AI for Multistructure Incidental Findings and Mortality Prediction at Chest CT in Lung Cancer Screening.

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Incidental extrapulmonary findings on chest CT scans are common and clinically significant.
• Automated analysis can improve the detection and prediction of these findings.

Purpose of the Study:

• To integrate AI for segmenting multiple structures, coronary artery calcium (CAC), and epicardial adipose tissue.
• To use automated feature extraction and machine learning to detect extrapulmonary abnormalities and predict all-cause mortality (ACM).

Main Methods:

• Post hoc analysis of chest CT scans from the National Lung Screening Trial (NLST).
• AI-based segmentation of 32 structures and quantification of 15 radiomic features per structure.
• Ensemble AI models used to predict 2-year and 10-year ACM and extrapulmonary significant incidental findings.

Main Results:

• 16% of participants had reported extrapulmonary significant incidental findings.
• 14% of participants died within 10 years.
• Coronary artery calcium (CAC) showed the highest feature importance for all prediction tasks.
• The 10-year ACM model achieved the best AUC performance (0.72).

Conclusions:

• A fully automated AI model can identify at-risk extrapulmonary structures on chest CT scans.
• AI models can predict all-cause mortality with interpretable risk scores.
• This approach enhances the clinical utility of chest CT scans for incidental findings and mortality prediction.