Related Experiment Video
Updated: Sep 30, 2025
09:06
Whole-brain Segmentation and Change-point Analysis of Anatomical Brain MRI—Application in Premanifest Huntington's Disease
Published on: June 9, 2018
12.3K
Glioma survival prediction from whole-brain MRI without tumor segmentation using deep attention network: a
Zhi-Cheng Li1,2, Jing Yan3, Shenghai Zhang1
1Institute of Biomedical and Health Engineering, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China.
European Radiology
|March 12, 2022
Summary
DeepRisk, a novel deep learning model, predicts diffuse glioma survival using whole-brain MRI without tumor segmentation. It offers comparable accuracy to models using segmented images and provides independent prognostic value.
Area of Science:
- Neuro-oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Diffuse gliomas are challenging to treat, with overall survival prediction crucial for treatment planning.
- Accurate prognostic models are needed to improve patient outcomes and guide clinical decisions.
- Current methods often rely on tumor segmentation, which can be time-consuming and variable.
Purpose of the Study:
- To develop and validate a deep learning model, DeepRisk, for predicting overall survival in diffuse glioma patients.
- To assess the model's performance using whole-brain MRI without the need for tumor segmentation.
- To evaluate DeepRisk's prognostic value compared to existing clinical and molecular factors.
Main Methods:
- A multicenter retrospective study involving 1556 diffuse glioma patients.
- Development of two deep learning models: DeepRisk (whole-brain MRI) and ResNet (segmented tumor images).
- Rigorous validation on multiple external test datasets using C-index, integrated Brier score (IBS), and AUCs.
Main Results:
- DeepRisk accurately predicted overall survival and stratified patients into distinct risk subgroups.
- DeepRisk demonstrated comparable performance to the ResNet model, with C-indices ranging from 0.77 to 0.83.
- The DeepRisk score showed independent prognostic value and improved classification accuracy when combined with clinicomolecular factors.
Conclusions:
- DeepRisk effectively predicts diffuse glioma survival directly from whole-brain MRI, eliminating the need for tumor segmentation.
- The model achieves high accuracy and offers significant incremental prognostic value.
- DeepRisk represents a promising tool for enhancing prognostic assessment in neuro-oncology.
