Accelerated Aging and Aging Velocity from Deep Learning-based Chest Radiograph-derived Age for Predicting...

Area of Science:

• Radiology
• Artificial Intelligence
• Gerontology

Background:

• Chronological age is a primary risk factor for mortality.
• Assessing biological age through medical imaging offers potential prognostic insights.
• Deep learning models can estimate biological age from radiographic data.

Purpose of the Study:

• To evaluate the prognostic capability of deep learning-derived radiographic age and aging velocity for predicting mortality in an Asian population.
• To investigate the association between accelerated aging and mortality risk.
• To determine if aging velocity predicts mortality independently of baseline characteristics.

Main Methods:

• Retrospective cohort study of 421,894 Korean adults with chest radiography data (2006-2020).
• Radiographic age estimated using the AgeNet deep learning model.
• Accelerated aging defined as radiographic age ≥ 5 years older than chronological age.
• Aging velocity calculated from serial radiographs.
• Cox and Fine-Gray models used for mortality prediction.

Main Results:

• Accelerated aging was significantly associated with increased all-cause and cause-specific mortality (HRs 1.26 for males, 1.52 for females).
• Aging velocity independently predicted mortality (cumulative mortality ratio per 1-SD increase: 1.24 for males, 1.35 for females).
• Accelerated aging velocity (≥ 1.5 years/year) significantly increased mortality risk (MRRs 1.51 for males, 1.71 for females).

Conclusions:

• Deep learning-derived radiographic age and aging velocity are independent predictors of all-cause and cause-specific mortality.
• Accelerated aging and faster aging velocity indicate higher mortality risk.
• Findings highlight the potential of radiographic age as a prognostic biomarker.