Introduction: Low back pain (LBP) is a prevalent and complex condition that poses significant medical, social, and economic burdens worldwide. The accurate and timely assessment and diagnosis of LBP, particularly non-specific LBP (NSLBP), are crucial to developing effective interventions and treatments for LBP patients. In this study, we aimed to investigate the potential of combining B-mode ultrasound image features with shear wave elastography (SWE) features to improve the classification of NSLBP patients.Methods: We recruited 52 subjects with NSLBP from the University of Hong Kong-Shenzhen Hospital and collected B-mode ultrasound images and SWE data from multiple sites. The Visual Analogue Scale (VAS) was used as the ground truth to classify NSLBP patients. We extracted and selected features from the data and employed a support vector machine (SVM) model to classify NSLBP patients. The performance of the SVM model was evaluated using five-fold cross-validation and the accuracy, precision, and sensitivity were calculated.Results: We obtained an optimal feature set of 48 features, among which the SWE elasticity feature had the most significant contribution to the classification task. The SVM model achieved an accuracy, precision, and sensitivity of 0.85, 0.89, and 0.86, respectively, which were higher than the previously reported values of MRI.Discussion: In this study, we aimed to investigate the potential of combining B-mode ultrasound image features with shear wave elastography (SWE) features to improve the classification of non-specific low back pain (NSLBP) patients. Our results showed that combining B-mode ultrasound image features with SWE features and employing an SVM model can improve the automatic classification of NSLBP patients. Our findings also suggest that the SWE elasticity feature is a crucial factor in classifying NSLBP patients, and the proposed method can identify the important site and position of the muscle in the NSLBP classification task.

Low back pain (LBP) is a problem that raises medical, social and economic concerns. Accurate and timely assessment and diagnosis of LBP, especially non-specific LBP (NSLBP), can help clinicians develop effective interventions and treatments for LBP patients. In this study, we integrated the B-mode ultrasound image feature from multiple sites with the shear wave elastography (SWE) feature of NSLBP patients, and then employed a support vector machines (SVM) model to classify NSLBP patients with the Visual Analogue Scale (VAS) as the ground truth. A total of 52 subjects were recruited from the University of Hong Kong-Shenzhen Hospital, and an optimal feature set with the size of 48 was obtained after feature extraction and feature selection, among which the SWE elasticity feature had most contribution in the classification task. The five-fold cross-validation accuracy, precision and sensitivity were 0.85, 0.89 and 0.86, respectively, higher than the previously reported values of MRI. The experimental results preliminarily demonstrate that the proposed methods can be applied to the automatic classification of NSLBP and find out the important site and position of the muscle in the NSLBP classification task.