磁振造影（Magnetic Resonance Imaging,MRI）越來越多地用於診斷腦組織病變，特別是擴散加權(Diffusion Weighted Image,DWI)在檢測缺血性腦中風中具有高度敏感性。然而，MRI會因設備，非自主運動，磁化率及金屬等的產生多種偽影，而這些偽影在DWI影像中的直方圖與梗塞有很大的重疊區域，容易使醫生判斷病人時造成誤判。對於日常臨床使用而言，醫生仍然需要手動或半手動地對腦區域中的病變進行評估,手動評估病理變化太麻煩且耗費時間，也難免受到個人主觀性的影響，然而目前為止提出的數種全自動分割方法中去除偽影與正確率仍保有很大的進步空間。在此項研究中，我們提出了一種基於深度學習的方法，首先取出影像的腦組織，經由旋轉T1加權影像(T1-Weighted Image,T1WI)找出最對稱的旋轉角度，然而將T1 map移位至影像正中心，再將DWI與表觀擴散係(Apparent Diffusion Coefficient,ADC) map對位至T1 map。依每張切片的影像強度，設立自適性門檻值，濾除大部分頭骨與雜訊。判斷小腦的影像平均強度是否與大腦平均強度是否相似，若不相同則會調整小腦平均強度。由於每個人大腦影像強度存在著變異性，所以我們將利用DWI map做線性回歸的方法，提取出梗塞機率較高的區域，並將非線性回歸所得的影像，作為第三種參考依據。將之前對位的DWI與ADC map和非線性回歸影像重疊，再將重疊的影像切割成數個小影像並加入切割的位置資訊、整個大腦的影像平均強度、第幾張切片，最後利用深度學習排除偽影與非梗塞並偵測出梗塞的位置，供醫生參考。該研究對於偽影的辨識，與醫生手動評估的速度上，達到了很好的準確率與速度。

Magnetic resonance imaging (MRI) is highly sensitive to stroke lesions. However,
MRI can produce a variety of artifacts due to equipment, involuntary movement, magnetic susceptibility, and metal. The histograms of these artifact may overlap with those of infact in the DWI image, which causes infarct segmentation errors. In the traditional clinical infarct segmentation, the doctors need to use manual or semi-automated methods to detect infarct lesions in the brain area. These nonautomated methods to assess pathological changes is cumbersome, time-consuming, and easily influenced by the assessor’s personal subjectivity. There is still a lot of room for improvement in removing artifacts and accuracy in the automatic segmentation method. In this study, we propose a cerebral infarct segmentation method based on deep learning. First, the brain tissue of the image is extracted, and the T1-Weighted (T1-W) image is rotated at various angles to find the angle with respect to which the T1-W image is most symmetrical. Then the T1-W image is centralized by a shift, and the DWI and ADC map are registered to T1-W image. Based on the image intensity of each slice, an adaptive threshold is set to filter out most of the skull and noise. The average image intensity of the cerebellum is compared with that of the cerebrum. If the cerebellar intensity is higher than the cerebral intensity, the average intensity of the cerebellum is adjusted to the same level as the cerebral intensity. To accommodate the inter-person variability in the brain image intensity, a new image is generated as the third reference image in addition to the DWI and ADC map. This new image contains regions with a high probability of infarction extracted from the DWI by nonlinear regression. The previously aligned DWI, ADC image, and the non-linear regression image are all divided into patches of 16 × 16 pixels. Each patch is accompanied with the position information of the patch, the average intensity of the entire brain image, and the slice number. A convolutional neural network (CNN) is constructed and trained with the patches of over 30 patients. The trained CNN achieves a good performance in identifying infarct, noninfarct, artifact patches at a high speed.

[1] (2019). 衛生福利部,107年死因統計年報電子書,衛生福利部. Available: https://dep.mohw.gov.tw/DOS/lp-4472-113.html
[2] J.-Z. Tsai, S.-J. Peng, Y.-W. Chen, K.-W. Wang, H.-K. Wu, Y.-Y. Lin, Y.-Y. Lee, C.-J. Chen, H.-J. Lin, and E. E. J. B. r. i. Smith, "Automatic detection and quantification of acute cerebral infarct by fuzzy clustering and histographic characterization on diffusion weighted MR imaging and apparent diffusion coefficient map," BioMed Research International, vol. 2014, 2014.
[3] J. Canny, "A computational approach to edge detection," IEEE Transactions on Pattern Analysis and Machine Intelligence, no. 6, pp. 679-698, 1986.
[4] J. C. Bezdek, Pattern recognition with fuzzy objective function algorithms. Springer Science & Business Media, 2013.
[5] P. Anbeek, K. L. Vincken, G. S. Van Bochove, M. J. Van Osch, and J. J. N. van der Grond, "Probabilistic segmentation of brain tissue in MR imaging," NeuroImage, vol. 27, no. 4, pp. 795-804, 2005.
[6] P. Anbeek, K. L. Vincken, M. J. van Osch, R. H. Bisschops, and J. J. M. i. a. van der Grond, "Automatic segmentation of different-sized white matter lesions by voxel probability estimation," Medical Image Analysis, vol. 8, no. 3, pp. 205-215, 2004.
[7] J.-Z. Tsai, S.-J. Peng, Y.-W. Chen, K.-W. Wang, C.-H. Li, J.-Y. Wang, C.-J. Chen, H.-J. Lin, E. E. Smith, and H.-K. J. P. o. Wu, "Automated segmentation and quantification of white matter hyperintensities in acute ischemic stroke patients with cerebral infarction," PloS one, vol. 9, no. 8, p. e104011, 2014.
[8] N. P. De La Ossa, M. Hernandez-Perez, S. Domenech, P. Cuadras, A. Massuet, M. Millan, M. Gomis, E. López-Cancio, L. Dorado, and A. J. C. D. Dávalos, "Hyperintensity of distal vessels on FLAIR is associated with slow progression of the infarction in acute ischemic stroke," Cerebrovascular Diseases, vol. 34, no. 5-6, pp. 376-384, 2012.
[9] L. Chen, P. Bentley, and D. J. N. C. Rueckert, "Fully automatic acute ischemic lesion segmentation in DWI using convolutional neural networks," NeuroImage, vol. 15, pp. 633-643, 2017.
[10] M. Ghafoorian, N. Karssemeijer, T. Heskes, I. W. van Uden, C. I. Sanchez, G. Litjens, F.-E. de Leeuw, B. van Ginneken, E. Marchiori, and B. J. S. R. Platel, "Location sensitive deep convolutional neural networks for segmentation of white matter hyperintensities," Scientific Reports, vol. 7, no. 1, p. 5110, 2017.
[11] D. R. Pereira, P. P. Reboucas Filho, G. H. de Rosa, J. P. Papa, and V. H. C. de Albuquerque, "Stroke lesion detection using convolutional neural networks," in 2018 International joint conference on neural networks (IJCNN), 2018, pp. 1-6: IEEE.
[12] K. Kamnitsas, C. Ledig, V. F. Newcombe, J. P. Simpson, A. D. Kane, D. K. Menon, D. Rueckert, and B. J. M. i. a. Glocker, "Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation," Medical image analysis, vol. 36, pp. 61-78, 2017.
[13] S. Ghosh, P. Boulanger, S. T. Acton, S. S. Blemker, and N. Ray, "Automated 3D muscle segmentation from MRI data using convolutional neural network," in 2017 IEEE International Conference on Image Processing (ICIP), 2017, pp. 4437-4441: IEEE.
[14] H. Vu, H.-C. Kim, and J.-H. Lee, "3D convolutional neural network for feature extraction and classification of fMRI volumes," in 2018 International Workshop on Pattern Recognition in Neuroimaging (PRNI), 2018, pp. 1-4: IEEE.
[15] S. Kumar, A. Negi, J. Singh, and H. Verma, "A deep learning for brain tumor mri images semantic segmentation using fcn," in 2018 4th International Conference on Computing Communication and Automation (ICCCA), 2018, pp. 1-4: IEEE.
[16] Z. Wang, Y. Sun, Q. Shen, and L. J. I. A. Cao, "Dilated 3D Convolutional Neural Networks for Brain MRI Data Classification," IEEE Access, vol. 7, pp. 134388-134398, 2019.
[17] R. Zhang, L. Zhao, W. Lou, J. M. Abrigo, V. C. T. Mok, W. C. W. Chu, D. Wang, and L. Shi, "Automatic segmentation of acute ischemic stroke from DWI using 3-D fully convolutional densenets," IEEE transactions on medical imaging, vol. 37, no. 9, pp. 2149-2160, 2018.
[18] L. Liu, S. Chen, F. Zhang, F.-X. Wu, Y. Pan, J. J. N. C. Wang, and Applications, "Deep convolutional neural network for automatically segmenting acute ischemic stroke lesion in multi-modality MRI," Neural Computing and Applications, pp. 1-14, 2019.
[19] J. Long, E. Shelhamer, and T. Darrell, "Fully convolutional networks for semantic segmentation," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 3431-3440.
[20] H. Chen, Q. Dou, L. Yu, J. Qin, and P.-A. J. N. Heng, "VoxResNet: Deep voxelwise residual networks for brain segmentation from 3D MR images," NeuroImage, vol. 170, pp. 446-455, 2018.
[21] H. Lutsep, G. Albers, A. DeCrespigny, G. Kamat, M. Marks, M. J. A. o. N. O. J. o. t. A. N. A. Moseley, and t. C. N. Society, "Clinical utility of diffusion‐weighted magnetic resonance imaging in the assessment of ischemic stroke," Annals of Neurology: Official Journal of the American Neurological Association and the Child Neurology Society, vol. 41, no. 5, pp. 574-580, 1997.
[22] M. Al-Khaled, C. Matthis, T. F. Münte, and J. J. N. Eggers, "The incidence and clinical predictors of acute infarction in patients with transient ischemic attack using MRI including DWI," Neuroradiology, vol. 55, no. 2, pp. 157-163, 2013.
[23] V. Newcombe, T. Das, and J. J. J. o. n. Cross, "Diffusion imaging in neurological disease," Journal of neurology, vol. 260, no. 1, pp. 335-342, 2013.
[24] E. Smith, M. Gurol, J. Eng, C. Engel, T. Nguyen, J. Rosand, and S. J. N. Greenberg, "White matter lesions, cognition, and recurrent hemorrhage in lobar intracerebral hemorrhage," Neurology, vol. 63, no. 9, pp. 1606-1612, 2004.
[25] M. M. Schoonheim, R. M. Vigeveno, F. C. R. Lopes, P. J. Pouwels, C. H. Polman, F. Barkhof, and J. J. J. H. b. m. Geurts, "Sex‐specific extent and severity of white matter damage in multiple sclerosis: Implications for cognitive decline," Human brain mapping, vol. 35, no. 5, pp. 2348-2358, 2014.
[26] M. M. Poels, K. Zaccai, G. C. Verwoert, M. W. Vernooij, A. Hofman, A. van der Lugt, J. C. Witteman, M. M. Breteler, F. U. Mattace-Raso, and M. A. J. S. Ikram, "Arterial stiffness and cerebral small vessel disease: the Rotterdam Scan Study," Stroke, vol. 43, no. 10, pp. 2637-2642, 2012.
[27] N. Altaf, P. S. Morgan, A. Moody, S. T. MacSweeney, J. Gladman, and D. P. J. R. Auer, "Brain white matter hyperintensities are associated with carotid intraplaque hemorrhage," Radiology, vol. 248, no. 1, pp. 202-209, 2008.
[28] 磁共振常見偽影的類型及消除辦法簡介. Available: https://kknews.cc/zh-tw/news/g36p58.html
[29] Y. Du, E. A. Allen, H. He, J. Sui, and V. D. Calhoun, "Brain functional networks extraction based on fMRI artifact removal: Single subject and group approaches," in 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2014, pp. 1026-1029: IEEE.
[30] M. Julien and Z. Yue-Min, "Model-based intensity nonuniformity correction in brain MRI," in Proceedings 7th International Conference on Signal Processing, 2004. Proceedings. ICSP'04. 2004., 2004, vol. 2, pp. 982-985: IEEE.
[31] S. T. M. Duong, M. M. Schira, S. L. Phung, A. Bouzerdoum, and H. Taylor, "Anatomy-Guided Inverse-Gradient Susceptibility Artifact Correction Method for High-Resolution FMRI," in 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018, pp. 786-790: IEEE.
[32] N. J. I. t. o. s. Otsu, man, and cybernetics, "A threshold selection method from gray-level histograms," IEEE transactions on systems, man, and cybernetics, vol. 9, no. 1, pp. 62-66, 1979.
[33] O. D. Trier, T. J. I. t. o. p. a. Taxt, and m. intelligence, "Evaluation of binarization methods for document images," IEEE transactions on pattern analysis and machine intelligence, vol. 17, no. 3, pp. 312-315, 1995.
[34] Ø. D. Trier, A. K. J. I. T. o. P. A. Jain, and M. Intelligence, "Goal-directed evaluation of binarization methods," IEEE Transactions on Pattern Analysis & Machine Intelligence, no. 12, pp. 1191-1201, 1995.
[35] T. L. Faber, E. M. J. I. T. o. P. A. Stokely, and M. Intelligence, "Orientation of 3-D structures in medical images," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 10, no. 5, pp. 626-633, 1988.
[36] K. V. Mardia, T. J. I. t. o. p. a. Hainsworth, and m. intelligence, "A spatial thresholding method for image segmentation," IEEE transactions on pattern analysis and machine intelligence, vol. 10, no. 6, pp. 919-927, 1988.
[37] L. Zou, J. Zheng, C. Miao, M. J. Mckeown, and Z. J. J. I. A. Wang, "3D CNN based automatic diagnosis of attention deficit hyperactivity disorder using functional and structural MRI," IEEE Access, vol. 5, pp. 23626-23636, 2017.
[38] R. Vinoth and C. Venkatesh, "Segmentation and Detection of Tumor in MRI images Using CNN and SVM Classification," in 2018 Conference on Emerging Devices and Smart Systems (ICEDSS), 2018, pp. 21-25: IEEE.
[39] V. Bhanumathi and R. Sangeetha, "CNN Based Training and Classification of MRI Brain Images," in 2019 5th International Conference on Advanced Computing & Communication Systems (ICACCS), 2019, pp. 129-133: IEEE.
[40] M. Haghighi, S. K. Warfield, and S. Kurugol, "Automatic renal segmentation in Dce-Mri using convolutional neural networks," in 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018), 2018, pp. 1534-1537: IEEE.
[41] S. Archa and C. S. Kumar, "Segmentation of Brain Tumor in MRI Images Using CNN with Edge Detection," in 2018 International Conference on Emerging Trends and Innovations In Engineering And Technological Research (ICETIETR), 2018, pp. 1-4: IEEE.
[42] S. M. Shelke and S. W. Mohod, "Automated Segmentation and Detection of Brain Tumor from MRI," in 2018 International Conference on Advances in Computing, Communications and Informatics (ICACCI), 2018, pp. 2120-2126: IEEE.
[43] I. Fantini, L. Rittner, C. Yasuda, and R. Lotufo, "Automatic detection of motion artifacts on MRI using Deep CNN," in 2018 International Workshop on Pattern Recognition in Neuroimaging (PRNI), 2018, pp. 1-4: IEEE.
[44] K. Thara and K. Jasmine, "Brain tumour detection in MRI images using PNN and GRNN," in 2016 International Conference on Wireless Communications, Signal Processing and Networking (WiSPNET), 2016, pp. 1504-1510: IEEE.
[45] A. G. Lakshmanan, A. Swarnambiga, S. Vasuki, and A. A. Raja, "Affine based image registration applied to MRI brain," in 2013 International Conference on Information Communication and Embedded Systems (ICICES), 2013, pp. 644-649: IEEE.
[46] A. Gholipour, N. Kehtarnavaz, R. Briggs, M. Devous, and K. J. I. t. o. m. i. Gopinath, "Brain functional localization: a survey of image registration techniques," IEEE Transactions on Medical Imaging vol. 26, no. 4, pp. 427-451, 2007.