Scaling nnU-Net for CBCT Segmentation
This work addresses segmentation challenges in dental CBCT imaging, but it is incremental as it builds upon the existing nnU-Net framework.
The paper tackled multi-structure segmentation on dental CBCT images by scaling the nnU-Net framework, achieving a mean Dice coefficient of 0.9253 and HD95 of 18.472, which secured first place in the ToothFairy2 challenge.
This paper presents our approach to scaling the nnU-Net framework for multi-structure segmentation on Cone Beam Computed Tomography (CBCT) images, specifically in the scope of the ToothFairy2 Challenge. We leveraged the nnU-Net ResEnc L model, introducing key modifications to patch size, network topology, and data augmentation strategies to address the unique challenges of dental CBCT imaging. Our method achieved a mean Dice coefficient of 0.9253 and HD95 of 18.472 on the test set, securing a mean rank of 4.6 and with it the first place in the ToothFairy2 challenge. The source code is publicly available, encouraging further research and development in the field.