Kuang, B. K. Menon, and W. Qiu, “Segmenting Hemorrhagic and Ischemic Infarct Simultaneously From Follow-Up Non-Contrast CT Images in Patients With Acute Ischemic Stroke,” IEEE Access, vol. 7, pp. 39842–39851, 2019, doi: 10.1109/ACCESS.2019.2906605.
[30] Z. Xu and C. Ding, “Combining convolutional attention mechanism and residual deformable Transformer for infarct segmentation from CT scans of acute ischemic stroke patients,” Front Neurol, vol. 14, p. 1178637, 2023, doi: 10.3389/fneur.2023.1178637.
[31] R. Sales Barros et al., “Automatic segmentation of cerebral infarcts in follow-up computed tomography images with convolutional neural networks,” J Neurointerv Surg, vol. 12, no. 9, pp. 848–852, Sep. 2020, doi: 10.1136/neurintsurg-2019-015471.
[32] K. Liang et al., “Symmetry-Enhanced Attention Network for Acute Ischemic Infarct Segmentation with Non-Contrast CT Images,” vol. 12907, 2021, pp. 432–441. doi: 10.1007/978-3-030-87234-2_41.
[33] H. El-Hariri et al., “Evaluating nnU-Net for early ischemic change segmentation on non-contrast computed tomography in patients with Acute Ischemic Stroke,” Comput Biol Med, vol. 141, p. 105033, Feb. 2022, doi: 10.1016/j.compbiomed.2021.105033.
[34] H. Ni et al., “Asymmetry Disentanglement Network for Interpretable Acute Ischemic Stroke Infarct Segmentation in Non-Contrast CT Scans,” Jun. 30, 2022, arXiv: arXiv:2206.15445. doi: 10.48550/arXiv.2206.15445.
[35] H. Nishi et al., “Automatic Ischemic Core Estimation Based on Noncontrast-Enhanced Computed Tomography,” Stroke, vol. 54, no. 7, pp. 1815–1822, Jul. 2023, doi: 10.1161/STROKEAHA.123.042689.
[36] H. Kuang, B. K. Menon, S. I. Sohn, and W. Qiu, “EIS-Net: Segmenting early infarct and scoring ASPECTS simultaneously on non-contrast CT of patients with acute ischemic stroke,” Med Image Anal, vol. 70, p. 101984, May 2021, doi: 10.1016/j.media.2021.101984.
[37] J. Zhang, T. Wan, E. MacDonald, B. Menon, A. Ganesh, and Q. Wu, “Synchronous Image-Label Diffusion Probability Model with Application to Stroke Lesion Segmentation on Non-contrast CT,” Jul. 18, 2023, arXiv: arXiv:2307.01740. doi: 10.48550/arXiv.2307.01740.

指導教授

蔡章仁(JANG-ZERN TSAI)

審核日期

2024-7-30

推文