A novel scoring approach for the Wolf Motor Function Test in stroke survivors using motion-sensing technology and machine learning: a preliminary study
Bo Sheng,Xiaohui Chen,Jian Cheng,Yanxin Zhang,Shane Xie,Jing Tao,Chaoqun Duan,Shane (Sheng Quan) Xie
DOI: https://doi.org/10.1016/j.cmpb.2023.107887
IF: 6.1
2023-10-27
Computer Methods and Programs in Biomedicine
Abstract:Background and objective Human-administered clinical scales, such as the Functional Ability Scale of the Wolf Motor Function Test (WMFT-FAS), are widely utilized to evaluate upper-limb motor function in stroke survivors. However, these scales are generally subjective and labor-intensive. To end this, we proposed a novel scoring approach for the motor function assessment. Methods The proposed novel scoring approach mainly contained one Microsoft Kinect v2, one customized motion tracking system, and one customized intelligent scoring system. Specifically, the Kinect v2 was used to capture stroke survivors' functional movements, the motion tracking system was developed for recording the gathered movement data, and the intelligent scoring system (kernel: feed-forward neural network, FFNN) was developed to evaluate movement quality and provide corresponding WMFT-FAS scores. Several methods have been applied to enhance the approach's usability, such as singular spectrum analysis and multi-ReliefF method. Results Sixteen stroke survivors and ten healthy subjects were recruited for validation. Inspiring results of the proposed approach were achieved when compared with the clinical scores provided by a physiotherapist: 0.924 ± 0.027 for accuracy, 0.875 ± 0.063 for F1-score, 0.915 ± 0.051 for sensitivity, 0.969 ± 0.013 for specificity, 0.952 ± 0.038 for AUC, 0.098 ± 0.037 for mean absolute error, and 0.214 ± 0.078 for root mean squared error. Conclusions The results indicate that the proposed novel scoring approach can provide objective and accurate assessment scores, which can help physiotherapists make individualized treatment decisions.
engineering, biomedical,computer science, interdisciplinary applications,medical informatics, theory & methods