Stroke survivors are often left with persistent hand dysfunction that limits their independence and quality of life. Now, a Johns Hopkins biomedical engineering master’s student team has developed a machine learning model that quantifies this hand dysfunction and could improve recovery outcomes.
The project was part of the Department of Biomedical Engineering’s Precision Care Medicine Course, and the team presented their innovation at the Whiting School of Engineering’s Design Day—an annual event showcasing students’ innovation and ability to translate theoretical knowledge into real-world solutions, held on April 28, 2026.
Currently, clinicians use the Fugl-Meyer Assessment (FMA), a performance-based impairment index, to evaluate motor function following a stroke. The lengthy evaluation is subjective and not precise enough to capture the rapid, dynamic coordination of hand movements.
Their new model will improve these drawbacks by using SenseHand, a medical device created by the Motor Recovery Research Lab in the Department of Physical Medicine and Rehabilitation at Hopkins, to collect quantitative data while a patient completes grasp-and-lift tasks. The lab is led by their clinical mentor, Preeti Raghavan.
From the data, they extract key features of motor control such as force timing, coordination between grip and load forces, and variability. Those features are integrated into machine learning models to identify patterns of impairment, group patients by severity, and track changes in motor function over the course of rehabilitation.
“In practice, the FMA remains the standard clinical benchmark, and our solution provides analysis of objective biomarkers to assess impairment and track functional recovery so that clinicians can provide personalized treatment,” said Saahil Sachdeva, team lead and biomedical engineering master’s student.
Their solution will reduce assessment time and could be performed remotely, both expanding access to care. In the long term, their model could be used to evaluate treatment therapies that reduce long-term disability from stroke.
Collaboration with clinical faculty helped the team focus on a real-world need and build a model that clinicians can easily understand and apply.
“Real-world healthcare problems are inherently complex and require a data-driven mindset. Rather than simplifying the data, we learned to work with that complexity to better reflect the realities of patient recovery,” said Sachdeva.
The team was advised by faculty mentors Joseph Greenstein, senior lecturer in biomedical engineering, Casey Overby Taylor, associate professor of biomedical engineering, and master’s students Maya Lane, and Aaron Roitman. Robert Nickl also served as a clinical mentor. The full team includes Prisha Jhala, Vani Padmakumar, Hritaal Saha, Ana Todesco, and Mytreyi Trivedi.