Quantitative Ultrasound Using Texture Analysis of Myofascial Pain Syndrome in the Trapezius

Abstract

Objective-The objective of this study is to assess the discriminative ability of textural analyses to assist in the differentiation of the myofascial trigger point (MTrP) region from normal regions of skeletal muscle. Also, to measure the ability to reliably differentiate between three clinically relevant groups: healthy asymptomatic, latent MTrPs, and active MTrP. Methods-18 and 19 patients were identified with having active and latent MTrPs in the trapezius muscle, respectively. We included 24 healthy volunteers. Images were obtained by research personnel, who were blinded with respect to the clinical status of the study participant. Histograms provided first-order parameters associated with image grayscale. Haralick, Galloway, and histogram-related features were used in texture analysis. Blob analysis was conducted on the regions of interest (ROIs). Principal component analysis (PCA) was performed followed by multivariate analysis of variance (MANOVA) to determine the statistical significance of the features. Results-92 texture features were analyzed for factorability using Bartlett's test of sphericity, which was significant. The Kaiser-Meyer-Olkin measure of sampling adequacy was 0.94. PCA demonstrated rotated eigenvalues of the first eight components (each comprised of multiple texture features) explained 94.92% of the cumulative variance in the ultrasound image characteristics. The 24 features identified by PCA were included in the MANOVA as dependent variables, and the presence of a latent or active MTrP or healthy muscle were independent variables. Conclusion-Texture analysis techniques can discriminate between the three clinically relevant groups.

Feasibility of a Support Vector Machine Classifier for Myofascial Pain Syndrome: Diagnostic Case‐Control Study

“…Shah et al describe hypoechoic regions associated with palpable MTrPs, but in our experience, these are extremely difficult to find clinically using B‐mode US. Furthermore, our previous work as well as that of Sikdar et al suggest that the abnormalities shown by quantitative US image analyses of the muscles affected by myofascial pain have regional abnormalities that may not be confined to the area of the palpable MTrP . For these reasons, we acquired a video from the entire upper fibers of the trapezius.…”

“…Images were acquired on an X‐porte US system (SonoSite Canada, Toronto, Ontario, Canada) with a linear US transducer. The acquisition procedure described below was the same used in previous studies . Videos were acquired at a frequency of 5 to 15 MHz at a depth of 2.7 cm.…”

“…After the preprocessing was completed, the texture features were extracted from the remaining ROIs. The texture features included the gray‐level co‐occurrence matrices and gray‐level run length matrices, and first‐order statistical features were extracted from the training data set, similar to previous works . In total, the number of features before model construction was 88 (76 gray‐level co‐occurrence matrices, 7 gray‐level run length matrices, and 5 first‐order features).…”

“…Despite these results, texture analytic techniques have not been applied further to characterize muscles of patients with MPS. Last, we have recently described texture features that can be used to effectively discriminate between participants who met the clinical criteria for MPS with active or latent MTrPs from healthy control participants …”

“…The aim of this study was to build on our recent work on texture features by using them in conjunction with machine learning in a CAD approach. More specifically, we set out to investigate the feasibility of a support vector machine (SVM) binary classification algorithm to differentiate between healthy and abnormal (latent/active) muscle.…”