Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Domain-Specialized Interactive Segmentation Framework for Meningioma Radiotherapy Planning

Junhyeok Lee, Han Jang, Kyu Sung Choi

TL;DR

Meningioma radiotherapy planning requires precise and reproducible tumor delineation, which is hindered by manual segmentation variability and tumor heterogeneity. The authors present Interactive-MEN-RT, a domain-specific interactive segmentation framework built on nnU-Net V2 and nnInteractive that supports 3D prompt-based refinement (points, boxes, lassos, scribbles) and encodes prompts as foreground/background channels to guide segmentation. Through task-specific fine-tuning and clinician-guided iterative refinement, Interactive-MEN-RT achieves strong performance on the BraTS 2025 Meningioma RT dataset, with $DSC$ up to 77.6% and $IoU$ up to 64.6% for certain prompts, outperforming general-purpose baselines. The work demonstrates the value of disease-specific IMIS for critical RT planning by reducing interobserver variability and enhancing efficiency, with publicly available code to facilitate adoption and further validation.

Abstract

Precise delineation of meningiomas is crucial for effective radiotherapy (RT) planning, directly influencing treatment efficacy and preservation of adjacent healthy tissues. While automated deep learning approaches have demonstrated considerable potential, achieving consistently accurate clinical segmentation remains challenging due to tumor heterogeneity. Interactive Medical Image Segmentation (IMIS) addresses this challenge by integrating advanced AI techniques with clinical input. However, generic segmentation tools, despite widespread applicability, often lack the specificity required for clinically critical and disease-specific tasks like meningioma RT planning. To overcome these limitations, we introduce Interactive-MEN-RT, a dedicated IMIS tool specifically developed for clinician-assisted 3D meningioma segmentation in RT workflows. The system incorporates multiple clinically relevant interaction methods, including point annotations, bounding boxes, lasso tools, and scribbles, enhancing usability and clinical precision. In our evaluation involving 500 contrast-enhanced T1-weighted MRI scans from the BraTS 2025 Meningioma RT Segmentation Challenge, Interactive-MEN-RT demonstrated substantial improvement compared to other segmentation methods, achieving Dice similarity coefficients of up to 77.6\% and Intersection over Union scores of 64.8\%. These results emphasize the need for clinically tailored segmentation solutions in critical applications such as meningioma RT planning. The code is publicly available at: https://github.com/snuh-rad-aicon/Interactive-MEN-RT

Domain-Specialized Interactive Segmentation Framework for Meningioma Radiotherapy Planning

TL;DR

Meningioma radiotherapy planning requires precise and reproducible tumor delineation, which is hindered by manual segmentation variability and tumor heterogeneity. The authors present Interactive-MEN-RT, a domain-specific interactive segmentation framework built on nnU-Net V2 and nnInteractive that supports 3D prompt-based refinement (points, boxes, lassos, scribbles) and encodes prompts as foreground/background channels to guide segmentation. Through task-specific fine-tuning and clinician-guided iterative refinement, Interactive-MEN-RT achieves strong performance on the BraTS 2025 Meningioma RT dataset, with up to 77.6% and up to 64.6% for certain prompts, outperforming general-purpose baselines. The work demonstrates the value of disease-specific IMIS for critical RT planning by reducing interobserver variability and enhancing efficiency, with publicly available code to facilitate adoption and further validation.

Abstract

Precise delineation of meningiomas is crucial for effective radiotherapy (RT) planning, directly influencing treatment efficacy and preservation of adjacent healthy tissues. While automated deep learning approaches have demonstrated considerable potential, achieving consistently accurate clinical segmentation remains challenging due to tumor heterogeneity. Interactive Medical Image Segmentation (IMIS) addresses this challenge by integrating advanced AI techniques with clinical input. However, generic segmentation tools, despite widespread applicability, often lack the specificity required for clinically critical and disease-specific tasks like meningioma RT planning. To overcome these limitations, we introduce Interactive-MEN-RT, a dedicated IMIS tool specifically developed for clinician-assisted 3D meningioma segmentation in RT workflows. The system incorporates multiple clinically relevant interaction methods, including point annotations, bounding boxes, lasso tools, and scribbles, enhancing usability and clinical precision. In our evaluation involving 500 contrast-enhanced T1-weighted MRI scans from the BraTS 2025 Meningioma RT Segmentation Challenge, Interactive-MEN-RT demonstrated substantial improvement compared to other segmentation methods, achieving Dice similarity coefficients of up to 77.6\% and Intersection over Union scores of 64.8\%. These results emphasize the need for clinically tailored segmentation solutions in critical applications such as meningioma RT planning. The code is publicly available at: https://github.com/snuh-rad-aicon/Interactive-MEN-RT

Paper Structure

This paper contains 19 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Data Preprocessing and Preparation
  4. Network Architecture
  5. Interaction Prompts
  6. User Interaction Types
  7. Training-time Simulation of Prompts
  8. Prompt Encoding
  9. Ablation Study: Transfer Learning
  10. Implementation Details
  11. Evaluation Metrics and Baselines
  12. Results
  13. Qualitative Assessment by Prompt Type
  14. Interactive Segmentation Performance
  15. Comparative Qualitative Analysis of Segmentation Methods
  16. ...and 4 more sections

Figures (4)

  • Figure 1: Representative meningioma cases illustrating anatomical diversity. (a) Convexity meningioma with clear margins; (b) Skull base meningioma encasing structures.
  • Figure 2: An overview of the Interactive-MEN-RT for meningioma gross tumor volume segmentation with interactive prompts.
  • Figure 3: Qualitative segmentation overlays for each method and the ground truth under point prompt interaction settings.
  • Figure 4: Qualitative 3D segmentations for each method and the ground truth.