Objectives
LIMSC will leverage retrospective datasets of multi-view images of patients collected prior to the project.
These videos will be processed with the current best 3D reconstruction algorithm [1], and SKEL [2] will be registered to the reconstructions to obtain 4D temporal trajectories.
The processing will take place on the GRICAD cluster (https://gricad-doc.univ-grenoble-alpes.fr/en/). Creating a dedicated processing pipeline will be key in the project, so that novel acquired data can be automatically processed, and a quality control can be effectuated.
As SKEL [2] was learned from an adult cohort and not adolescents, improvements will need to be added to the model and the registration pipeline. Also, as the back of AIS patients might exhibit gibossities and SKEL does not model them, further improvements will be required.
[1] Toussaint, Briac, Diego Thomas, and Jean-Sébastien Franco. "ProbeSDF: Light Field Probes For Neural Surface Reconstruction." In Proceedings of the Computer Vision and Pattern Recognition Conference, pp. 11026-11035. 2025.
[2] Keller, Marilyn, Keenon Werling, Soyong Shin, Scott Delp, Sergi Pujades, C. Karen Liu, and Michael J. Black.
"From Skin to Skeleton: Towards Biomechanically Accurate 3D Digital Humans." ACM Transactions on Graphics (TOG) 42, no. 6 (2023): 1-12.
