Related Experiment Video
Updated: Jun 21, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Automated segmentation of brain metastases with deep learning: A multi-center, randomized crossover, multi-reader
Xiao Luo1,2, Yadi Yang1,2, Shaohan Yin1,2
1State Key Laboratory of Oncology in South China, Guangdong Provincial Clinical Research Center for Cancer, Sun Yat-sen University Cancer Center, Guang zhou, Guangdong Province, China.
Neuro-Oncology
|July 11, 2024
Summary
This study developed an AI system for brain metastasis segmentation. The system significantly improved accuracy and reduced delineation time for radiologists, showing clinical utility.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Clinical validation of AI for brain metastasis (BM) segmentation is limited.
- Developing and evaluating an AI system for accurate BM segmentation is crucial.
Purpose of the Study:
- To develop and clinically validate a deep-learning-based system for brain metastasis segmentation (BMSS).
- To assess the performance of the BMSS in improving segmentation accuracy and efficiency.
Main Methods:
- Developed a deep-learning BM segmentation system (BMSS) using 10,338 BM from 488 patients.
- Conducted a multi-reader study with 50 prospective cases across 5 centers.
- Compared radiologist performance (residents and attendings) in assisted vs. unassisted segmentation modes.
Main Results:
- BMSS achieved a median Dice Similarity Coefficient (DSC) of 0.91.
- BMSS assistance improved median DSC from 0.87 to 0.92 (P < .001) and reduced contouring time by 42%.
- Resident radiologists showed greater accuracy improvement with BMSS assistance compared to attending radiologists.
Conclusions:
- The developed BMSS significantly enhances the efficiency and accuracy of brain metastasis delineation.
- The AI system demonstrates optimal application for improving clinical workflow in neuro-oncology.
