Related Experiment Video
Updated: Oct 30, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.1K
A deep learning-based dual-omics prediction model for radiation pneumonitis.
Liang Bin1, Tian Yuan1, Su Zhaohui2
1Department of Radiation Oncology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China.
Medical Physics
|July 5, 2021
Summary
A new deep learning dual-omics model significantly improves radiation pneumonitis prediction in thoracic radiotherapy patients by integrating dose and ventilation imaging data. This approach offers superior accuracy over single-omics models.
Area of Science:
- Radiotherapy
- Medical Imaging
- Artificial Intelligence
Background:
- Radiation pneumonitis (RP) is a primary toxicity concern in thoracic radiotherapy.
- Predicting RP accurately is crucial for optimizing patient treatment and outcomes.
Purpose of the Study:
- To develop and validate a deep learning-based dual-omics model for enhanced RP prediction.
- To integrate original dose (OD) distribution and ventilation image (VI) data for improved predictive performance.
Main Methods:
- Utilized four-dimensional computed tomography (4DCT) to derive VI and OD.
- Employed a 3D Convolution (C3D) network for feature extraction from functional dose (FD), VI, and OD.
- Applied entropy-based methods for feature selection and binary classifiers for model construction.
- Validated models using cross-validation, bootstrap, and nested sampling.
Main Results:
- The dual-omics model achieved an Area Under Curve (AUC) of 0.874, outperforming single-omics models (VI AUC: 0.780, OD AUC: 0.810).
- Analysis of 4DCT-based VI revealed inhomogeneous pulmonary function in patients.
- Feature selection effectively removed singular features, enhancing model stability.
Conclusions:
- The proposed dual-omics model demonstrates superior performance in predicting RP compared to single-omics approaches.
- Integration of bimodality data (OD and VI) and in-depth data exploration contribute to improved prediction accuracy.
- This deep learning model offers a promising tool for personalized radiotherapy planning.
