PROJECT SUMMARY/ABSTRACT
It is widely accepted that failure to restore pre-injury biomechanics after anterior cruciate ligament
reconstruction (ACL-R) surgery is one of the key contributing factors to the high prevalence of post-traumatic
osteoarthritis (PTOA). Precision rehabilitation, which refers to the delivery of the right feedback to the right
patient at the right time, is now a feasible approach for PTOA prevention given recent advances in wearable
sensing and computer vision technologies. Flexible and unobtrusive skin patches can objectively quantify
movement out of the clinic and deliver real-time haptic feedback, while simple videos from smartphones can
assess physical therapy quality and deliver corrective visual or auditory feedback. To effectively apply
emerging smart-health technologies toward PTOA prevention, the multi-modal and multivariate data produced
by these sensors must be distilled to identify digital biomarkers of PTOA that can be targeted with biofeedback
therapy in the future. Accordingly, the central objective of this proposed work is to determine if characteristics
of gait extracted from video and wearable sensors (digital biomarkers) can predict longitudinal changes in
cartilage microstructure (early PTOA) extracted from quantitative Magnetic Resonance Imaging (qMRI). Our
central hypothesis is that future risk of PTOA can be predicted in the first few months after surgery using
passively collected data from wearable sensors and video. This hypothesis is supported by our previous work
on pre-arthritic subjects, where we demonstrated that wearable sensing data could predict detrimental changes
in cartilage microstructure that are indicative of OA risk. To accomplish the overall objective of this work,
physical therapy, natural environment ambulation, and cartilage health will be monitored longitudinally.
Exercise correctness during pre- and post-operative physical therapy will be quantified using computer vision
and machine learning algorithms. Out-of-lab movement will be monitored at baseline (3 weeks), 3, and 9
months after surgery with epidermal sensors placed on the thighs and shanks. Quantitative MRI data will be
collected at baseline (3 weeks), 3 and 18 months after the surgery. Specifically, we will determine (1) if gait
symmetry restoration measured by wearable sensors can predict qMRI changes up to 18 months post-surgery
and (2) if physical therapy quality, to the extent that is quantifiable with passive computer vision algorithms, can
predict gait symmetry restoration up to 9 months post-surgery. This work is innovative because it breaks with
the current norms of studying the role of biomechanics in PTOA in the laboratory. Instead, we will use
wearable sensing, computer vision, and machine learning to generate previously unavailable knowledge on the
role of natural environment biomechanics. If successful, this work could enable personalized, technology-
assisted rehabilitation—a paradigm shift in clinical care. Additionally, the discovery of new PTOA biomarkers
could improve the efficiency of clinical trials for new surgical techniques, while the proposed framework is also
extensible to the study and prevention of primary OA, and possibly other orthopaedic conditions.
