The purpose of this investigation was to explore the potential for using a neural network approach in objectively determining the status of ACL (anterior cruciate ligament) rehabilitation. Nine male subjects (three each of non-injured, in-rehab, post-rehab conditions) were matched for size and strength. The testing protocol was established by a physical therapist: each leg was tested at three different speeds on a Cybex machine with electromyography (EMG) and torque data collected concurrently. Neural net training and testing sets were created from the values of peak torque, time to peak torque, EMG mean power frequency, the EMG root mean square (RMS), and a value representing the ACL condition of the tested leg from each of 15 repetitions. The network was then run until reaching a pre-set mean square error value of 0.05 during training. Subsequent test sets were correctly classified with 99% accuracy, demonstrating the potential of the method as a diagnostic aid for detecting and identifying changes over time in the rehabilitation patterns of individuals having ACL injuries.
|Number of pages||3|
|Journal||Biomedical Sciences Instrumentation|
|State||Published - 1996|
|Event||Proceedings of the 1996 33rd Annual Rocky Mountain Bioengineering Symposium & 33rd International ISA Biomedical Sciences Instrumentation Symposium - Colorado Springs, CO, USA|
Duration: Apr 12 1996 → Apr 13 1996
- ACL injuries
- Neural nets