Both chronological age and individual differences in aging are the two indispensable components for predicting...

Area of Science:

• Biomedical Engineering
• Gerontology
• Artificial Intelligence

Background:

• Chronological age (CA) is often used as a proxy for biological age (BA) in AI models, but fails to capture individual aging variations.
• This limitation leads to the age delta correlation (ADC) phenomenon, where predicted BA is biased by CA.

Purpose of the Study:

• To develop a novel method for predicting vascular biological age (vBA) that incorporates individual aging heterogeneity.
• To introduce a new index, relative individual risk difference (RIRD), to quantify individual aging variations.

Main Methods:

• Utilized arteriosclerosis detection data for vBA prediction.
• Developed a supervised regression model using a selected healthy population group.
• Introduced RIRD to quantify individual aging differences and correct vBA predictions.

Main Results:

• The corrected vBA demonstrated comparable performance to the Klemera-Doubal method (KDM) in identifying non-healthy individuals.
• The corrected vBA significantly reduced the age delta correlation (ADC) phenomenon, unlike uncorrected BA.
• The study confirmed that both CA and individual aging differences are crucial for accurate BA prediction.

Conclusions:

• Incorporating individual aging heterogeneity is essential for developing robust BA prediction models.
• The proposed RIRD-based method offers a more accurate and less biased approach to vBA prediction.
• Future AI models for BA should integrate individual aging variations alongside chronological age.