|
|
|
Moderator: Jonathan Bravman, MD
Moderator: David R McAllister, MD
|
|
|
| |
|
Introduction:
Across the globe, knee surgeries comprise the majority of surgical procedures in athletes, with anterior cruciate ligament reconstructions (ACLR) accounting for a substantial proportion of these operations. Given the relatively high frequency of these procedures within soccer and football, the management and rehabilitation of ALCR in this population presents a unique area of focus for novel rehabilitation protocols and technologies.
Arthrogenic muscle inhibition (AMI) is a well-recognized and relatively common complication following anterior cruciate ligament (ACL) reconstruction, occurring to some degree in nearly 60% of patients. AMI most commonly manifests as impaired quadriceps activation, which can result in gait abnormalities and present a significant barrier to effective postoperative rehabilitation. In professional athletes, delays in neuromuscular recovery may have substantial consequences, including prolonged return-to-play timelines, diminished team performance, and potential financial implications related to athlete compensation. Accordingly, strategies aimed at mitigating AMI and expediting rehabilitation are of particular importance in this population.
Previous work investigating both motor imagery and the integration of biofeedback into rehabilitation has demonstrated promise in mitigating deficits associated with arthrogenic muscle inhibition (AMI) and improving functional recovery. The purpose of the present study is to evaluate the efficacy of a novel electroencephalogram (EEG)–based biofeedback training protocol that integrates traditional motor imagery with real-time EEG feedback to enhance postoperative outcomes following anterior cruciate ligament reconstruction (ACLR).
This technology has previously been utilized by professional clubs across Major League Soccer (MLS), LaLiga, and Serie A to enhance athletic performance among non-injured players. By adapting this established performance-training platform for postoperative rehabilitation, the current study seeks to extend the application of this technology to the recovery setting. Leveraging EEG-guided motor imagery during rehabilitation may improve neuromuscular recovery, accelerate return-to-sport timelines, and facilitate a more efficient return to competitive play following ACLR.
Methods:
A preliminary exploratory comparison of two patients was performed between the first intervention subject and a matched control from an ongoing randomized, blinded clinical trial (Intervention: 29-year-old male, 86 kg; Control: 29-year-old female, 68kg). By April 2026, we expect to have reportable data from 10 participants. All participants followed a standardized postoperative physical therapy protocol, with the intervention group completing additional 20-minute EEG biofeedback sessions twice weekly for 8 weeks, starting at the first therapy visit. Training involves motor-imagery visualization with real-time EEG feedback to reinforce activation of motor pathways. Markerless motion capture recorded overground gait, bilateral squats, and forward lunges at 2, 4, and 6 months postoperatively, with outcome assessors blinded to group allocation. Knee kinematics, range of motion, peak knee flexion angles, and peak knee flexion moments were reported. Surface EMG was used to detect quadriceps and hamstring activation throughout the testing protocol.
Results:
Both patients demonstrated gains in stance-phase knee ROM over the follow-up period (Fig 1), though the intervention patient showed greater improvement. Stance-phase ROM more than doubled from 2 to 6 months (+7.2°), driven by increased early-stance flexion and improved mid-stance extension (+3.7°). The control patient showed smaller ROM gains (+2.9°) but similarly improved mid-stance extension by 6 months (+6.6°). First-peak knee flexion moments, however, diverged. By 6 months, the intervention patient increased from 0.13 to 0.29 %BW·m, indicating improved quadriceps loading, whereas the control patient decreased from 0.37 to 0.22 %BW·m, consistent with continued quadriceps avoidance. The intervention patient showed elevated semitendinosus activation at 4 months during lunges and squats, with a subsequent reduction at 6mo. This pattern was mirrored by the co-contraction index, which increased at 4mo and subsequently decreased at 6 months for both tasks. In contrast, the control subject exhibited higher rectus femoris activation at 4 months during both the lunge and squat, with a subsequent decrease at 6months, reflecting an opposing temporal pattern.
Conclusion:
Preliminary results suggest EEG biofeedback may enhance early quadriceps activation and stance-phase knee mechanics following ACLR. Overall, the intervention patient demonstrated more substantial gains in both ROM and flexion moment, suggesting improved quadriceps engagement and progression toward normative gait mechanics. Further data is needed to confirm these effects.
|
| |
|
Introduction
Artificial intelligence (AI) is rapidly reshaping orthopaedic care and clinical practice, with growing
applications in imaging, predictive modeling, and surgical decision-making. Rotator cuff tears represent a
prevalent and formidable challenge in the field of shoulder surgery. Achieving an accurate diagnosis is
paramount, as it lays the foundation for tailored treatment planning and significantly influences the
prognosis. Moreover, implementing effective strategies for re-tear prevention is crucial in ensuring
optimal surgical outcomes. Addressing these elements with rigor not only enhances patient recovery but
also reduces the long-term burden of shoulder dysfunction.AI-based imaging models have demonstrated
improved accuracy in detecting and characterizing rotator cuff pathology, while deep learning algorithms
have been developed to predict re-tear risk following repair. Despite the progress made in the field, the
existing literature remains disjointed, with the majority of studies focusing on specific, isolated aspects of
care. This systematic review seeks to integrate the prevailing evidence regarding the application of
artificial intelligence in the management of rotator cuff tears throughout the perioperative continuum,
encompassing preoperative imaging, surgical planning, prognostic modeling, and postoperative
rehabilitation.
Methods
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews
and Meta-Analyses (PRISMA) guidelines. A comprehensive literature search was performed across
PubMed, Scopus, and Embase from database inception to May 2025. The search strategy combined terms
related to artificial intelligence, machine learning, and deep learning with terms for rotator cuff tears and
rotator cuff repair. Eligible studies included those addressing AI-based imaging diagnosis, prognostic
modeling, surgical planning, or postoperative outcome prediction.
The following data were extracted and recorded: image modality/plane, input features, age, gender,
diagnosis (rotator cuff tears), ground truth references, AI algorithm, pretrained CNN, size of training set,
size of testing set, size of validation set or validation method, and model performance (accuracy,
sensitivity, specificity, AUC). The data extracted from the included studies were narratively reviewed.
Results
A total of 59 relevant studies were incorporated into the analysis. The results demonstrate that artificial
intelligence (AI) exhibits strong diagnostic accuracy for rotator cuff tears across various imaging
modalities. Convolutional neural network (CNN)-based models utilizing magnetic resonance imaging
(MRI) achieved diagnostic accuracies exceeding 90% in multiple studies, with ultrasound-based models
exhibiting comparable performance. Several studies applied AI to surgical planning, using preoperative
imaging and clinical features to guide decision-making and optimize repair strategies. Prognostic
applications focused on re-tear risk and functional outcomes: deep learning models achieved AUCs of
0.87–0.92 for re-tear prediction. In contrast, machine learning approaches predicted postoperative
improvement with accuracies up to 96.9% internally and 79.6% externally validated. Across studies,
ensemble methods (XGBoost, LightGBM) consistently outperformed logistic regression, underscoring the
potential of AI to extend beyond diagnosis into prognosis and recovery trajectories.
Discussion
This systematic review illustrates the expanding role of artificial intelligence throughout the continuum of
rotator cuff management, encompassing preoperative imaging interpretation to postoperative outcome
prediction. An analysis of over 230,000 patients and imaging datasets reveals that AI consistently
demonstrates robust diagnostic accuracy, particularly in the detection of conditions via MRI and
ultrasound. These findings underscore the potential of AI as a dependable complement to the work of
radiologists. Prognostic models, including deep learning algorithms predicting retear risk after repair,
showed AUC values exceeding 0.80 in several studies, highlighting their promise in stratifying high-risk
patients. Applications in surgical planning and rehabilitation remain underexplored, though preliminary
work suggests feasibility in quantifying tendon healing and guiding individualized recovery. These
findings suggest AI can augment each step of rotator cuff management by enhancing diagnostic precision,
informing surgical decision-making, and predicting outcomes. Nonetheless, heterogeneity in datasets and
lack of external validation remain significant barriers to clinical adoption. Future multicenter prospective
studies are needed to validate these tools and integrate AI into clinical workflows.
|