UCSF

In multiple sclerosis (MS), upper extremity dysfunction is highly prevalent. Hand function is critical to activities of daily living (ADL), and a change in function can significantly impact the ability to participate in self-care, occupational and recreational activities. Current assessments of hand function are exclusively performed in the clinic, and do not adequately quantify the quality and variety of movement needed to identify specific dexterity impairments. Moreover, recent research has identified a relationship between diminished hand function and worsening disease progression, suggesting that assessments of hand function with this predictive component are needed. While previous work has been done to explore the use of digital tools to meet this need, large-scale uptake has been stymied due to rapidly changing software and hardware and well as the cost of devices such as wearable sensors. Self-uploaded videos represent a more patient-friendly target for collecting functional data regularly, without cost and technology burdens. The data generated in this body of work demonstrate that human pose estimation using patient-generated videos is not only a viable method of quantifying dexterity in relationship to common clinical methods of assessment, but is also cost-effective and scalable. Exploratory data show that pose estimation is capable of identifying changes in dexterity longitudinally and that it has the potential to serve as a clinical biomarker for future trials. I also present feasibility and participant acceptability data which demonstrate that this method was well-received and simple to use for participants. This method is capable of identifying changes in dexterity longitudinally and captures patient-reported change in daily tasks. Collectively, these findings advance our understanding of digital tools for frequent, granular, and user-friendly dexterity assessments in MS.