Graduate Student Seminar Series – Vishal Pendse

Graduate Student Seminar Series
Please ensure you invite your Principal Investigator by adding their email via the ‘Add Guest’ button and they will also be notified of your presentation.
Location: MS2158 – 1 King’s College Circle
Presentation Title: PACT: A Predictive Algorithm for Automated Upper-Limb Prosthetic Socket Design
Abstract: Well-fitting sockets are crucial for successful upper-limb prosthesis use, yet current digital socket design workflows are not standardized and demand considerable clinician effort. In this study, we introduce the Predictive Algorithm for Customized Transradial Socket Design (PACT), which generates socket models from a 3D limb scan. It works by retrieving the most similar limb–socket pair from a reference library of prosthetist-designed prosthetic sockets and applying isotropic and anisotropic scaling adjustments to match the input limb. To validate the algorithm, the PACT-predicted sockets for 19 participants were compared to their prosthetist-designed ones (the clinical “gold standard”) using the surface Euclidean (L2) distances, volume differences, and a 100-slice cross-sectional-area analysis. Localized discrepancies were mapped via signed-distance colorization and clustered with DBSCAN. PACT’s outputs differed from prosthetist designs by 2.11 ± 0.51 mm on the surface and 2.74 ± 2.56% in volume on average; slice-wise area differences were within ±10% for most of the socket length, with larger errors near the proximal trimline and distal tip. Recurrent localized discrepancies were found to be concentrated at the anterior-distal trimline (15/19 cases) and anterior–posterior compression (11/19 cases), indicating clear targets for rule-based or measurement-informed refinements. Subgroup patterns suggested an age-related bias (undersizing in pediatric, oversizing in adults). Overall, PACT quickly delivers (13.2 ± 0.7 s) a first-draft transradial socket within commonly cited clinical fit tolerances. Focus on specific regions, along with metadata such as tissue stiffness, age, and clinician-led measurements, can improve personalization and generalizability in future iterations of the PACT.
Supervisor Name: Jan Andrysek
Year of Study: 5
Program of Study: PhD
Powered by Calendly.com