Prostate MR/TRUS image segmentation and registration methods based on supervised learning
-
摘要: 前列腺核磁超聲圖像配準融合有助于實現前列腺腫瘤的靶向穿刺。傳統的配準方法主要是針對手動分割的前列腺核磁（Magnetic resonance, MR）和經直腸超聲（Trans-rectal ultrasound, TRUS）圖像上對應的生理特征點作為參考點，進行剛體或非剛體配準。針對超聲圖像因成像質量低導致手動分割配準效率低下的問題，提出一種基于監督學習的前列腺MR/TRUS圖像自動分割方法，與術前核磁圖像進行非剛體配準。首先，針對圖像分割任務訓練前列腺超聲圖像的活動表觀模型（Active appearance model, AAM），并基于隨機森林建立邊界驅動的數學模型，實現超聲圖像自動分割。接著，提取術前分割的核磁圖像與自動分割的超聲圖像建立輪廓的形狀特征矢量，進行特征匹配與圖像配準。實驗結果表明，本文方法能準確實現前列腺超聲圖像自動分割與配準融合，9組配準結果的戴斯相似性系數（Dice similarity coefficient, DSC）均大于0.98，同時尿道口處特征點的平均定位精度達1.64 mm，相比傳統方法具有更高的配準精度。Abstract: At present, the diagnosis of prostate cancer mainly relies on the level of prostate-specific antigen (PSA) followed by a prostate biopsy. The technology, transrectal ultrasound (TRUS), has been the most popular method for diagnosing prostate cancer because of its advantages, such as real-time, low cost, easy operation. However, the low imaging quality of ultrasound equipment makes it difficult to distinguish regions of malignant tumors from those of healthy tissues from low-quality images, which results in missing diagnoses or overtreating conditions. In contrast, magnetic resonance (MR) images of the prostate can quickly locate the position of malignant tumors. It is crucial to register the annotated MR images and the corresponding TRUS image to perform a targeted biopsy of the prostate tumor. The registration fusion of prostate magnetic resonance and transrectal ultrasound images helps to improve the accuracy of the prostate lesions targeted biopsy. Traditional registration methods that are usually manually selected, specific anatomical landmarks in segmented areas used as a reference, and performed rigid or nonrigid registration, which is inefficient because of the low quality of prostate TRUS images and the substantial differences in pixel intensity of the prostate between MR and TRUS images. This paper proposed a novel prostate MR/TRUS image segmentation and the automatic registration method was based on a supervised learning framework. First, the prostate active appearance model was trained to be applied in the prostate TRUS images segmentation task, and the random forest classifier was used for building a boundary-driven mathematical model to realize automatic segmentation of TRUS images. Then, some sets of MR/TRUS images contour landmarks were computed by matching the corresponding shape descriptors used for registration. The method was validated by comparing the automatic contour segmentation results with standard results, and the registration results with a traditional registration method. Results showed that our method could accurately realize the automatic segmentation and registration of prostate TRUS and MR images. The DSC (Dice similarity coefficient, DSC) accuracy of nine sets of registration results is higher than 0.98, whereas the average location accuracy of the urethral opening is 1.64 mm, which displays a better registration performance.
-
Key words:
- prostate /
- image registration /
- image segmentation /
- random forest /
- active appearance model
-
圖 1 初始姿態
${{{T}}_{{\rm{ini}}}}$ 對收斂結果的影響。（a～d）初始姿態參數過大無法收斂；（e～f）初始姿態滿足收斂條件Figure 1. Effect of the initial position parameter
${{{T}}_{{\rm{ini}}}}$ on the results of convergence: (a?d) large initial position parameter resulted error convergence; (e?f) initial position parameter met convergence conditions
圖 2 前列腺隨機森林模型的訓練和預測過程
Figure 2. Training and prediction of the prostate random forest model
圖 4 姿態
${{{T}}_k}$ 下的坐標變化關系Figure 4. Coordinate transformation relationship at position
${{{T}}_k}$ 
圖 6 前列腺超聲圖像的自動分割過程。（a）前列腺TRUS圖像；（b）參數尋優結果；（c）圖像分割結果；（d）分割對比結果
Figure 6. Automatic segmentation process of prostate TRUS images: (a) prostate TRUS image; (b) parameters optimization result; (c) segmentation result; (d) image segmentation comparison results
圖 7 待配準的前列腺MR/TRUS圖像。（a）MR圖像輪廓點；（b）TRUS圖像輪廓點
Figure 7. Prostate MR/TRUS images to be registered: (a) contour points on MR image; (b) contour points on TRUS image
圖 10 圖像配準融合結果。（a）MR圖像的變換結果；（b）MR/TRUS圖像融合結果
Figure 10. Registration results: (a) transformation result of MR image; (b) registration result of MR/TRUS images
圖 11 前列腺超聲圖像分割結果。（a1～a5）待分割的TRUS圖像；（b1～b5）隨機森林預分割結果；（c1～c5）輪廓分割收斂過程
Figure 11. Segmentation results of prostate US images: (a1?a5) initial TRUS images; (b1?b5) pre-segmentation results of random forest; (c1?c5) convergence processes of contour segmentation
表 1 尿道特征點定位結果對比
Table 1. Comparison of location results of urethral points
Sample Method proposed Literature method DSC TE/mm DSC TE/mm 1 0.9938 1.76 0.9939 1.93 2 0.9873 1.24 0.9816 3.84 3 0.9897 1.92 0.9893 1.27 4 0.9906 1.58 0.9872 2.55 5 0.9871 1.47 0.9884 1.42 AP 0.9897 1.59 0.9880 2.20 ${d_2}$ 0.0024 0.23 0.0039 0.93 
下載: 導出CSV
259luxu-164<th id="5nh9l"></th> <strike id="5nh9l"></strike> <th id="5nh9l"><noframes id="5nh9l"><th id="5nh9l"></th> <strike id="5nh9l"></strike> <th id="5nh9l"><noframes id="5nh9l"> <th id="5nh9l"></th> <strike id="5nh9l"><noframes id="5nh9l"><span id="5nh9l"></span> <span id="5nh9l"><noframes id="5nh9l"><span id="5nh9l"></span> <strike id="5nh9l"><noframes id="5nh9l"><strike id="5nh9l"></strike> <span id="5nh9l"><noframes id="5nh9l"> <span id="5nh9l"><noframes id="5nh9l"> <span id="5nh9l"></span> <span id="5nh9l"><video id="5nh9l"></video></span> <th id="5nh9l"><noframes id="5nh9l"><th id="5nh9l"></th> -
參考文獻
[1] Deng Y S, He Y H, Zhou X F. Development of prostate targeted puncture technology. J Mimimally Invasive Urology, 2018, 7(6): 428鄧益森, 何宇輝, 周曉峰. 前列腺靶向穿刺技術發展概況. 微創泌尿外科雜志, 2018, 7(6):428 [2] Guichard G, Larré, Gallina A, et al. Extended 21-sample needle biopsy protocol for diagnosis of prostate cancer in 1000 consecutive patients. Eur Urol, 2007, 52(2): 430 doi: 10.1016/j.eururo.2007.02.062 [3] Zhou Z E, Yan W G, Zhou Y, et al. Recent progress in MRI-ultrasound fusion for guidance of targeted prostate biopsy. Chin J Surg, 2016, 54(10): 792 doi: 10.3760/cma.j.issn.0529-5815.2016.10.016周智恩, 嚴維剛, 周毅, 等. MRI-超聲融合引導下前列腺靶向穿刺活檢的最新進展. 中華外科雜志, 2016, 54(10):792 doi: 10.3760/cma.j.issn.0529-5815.2016.10.016 [4] Qu H W, Liu H, Cui Z L, et al. Focusing on MRI-suspected lesions in targeted transrectal prostate biopsy guided by MRI-TRUS fusion imaging for the diagnosis of prostate cancer. Natl J Androl, 2016, 22(9): 782曲華偉, 劉輝, 崔子連, 等. 重點穿刺MRI可疑病灶區域在MRI/TRUS融合成像引導靶向前列腺穿刺中的診斷價值. 中華男科學雜志, 2016, 22(9):782 [5] Schlenker B, Apfelbeck M, Buchner A, et al. MRI-TRUS fusion biopsy of the prostate: quality of image fusion in a clinical setting. Clin Hemorheol Microcirculat, 2018, 70(4): 433 [6] Mitra J, Martí R, Oliver A, et al. Prostate multimodality image registration based on B-splines and quadrature local energy. Int J Comput Assisted Radiol Surg, 2012, 7(3): 445 doi: 10.1007/s11548-011-0635-8 [7] Sun Y, Yuan J, Qiu W, et al. Three-dimensional nonrigid MR-TRUS registration using dual optimization. IEEE Trans Med Imag, 2015, 34(5): 1085 doi: 10.1109/TMI.2014.2375207 [8] Moradi M, Janoos F, Fedorov A, et al. Two solutions for registration of ultrasound to MRI for image-guided prostate interventions // 2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society. San Diego, 2012: 1129 [9] Fedorov A, Khallaghi S, Sánchez C A, et al. Open-source image registration for MRI-TRUS fusion-guided prostate interventions. Int J Comput Assisted Radiol Surg, 2015, 10(6): 925 doi: 10.1007/s11548-015-1180-7 [10] Ni D, Wu H L. MRI-TRUS multi-modality image fusion for targeted prostate biopsy. J Shenzhen Univ Sci Eng, 2016, 33(2): 111 doi: 10.3724/SP.J.1249.2016.02111倪東, 吳海浪. 基于核磁-超聲融合的前列腺靶向穿刺系統. 深圳大學學報: 理工版, 2016, 33(2):111 doi: 10.3724/SP.J.1249.2016.02111 [11] Wang W R. Research on Prostate Puncture Assisted by MR and TRUS Image[Dissertation]. Harbin: Harbin Institute of Technology, 2018王煒榮. MR與TRUS圖像輔助前列腺穿刺技術研究[學位論文]. 哈爾濱: 哈爾濱工業大學, 2018 [12] Du C. MR and TRUS Image Denoising and Segmentation Methods in Prostate Puncture Guidance[Dissertation]. Harbin: Harbin Institute of Technology, 2019杜超. 前列腺穿刺引導中的MR和TRUS圖像去噪與分割方法[學位論文]. 哈爾濱: 哈爾濱工業大學, 2019 [13] Cootes T F, Edwards G J, Taylor C J. Active appearance models. IEEE Trans Pattern Anal Mach Intellig, 2001, 23(6): 681 doi: 10.1109/34.927467 [14] Bookstein F L. Principal warps: thin-plate splines and the decomposition of deformations. IEEE Trans Pattern Anal Mach Intellig, 1989, 11(6): 567 doi: 10.1109/34.24792 [15] Svetnik V, Liaw A, Tong C, et al. Random forest: a classification and regression tool for compound classification and QSAR modeling. J Chem Inf Comput Sci, 2003, 43(6): 1947 doi: 10.1021/ci034160g [16] Rohr K, Fornefett M, Stiehl H S. Spline-based elastic image registration: integration of landmark errors and orientation attributes. Comput Vision Image Understand, 2003, 90(2): 153 doi: 10.1016/S1077-3142(03)00048-1 [17] Evangelidis G D, Psarakis E Z. Parametric image alignment using enhanced correlation coefficient maximization. IEEE Trans Pattern Anal Mach Intellig, 2008, 30(10): 1858 doi: 10.1109/TPAMI.2008.113 [18] Ghosh P, Mitchell M, Tanyi J A, et al. Incorporating priors for medical image segmentation using a genetic algorithm. Neurocomputing, 2016, 195: 181 doi: 10.1016/j.neucom.2015.09.123 [19] Cosío F A. Automatic initialization of an active shape model of the prostate. Med Image Anal, 2008, 12(4): 469 doi: 10.1016/j.media.2008.02.001 [20] Belongie S, Malik J, Puzicha J. Shape matching and object recognition using shape contexts. IEEE Trans Pattern Anal Mach Intellig, 2002, 24(4): 509 doi: 10.1109/34.993558 [21] Munkres J. Algorithms for the assignment and transportation problems. J Soc Ind Appl Math, 1957, 5(1): 32 doi: 10.1137/0105003 [22] ?erifovi?-Trbali? A, Demirovi? D, Prlja?a N, et al. Intensity-based elastic registration incorporating anisotropic landmark errors and rotational information. Int J Comput Assisted Radiol Surg, 2009, 4(5): 463 doi: 10.1007/s11548-009-0358-2 [23] Keys R. Cubic convolution interpolation for digital image processing. IEEE Trans Acoust Speech Signal Process, 1981, 29(6): 1153 doi: 10.1109/TASSP.1981.1163711 [24] Huttenlocher D P, Klanderman G A, Rucklidge W J. Comparing images using the Hausdorff distance. IEEE Trans Pattern Anal Mach Intellig, 1993, 15(9): 850 doi: 10.1109/34.232073 [25] Dice L R. Measures of the amount of ecologic association between species. J Ecol, 1945, 26(3): 297 doi: 10.2307/1932409 [26] Mitra J, Marti R, Oliver A, et al. A comparison of thin-plate splines with automatic correspondences and B-splines with uniform grids for multimodal prostate registration. Proc SPIE - Int Soc Opt Eng, 2011, 7964(2): 150 -
點擊查看大圖
計量
- 文章訪問數: 3984
- HTML全文瀏覽量: 1068
- PDF下載量: 87
- 被引次數: 0
