The Study Of Key Features Extraction And Quantization Of Spinal Image In Computer-aided Clinical Diagnosis

Lower back pain(LBP) is widely prevalent all over the world and more than 80% of the people suffer from it. Although the causes of LBP remains to be determined, it is generally believed that the degeneration of the intervertebral disc is related to the original pain. Traditionally,the T2-weighted magnetic resonance image(MRI) which can reveal disc morphology coupled with a clinical exam assessing functional range of motion(ROM) is used to evaluate spinal health. However, these subjective measures fail to correlate well with the pain or provide useful patient stratification. In consequence, the quantification of spinal motion and objective MRI measures of disc health are needed to be improved. In this paper, we researched on the classification of intervertebral disc degeneration and intervertebral motion patterns based on MRI and quantitative fluoroscopy(QF) respectively. The contents are divided into the following three aspects:First, a Gabor filter based method for localizing intervertebral disc(IVD) was presented which contains 3 steps. During the first step, a series of Gabor-filtered spine images were obtained in order to extract the information of spine edges and disc edges; then, based on which the areas of spines could be determined. After that, in conjunction with the prior knowledge,the exact location of IVDs in the spines were calculated.Second, according to the localization results an elliptical fitting was performed followed by a comparison of the distribution of gray value and Gabor coefficients on this area; we proposed a method to distinguish the nucleus pulposus and annulus fibrosis. Then the classification of IVD degeneration was achieved by associating gray information of nucleus pulposus and annulus fibrosis with the information of height.Third, based on Forbin’s theory, two methods including mutual information(MI) method and Kanade-Lucas-Tomasi(KLT) were utilized in order to tracking the vertebra. And finally the tracking of vertebra both in coronal plane of side-bending and sagittal plane of flexion and extension with KLT were realized. And then the clinical parameters were calculated.The experiments of the MRI were performed in a dataset of 37 patients, the results showed that our method is simpler meanwhile with a close accuracy of 96.6% compared with the existing methods. In addition, the precision has been increased to a level of 1.46 mm. And the degeneration classification of level 1-5 could be realized by our classification algorithm. While in the study of lumbar intervertebral motion patterns, the experimental results demonstrated that KLT based tracking method performed better in terms of accuracy and efficiency than MI method. A model simulation results verified the accuracy of parameters obtained by our methods.