Scaling nnU-Net for CBCT Segmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Isensee, Fabian, Kirchhoff, Yannick, Kraemer, Lars, Rokuss, Maximilian, Ulrich, Constantin, Maier-Hein, Klaus H.
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909411444785152
author Isensee, Fabian
Kirchhoff, Yannick
Kraemer, Lars
Rokuss, Maximilian
Ulrich, Constantin
Maier-Hein, Klaus H.
author_facet Isensee, Fabian
Kirchhoff, Yannick
Kraemer, Lars
Rokuss, Maximilian
Ulrich, Constantin
Maier-Hein, Klaus H.
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2411_17213
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Scaling nnU-Net for CBCT Segmentation
Isensee, Fabian
Kirchhoff, Yannick
Kraemer, Lars
Rokuss, Maximilian
Ulrich, Constantin
Maier-Hein, Klaus H.
Computer Vision and Pattern Recognition
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.
title Scaling nnU-Net for CBCT Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2411.17213