Related Experiment Video
Updated: May 15, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.6K
Head and Neck Tumor Segmentation on MRIs with Fast and Resource-Efficient Staged nnU-Nets.
Elias Tappeiner1, Christian Gapp1, Martin Welk1
1UMIT Tirol - Private University for Health Sciences and Health Technology, Eduard-Wallnöfer-Zentrum 1, Hall in Tirol 6060, Austria.
Head and Neck Tumor Segmentation for Mr-Guided Applications : First MICCAI Challenge, HNTS-MRG 2024, Held in Conjunction with MICCAI 2024, Marrakesh, Morocco, October 17, 2024, Proceedings
|April 10, 2025
Summary
This study introduces a fast, two-stage AI method for segmenting head and neck tumors in MRI scans, improving adaptive radiotherapy planning. The approach achieved competitive results, enhancing tumor segmentation accuracy for pre-RT and mid-RT planning.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiotherapy
Background:
- Conventional CT-based radiotherapy planning lacks soft tissue contrast compared to MRI.
- Manual tumor segmentation in head and neck MRI is time-consuming, hindering adaptive radiotherapy.
- The HNTS-MRG challenge aims to advance automatic tumor segmentation for MR-guided applications.
Purpose of the Study:
- To develop an efficient automatic tumor segmentation method for head and neck MRI.
- To improve pre-RT and mid-RT planning in MR-guided radiotherapy.
- To address the limitations of manual segmentation in head and neck cancer treatment.
Main Methods:
- Implemented a two-stage segmentation approach using the nnU-Net architecture with residual encoders.
- Utilized a novel two-stage strategy where initial segmentation results guide a refinement stage.
- Incorporated pre-RT plan information as input for the mid-RT segmentation refinement network.
Main Results:
- Achieved a Dice Coefficient of 80.97% for pre-RT segmentation using the first-stage nnU-Net.
- Significantly improved mid-RT segmentation performance by using the second-stage refinement network.
- Demonstrated competitive and enhanced segmentation accuracy for both pre-RT and mid-RT tasks.
Conclusions:
- The proposed two-stage nnU-Net approach offers a fast and resource-efficient solution for head and neck tumor segmentation.
- The method shows significant potential for advancing daily adaptive radiotherapy through improved MRI segmentation.
- Code and model weights are publicly available to facilitate further research and application.
