Development and application of a prosthetist-specific rectification template based on artificial intelligence for the fabrication of transfemoral prosthetic sockets

The prosthetic socket is the most critical component of a lower limb prosthesis, requiring precise customization to the individual’s residual limb. This study presents proof-of-concept for a novel artificial intelligence (AI)-driven rectification template for transfemoral sockets tailored to a single prosthetist. Using a dataset of nine persons with transfemoral amputation, the study workflow required the manual casting and 3D scanning of unrectified and rectified plaster positives, anatomical landmark identification, and unsupervised training of an algorithm using Principal Component Analysis (PCA). The AI captured both explicit and implicit rectification strategies, generating an average rectification template and synergistic modes of variation. Validation was conducted via a leave-one-out approach, comparing AI-generated versus manually crafted rectified positives using clinically-relevant metrics: perimeter and volume differences. The first four PCA modes explained 78% of rectification variability, with key modifications observed in distal and medial regions. Volume differences between AI and manually rectified positives were within clinically acceptable limits for all participants, with 44% rated “good” and 56% “acceptable”. This proof-of-concept provides a new perspective on the application of AI to replicate prosthetist-specific rectification strategies for transfemoral sockets. It may potentially help streamline fabrication by capturing both explicit and implicit prosthetist knowledge. The approach may be useful in clinical training, documentation, and socket fabrication, particularly in resource-limited settings, and may contribute to more consistent and efficient prosthetics care.