New paper! Self-Supervised Deformable Masked Autoencoders for Cranial Defects

June 3, 2024

This conference paper appears in the IEEE EMBC proceedings (2024). It proposes self-supervised deformable masked autoencoders that learn structural priors from unlabelled data to reconstruct cranial defects without dense ground truth, reducing annotation needs and producing high-quality reconstructions for CAD pipelines.

Read the paper

InnoDiament Award!

October 20, 2023

DeepImplant

Marek Wodzinski has been awarded the "Scientist of the Future" award organized by the Smart Development Forum 🥉. The award is connected with the DeepImplant project about developing novel artificial intelligence methods dedicated to personalized cranial implant modeling.

Moreover, Marek had the opportunity to present the overview of the DeepImplant project to a wider audience during the congress in Uniejów!

Early acceptance at MICCAI! From Sparse Points to High-Resolution Cranial Surfaces with Transformers

October 1, 2023

DeepImplant

This MICCAI paper (2023) presents an iterative, transformer-based point-cloud completion pipeline that converts low-resolution point sets into high-fidelity cranial surfaces while keeping computation efficient, producing implant-ready meshes that improve automated CAD outputs.

DeepImplant Project Started!

March 1, 2023

DeepImplant

We are excited to announce the official launch of DeepImplant — a multidisciplinary project that brings together AI, medical imaging, and mixed reality to radically speed up and simplify the design of patient-specific cranial implants. DeepImplant aims to automate CAD modeling so clinicians and technicians can generate personalized implants in just a few seconds without specialist biomechanical knowledge, and to let users verify modeling correctness interactively using mixed reality.

What the project will deliver

Fast, automated CAD modeling that produces implant-ready geometry in seconds, reducing weeks of work to minutes
Accessible workflows that hide modeling complexity behind intuitive AI tools so non-expert users can create safe, clinically useful implants
Mixed reality verification so surgeons and engineers can inspect and validate designs in 3D before manufacturing
Clinical-ready integration with accepted medical-manufacturing processes to shorten time from imaging to surgery

Selected initial research

Wodziński, M.; Daniol, M.; Socha, M.; Hemmerling, D.; Stanuch, M.; Skalski, A. (2022). Deep Learning-Based Framework for Automatic Cranial Defect Reconstruction and Implant Modeling. Computer Methods and Programs in Biomedicine. Read
Wodziński, M.; Daniol, M.; Hemmerling, D. (2021). Improving the Automatic Cranial Implant Design in Cranioplasty by Linking Different Datasets. In: MICCAI 2023 Challenges. Read
Kwarciak, K.; Wodziński, M. (2023). Deep Generative Networks for Heterogeneous Augmentation of Cranial Defects. In: ICCV 2023 Workshop. Read