Radiomic Analysis of CT Predicts Tumor Response in Human Lung Cancer with Radiotherapy
Mengmeng Yan1,2, Weidong Wang3,4
1Urban Vocational College of Sichuan, Chengdu, China.
Journal of Digital Imaging
|October 7, 2020
Summary
Machine learning using radiomic features from CT scans can predict lung cancer treatment response. Two radiomic features, flatness and coefficient of variation, show promise as biomarkers.
Area of Science:
- Medical imaging analysis
- Artificial intelligence in oncology
- Radiotherapy research
Background:
- Radiomics analysis offers potential for monitoring treatment response in lung cancer.
- Machine learning (ML) models can analyze complex imaging data.
Purpose of the Study:
- To evaluate ML models for predicting tumor response in lung cancer patients undergoing radiotherapy.
- To identify radiomic features that serve as biomarkers for treatment outcome.
Main Methods:
- Retrospective analysis of CT images from 100 lung cancer patients treated with radiotherapy.
- Radiomics feature extraction and selection.
- Training and testing a support vector machine (SVM) classifier.
Main Results:
- A SVM classifier utilizing two radiomic features (flatness and coefficient of variation) achieved an AUC of 0.91 on the test set.
- These features were derived from the lung tumor's volume of interest.
Conclusions:
- Flatness and coefficient of variation are potential imaging biomarkers for predicting lung cancer treatment response.
- ML-based radiomics analysis of CT images can aid in personalized radiotherapy treatment planning.
