MRI影像之人類脊椎切割與疾病診斷技術的研究

A Study of Human Spine Segmentation and Disease Diagnosis Techniques for MRI Images

房宸安(Cheng-An Fang)

指導教授 ： 吳憲珠

國立臺中科技大學/資訊與流通學院/資訊工程系碩士班/碩士(2012年)

摘要

在本論文中，主要探討醫學影像切割技術於脊椎疾病之研究，脊椎的疾病大致可以分為：椎間盤退化、椎間盤突出、脊椎壓迫性骨折以及骨刺等，而這些疾病主要是發生在中老年身上，因為隨著年齡不斷的增加，人類的脊椎會慢慢老化，脊椎的疾病將是成為大眾的問題。椎間盤疾病都可以透過一些醫療儀器來掃描，而在醫學影像裡，大致可分為兩個大項，分別是CT以及MRI影像。一般脊椎在醫學影像上主要是以MRI的T2影像來做疾病的判斷，而這些判斷的工作皆需要專業醫生來辨識，因此不但耗時，並且也增加醫生的工作量。所以利用電腦自動化診斷脊椎疾病是未來的趨勢，而自動診斷的依據往往是根據觀察椎間盤的面積、形狀的變化、或者是任何疾病的特徵，因此需要截取出最具影響力之特徵，使其達到有效切割與診斷之效果。
在第三章中，本論文提出一個自動化切割椎間盤以及自動診斷椎間盤突出的方法。其主要步驟分為：切割椎間盤、切割神經、自動診斷椎間盤突出。切割椎間盤方法為先去除背景和雜訊，之後分析椎間盤，這裡提出四個椎間盤特徵，用來濾除非椎間盤的物件，最後留下影像中的完整椎間盤。切割神經方法為利用影像中的像素分佈保留較亮的像素值，再利用型態學的方法處理雜訊，最後留下影像中的神經特徵。最後診斷椎間盤突出方法為找出椎間盤以及神經特徵重疊的地方，並判斷重疊的物件大小，並把椎間盤突出的物件標記出來。本技術能有效的找出異常的椎間盤部位其準確度平均最高為99.88%。

關鍵字

醫學影像切割與診斷 ； 脊椎影像 ； 椎間盤突出 ； 椎間盤

並列摘要

This thesis focuses on medical imaging segmenting technology in the study of spinal diseases. Diseases of the spine can be divided into: disc degeneration, disc herniation, vertebral compression fractures and bone spurs. These diseases mainly occur in aged and middle-aged people. Because with the increase of age, the human spine is slowly aging, diseases of the spine will become public. Disc diseases can be scanned through some medical equipment; and medical imaging can be divided into two large categories CT and MRI images. In medical imaging, the usual practice is to make use of the MRI T2 images to judge the diseases of the spine. Since making the correct judgments depends heavily on professional doctors, this is not only time-consuming but also increases the workload of doctors. Therefore, the future trend is to use computer to do automatic diagnosis of spinal diseases, and automatic diagnosis is based on observing the area, shape varieties, and other features of the disease of the intervertebral disc. Thus, it is essential to extract the most significant feature to effectively segment and diagnose intervertebral disc.
In Chapter 3, this thesis proposes an automatic method to segment the intervertebral disc and to diagnose the herniation of the intervertebral disc. This method contains the following major steps: segmenting the intervertebral disc, segmenting the nerve, and automatic diagnosis of the herniation of the intervertebral disc. The method of segmenting the intervertebral disc includes, first, removing the background and noises in the spine images, and then, analyzing the intervertebral disc. Here non- intervertebral disc objects are filtered by four intervertebral disc features, and finally the integral intervertebral disc objects will be kept. The method of segmenting the nerve includes, first, saving the brighter pixel value, and then removing the noises by morphology method. Finally, the nerve features will be kept in the image. The method of automatic diagnosis of the herniation of the intervertebral disc includes finding the overlapping objects between the intervertebral disc and the nerve image, and then labeling the objects of the intervertebral disc. This technology can effectively find out the abnormal parts of the intervertebral disc. Its average accuracy is as high as 99.88%.

並列關鍵字

medical image segmentation and diagnosis ； spine image ； herniation of intervertebral disc ； intervertebral disc

參考文獻

[1] M. Adams and P. Roughley, “What is Intervertebral Disc degeneration, and What Causes It?,” Spine, Vol. 31, No.18, 2006, pp. 2151-2161.

[2] R. Alomari, J. J. Corso and V. Chaudhary, “Labeling of Lumbar Discs Using Both Pixel and Object-Level Features With a Two-Level Probabilistic Model,” IEEE Transactions on Biomedical Engineering, Vol. 30, No. 1, 2011, pp. 0062-0278.

[5] J. J. Corso, R. S. Alomari, V. Chaudhary and G. Dhillon, “Lumbar Disc Localization and Labeling with a Probabilistic Model on Both Pixel and Object Features,” in Proc. Med. Image Computing Computer Assist. Intervent. (MICCAI), Vol. 5241, No. 1, 2008, pp. 202-210.

[7] S. Ghebreab and W. M. Smeulders, “Combining Strings and Necklaces for Interactive Three-Dimensional Segmentation of Spinal Images Using an Integral Deformable Spine Model,” IEEE Transactions on Biomedical Engineering, Vol. 51, No. 10, 2004, pp.1821-1828.

[8] L. Grady and E. L. Schwartz, “Isoperimetric Graph Partitioning for Image Segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 28, No. 3, 2006, pp. 469-475.