Abstract
Background
Patients who have undergone Anterior Cruciate Ligament Reconstruction (ACLR) have a 6-24% chance of either re-tearing or having subsequent knee surgery. To date there have been no practical validated risk prediction models that can be easily implemented into clinical workflow for re-injury risk. Micro-Doppler radar (MDR) provides a promising solution.
Objective
The purpose of this study was to investigate the predictive ability of MDR to identify persons with a previous ACLR relative to an age and sex matched healthy control.
Methods
ACLR patients (n=81) and controls (n=100) performed drop box jump, sit to stand (STS), and walking trials as MDR signatures were collected. A 1D Convolutional Neural Network was developed to evaluate each activity individually followed by the development of a fusion model validation using all three activities.
Results
The STS model individually achieved the highest overall accuracy of 82.3%, with a sensitivity of 71.6% and specificity of 91.0%. The fusion model using all activities achieved a peak overall accuracy to detect ACLR of 86.2%, 80.3% sensitivity, and 91% specificity.
Conclusions
Currently, there is no clinically validated, efficient approach to objectively evaluate human motion at the point of care. When coupled with machine learning, MDR accurately differentiates ACLR from control groups by identifying complex biomechanical asymmetries, with classification performance comparable to or exceeding that of motion capture. Future research is needed to determine if MDR can be used in conjunction with risk prediction modeling.
Key points
Micro-Doppler radar provides a promising new solution to identify important human motion asymmetries in clinical settings. Here we evaluated a group of patients who have a history of Anterior Cruciate Ligament reconstruction versus a control group. Simple movements performed in the presence of the micro-Doppler radar system were used to identify the 2 groups with accuracy comparable or superior to motion capture systems.
Full Text Availability
The license terms selected by the author(s) for this preprint version do not permit archiving in PMC. The full text is available from the preprint server.
