Abstract
BACKGROUND: Ballet is a highly technical and physically demanding dance form involving extensive end-range lumbar movements and emphasizing movement smoothness and gracefulness. A high prevalence of non-specific low back pain (LBP) is found in ballet dancers, which may lead to poor controlled movement and possible pain occurrence and reoccurrence. The power spectral entropy of time-series acceleration is a useful indicator of random uncertainty information, and a lower value indicates a greater smoothness or regularity. The current study thus applied a power spectral entropy method to analyze the movement smoothness in lumbar flexion and extension in healthy dancers and dancers with LBP, respectively. METHOD: A total of 40 female ballet dancers (23 in the LBP group and 17 in the control group) were recruited in the study. Repetitive end-range lumbar flexion and extension tasks were performed and the kinematic data were collected using a motion capture system. The power spectral entropy of the time-series acceleration of the lumbar movements was calculated in the anterior-posterior (AP), medial-lateral (ML), vertical (VT), and three-directional (3D) vectors. The entropy data were then used to conduct receiver operating characteristic curve analyses to evaluate the overall distinguishing performance and thus cutoff value, sensitivity, specificity, and area under the curve (AUC) were calculated. RESULTS: The power spectral entropy was significantly higher in the LBP group than the control group in the 3D vector in both lumbar flexion and lumber extension (flexion: p = 0.005; extension: p < 0.001). In lumbar extension, the AUC in the 3D vector was 0.807. In other words, the entropy provides an 80.7% probability of distinguishing between the two groups (i.e., LBP and control) correctly. The optimal cutoff entropy value was 0.5806 and yielded a sensitivity of 75% and specificity of 73.3%. In lumbar flexion, the AUC in the 3D vector was 0.777, and hence the entropy provided a probability of 77.7% of distinguishing between the two groups correctly. The optimal cutoff value was 0.5649 and yielded a sensitivity of 90% and a specificity of 73.3%. CONCLUSIONS: The LBP group showed significantly lower lumbar movement smoothness than the control group. The lumbar movement smoothness in the 3D vector had a high AUC and thus provided a high differentiating capacity between the two groups. It may therefore be potentially applied in clinical contexts to screen dancers with a high risk of LBP.