Finding the needle in the haystack of isokinetic knee data: Random Forest modelling improves information about ACLR-related deficiencies.

The difficulties of rehabilitation after anterior cruciate ligament (ACL) injuries, subsequent return-to-sport (RTS) let alone achieving pre-injury performance, are well known. Isokinetic testing is often used to assess strength capacities during that process. The aim of the present machine learning (ML) approach was to examine which isokinetic data differentiates athletes post ACL reconstruction (ACLR) and healthy controls. Two Random Forest models were trained from data of unilateral concentric and eccentric knee flexor and extensor tests (30°/s, 150°/s) of 366 male (63 post ACLR) as well as 183 female (72 post ACLR) athletes. Via a cross-validation predictive performance was evaluated and the Random Forest showed outstanding results for male (AUC = 0.90, sensitivity = 0.76, specificity = 0.88) and female (AUC = 0.92, sensitivity = 0.85, specificity = 0.89) athletes. The Accumulated Local Effects plot was used to determine the impact of single features on the predictive likelihood. For both male and female athletes, the ten most impactful features either referred to the disadvantageous (injured, non-dominant in control group) leg or to lateral differences. The eccentric hamstring work at 150°/s was identified as the most impactful single parameter. We see potential for improving the RTS process by incorporating and combining measures, which focus on hamstring strength, leg symmetry and contractional work.