Plausible but Unimplantable: Foundation Models vs Cranial Implants
An audit of three released shape completion models on 239 skull defects found no output that could be placed in a patient.
October 25, 2026
Guides, project breakdowns, and lessons learned from building medical imaging platforms — written by the PYCAD team.
An audit of three released shape completion models on 239 skull defects found no output that could be placed in a patient.
A post-training mixed-precision method that checks calibration and uncertainty maps, not only Dice, before it quantizes a layer.
A text-conditioned 3D diffusion model cleans up masks from TotalSegmentator or MedSAM, on CT and MRI, without reading the image.
Averaging InnerEye, FastSurfer and HippoSeg gives calibrated hippocampus masks plus an entropy score that sends doubtful cases to review.
A training-free SMPL fit fills in arms and hands cut off by the CT field of view, so body size metrics stop coming out short.
Open-source desktop tool that keeps metrics, reader scores and radiomic features tied to the same images. Apache 2.0.
We build custom medical imaging platforms — advanced DICOM viewers, AI segmentation, and the clinical systems around them.
Copyright © 2026 PYCAD. All Rights Reserved.