使用線性與非線性分類演算法測量 Spinal Canal之 CSA

近年來，醫學影像技術不斷的蓬勃發展，其中核磁共振造影（Magnetic Resonance Imaging）技術逐漸成熟，使其在醫學診斷上被廣泛地利用；核磁共振造影利用磁場改變氫原子的排列方式分析接收其所釋放的電磁波，以達到繪製人體內部組織精確影像的目的，對於醫學診斷可提供重要的資訊。脊椎狹隘症（Lumbar Spinal Stenosis）時常發生在65歲後的成年人身上，尤其是腰椎已經開始病變或退化的老人家，神經外科醫生必須透過核磁共振造影的成像來幫助其診斷病症，且在手術後的恢復狀況，也必須透過MR影像來判斷；通常醫生在觀察手術後的恢復情形時，必須透過一些市售的軟體以人工的方式對每一張核磁共振影像操作，此舉相當費時費力。 Cross section area of spinal canal是醫生用來判斷脊椎狹隘情形的重要指標，因此在本研究中，我們利用一些演算法包含k-NN、FLDA以及線性和非線性的SVM演算法以幫助圈選及量測該面積區域。由於該面積區域的形狀通常呈現不規則形而且核磁共振所能提供的頻譜資訊較少，本研究中方法雖然尚未能完全取代人工測量的方式，但未來可以成為準確的診斷工具。

關鍵字

spinal canal ；監督式分類學習法； k-最近鄰演算法；費雪線性判別分析演算法； support vector machine ； kernel函數

並列摘要

Remote sensing is a technique to get information about the region of interest without physical contact. Medical imaging could be seen as one of popular applications in remote sensing. With recent advances in medical imaging instruments magnetic resonance imaging (MRI) has become a reliable diagnostic technique, which provides a good contrast among different tissues of human body for clinical practice. Due to its strength to detect soft tissue, MRI has been widely used for diagnosis of diseases or symptoms. One of such examples is the cross-section area (CSA) of spinal canal. It has been an important indicator for lumbar spinal stenosis (LSS), which remains the leading preoperative diagnosis for adults older than 65 years. Presently this region can only be manually defined by doctors in axial T2-weighted MR images and measured by commercial software. This process causes the confines of spinal canal inconsistent and inaccurate. More importantly, the enclosed region is not objective neither reproducible. In this research, the linear and kernel based classification algorithms were investigated for measurement of CSA. Due to its irregularity in spatial shape and lack of spectral information in CSA, this approach has not been deployed to provide a robust measurement of CSA. The target algorithms include k-nearest neighbor (k-NN), Fisher’s linear discriminant analysis (FLDA), support vector machine (SVM) both in linear and kernel bases. To classify the desired section, we took the advantage of characteristic of spinal nerve roots and the cerebrospinal fluid (CSF) in axial T1-weighted and axial T2-weighted MR images acquired in clinical examination as features. The experimental results demonstrate that measurement of CSA by using linear and kernel based classification algorithms could be an accurate and consistent diagnosis tool.

並列關鍵字

spinal canal ； supervised classification ； k-Nearest Neighbor ； Fisher’s linear discriminant analysis ； support vector machine ； kernel function

參考文獻

[1] Nikolai Bogduk, Clinical anatomy of the lumbar spine and sacrum, fourth edition, 2005, pp. 55-57.

[2] Daffner SD, and Wang JC, “The pathophysiology and nonsurgical treatment of lumbar spinal stenosis,” Instr Course Lect., 2009, 58:657-68.

[4] Anil. K. Jain, Robert P.W. Duin, and Jiangchang Mao, “Statistical Pattern Recognition: A Review,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 22, No.1, January 2000.

[5] Sahibsingh A. Dudani, “The Distance-Weighted k-Nearest Neighbor Rule,” IEEE Transactions on System, Man, and Cybernetics, April 1976.

[6] V. Vapnik and A. Lerner, “Pattern recognition using generalized portrait Computer Vision and Pattern Recognition method,” Automation and Remote Control, 24, 1963.

國際替代計量

使用線性與非線性分類演算法測量 Spinal Canal之 CSA

全文下載

主題瀏覽