This paper has been published in Scientific Reports (2025). It presents an automated pipeline that translates MR images to CT-like contrast and applies super-resolution to enable accurate, unsupervised skull segmentation, dramatically speeding preparation of 3D skull geometry for implant design. The method reduces reliance on manual labels and improves robustness on clinical data, helping DeepImplant extract bone geometry faster and more reliably.

This paper appears in the ACCV proceedings (2024). It uses contrastive learning to translate MR scans into CT-like images so skull segmentation can be performed without CT, accelerating extraction of bone geometry needed for implant planning.

Read the paper

This paper has been published in Computer Methods and Programs in Biomedicine (2025). It introduces a deep-learning approach that enforces anatomical symmetry to reconstruct missing skull regions automatically, producing plausible implant candidates with minimal manual input. The symmetry-aware model simplifies the CAD modeling step and accelerates the path from imaging to implant-ready designs.

This paper has been published in Computers in Biology and Medicine (2024). It demonstrates that strong, diverse augmentation strategies - from registration-based transforms to latent-diffusion augmentations-— substantially improve reconstruction robustness for rare and noisy defects. The result is faster, more reliable automatic implant modeling that generalizes better to real clinical scans.