CT-Based Radiomics Enhance Respiratory Function Analysis for Lung SBRT.

Area of Science:

• Pulmonary Medicine
• Radiology
• Medical Imaging Analysis

Background:

• Radiomics offers non-invasive quantitative imaging features for clinical applications.
• Quantifying lung function via radiomics can aid in managing pulmonary lesions.
• This study focuses on predicting diffusing capacity of lungs to carbon monoxide (DLCO) using radiomic features.

Purpose of the Study:

• To evaluate the capability of radiomic features in predicting pulmonary function parameters, particularly DLCO.
• To assess the predictive performance of clinical, radiomic, and combined clinical-radiomic models for DLCO abnormality.
• To explore the potential of radiomics in predicting lung function changes after Stereotactic Body Radiation Therapy (SBRT).

Main Methods:

• Retrospective analysis of 98 patients undergoing SBRT for lung cancer or nodules.
• Extraction of radiomic features from 4D CT scans of lung parenchyma, excluding lesions.
• Development of three models (clinical, radiomic, clinical-radiomic) using different CT phases to predict DLCO abnormality, assessed by AUC.

Main Results:

• The Charlson Comorbidity Index (CCI) was the most significant clinical predictor of baseline DLCO.
• Clinical-radiomic models outperformed individual clinical or radiomic models in predicting both baseline and post-treatment DLCO abnormality.
• Clinical-radiomic models achieved AUCs of 0.72-0.75 for baseline DLCO prediction and 0.91-0.95 for post-treatment DLCO prediction.

Conclusions:

• Radiomic features from healthy lung parenchyma on 4D CT scans correlate with baseline pulmonary function.
• Radiomics provides valuable information for surrogate models of lung function assessment.
• Preliminary findings suggest radiomics models can predict post-SBRT lung function, requiring further validation in larger prospective studies.