Knee osteoarthritis (KOA), a degenerative joint disease characterized by the progressive breakdown of articular cartilage, is a leading cause of disability worldwide. Due to the irreversible nature of cartilage damage, total knee replacement is currently the prominent end-stage treatment. However, because knee joint loading affects KOA progression, there is growing interest in non-surgical interventions like gait retraining - teaching patients to walk in a way that could reduce loading on the medial compartment of the knee - to delay the need for surgery. One intuitive strategy is to adjust the foot progression angle, or the angle of the foot relative to the direction of motion.

Despite its promise, gait retraining faces three key obstacles. First, effectively prescribing a patient's new foot progression angle remains challenging: identical toe-in modifications can yield vastly different outcomes and require patient-specific tuning that typically relies on time-consuming and equipment-dependent laboratory analyses. Second, measuring whether patients are actually achieving the intended foot progression angle (rather than direct in-vivo load measures) is complicated by the need for reliable motion tracking outside specialized clinics. Third, delivering real-time feedback can improve gait modifications but requires unobtrusive hardware and accurate signals.

To address these issues, we developed a regression model that uses minimal clinical data to predict knee adduction moment (KAM) reductions from toe-in gait []. Because KAM is closely linked to mediolateral load distribution in the knee, this tool helps clinicians predict the outcome of personalize interventions without the need for traditional gait lab equipment.

We then investigated how measurement accuracy influences gait retraining through a user study in which participants wore a commercial gait retraining device to track foot progression angle while receiving vibrotactile feedback. The results of this study highlighted a reduction in training efficacy when the commercial device produced inconsistent signals due to low tracking accuracy []. Using the commercial device led to the creation of ARIADNE, an open-source wearable vibrotactile device that, when adhered to opposing sides of a limb, can provide precise, bidirectional corrective cues using two linear resonant actuators []. ARIADNE addresses donning inconsistencies observed with commercial devices, specifically related to attachment method and placement on the limb []. Additionally, it quantifies the impact of device placement and individual tissue properties on both the actual and perceived vibration using measurements from its two accelerometers []. In a subsequent study, ARIADNE provided precise vibrotactile cues during gait retraining, and participants who received feedback showed improved learning of a new gait compared to those who did not use feedback [].

Collectively, these contributions help bridge the gap between research on gait retraining as a method of reducing joint loading in KOA and its broader clinical adoption, offering a pathway to more personalized and effective interventions.

This research project involves collaborations with Reed Ferber (University of Calgary), Eni Halilaj (Carnegie Mellon University), Owen Pearl (Nike Sport Research Lab), and Peter B. Shull (Shanghai Jiao Tong University).