AI-based CT Body Composition Identifies Myosteatosis as Key Mortality Predictor in Asymptomatic Adults

Area of Science:

• Radiology
• Artificial Intelligence
• Public Health

Background:

• Body composition data is typically limited to diseased or elderly populations.
• The prognostic value of body composition in asymptomatic adults is not well understood.

Purpose of the Study:

• To utilize AI-driven body composition metrics from routine abdominal CT scans.
• To investigate the association between obesity, liver steatosis, myopenia, and myosteatosis with mortality risk in asymptomatic adults.

Main Methods:

• Retrospective study of asymptomatic outpatients undergoing colorectal cancer screening (April 2004 - December 2016).
• AI (U-Net algorithm) extracted body composition metrics (muscle area/density, fat area, liver density) from CT scans.
• Abnormal body composition defined by liver steatosis, obesity, myosteatosis, or myopenia.

Main Results:

• 86% of deceased patients exhibited abnormal body composition.
• Myosteatosis was present in 55% of deceased patients, with a 15.5% absolute 10-year risk.
• Myosteatosis, obesity, liver steatosis, and myopenia were all linked to increased mortality risk.
• After multivariable adjustment, myosteatosis remained a significant predictor of mortality (HR, 1.89).

Conclusions:

• AI-based body composition analysis from routine CT scans is feasible.
• Myosteatosis is identified as a crucial predictor of mortality in asymptomatic adults.
• This approach can enhance risk stratification and inform preventative strategies.