Scalable digital solutions for reconstructive surgery. Rapid prediction of 3D face and body morphology in low-resource conditions

Abstract

This project advances AI-driven 3D and AR content creation for reconstructive and aesthetic surgery by developing a mobile-first, privacy-preserving platform for the automated generation of high-quality 3D face and body models from 2D images. Today's tools depend on special scanners, full video capture, or cloud processing, making them costly, hard to access, and a concern for patient privacy. We address these limits using generative AI, Gaussian splatting, and neural rendering to reconstruct reliable 3D models from just a handful of images. The work combines research on sparse-view reconstruction with methods for generating pseudo-ground-truth models and building real-world capture datasets. We also study how to run these models directly on smartphones and tablets, so 3D modeling works in real time without sending data to the cloud and patient privacy is protected by design. The goal is a scalable and affordable solution that lowers the barrier for clinics.