Deep learning-quantified body composition from positron emission tomography/computed tomography and cardiovascular...

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Body Composition Assessment

Background:

• Cardiac Positron Emission Tomography (PET)/Computed Tomography (CT) Myocardial Perfusion Imaging (MPI) is crucial for diagnosing cardiometabolic syndrome.
• Low-dose CT scans in PET/CT MPI offer potential for body tissue composition analysis.
• Automated quantification of skeletal muscle (SM) and adipose tissue from these scans is an emerging area.

Purpose of the Study:

• To automatically quantify skeletal muscle (SM), bone, and adipose tissue (epicardial, subcutaneous, visceral, intermuscular) from cardiac PET/CT scans using deep learning.
• To evaluate the association of these quantified body tissues with the risk of death or myocardial infarction (MI).

Main Methods:

• Deep learning algorithms were employed to segment and quantify SM, bone, and various adipose tissue depots (EAT, SAT, VAT, IMAT) from PET MPI scans across three sites.
• Sex-specific thresholds for abnormal tissue values were established.
• Unadjusted and multivariable Cox regression models were used to assess the association between body composition and outcomes (death or MI).

Main Results:

• The study analyzed 10,085 patients, with automated body tissue segmentation completed rapidly (102 ± 4 seconds).
• Higher visceral adipose tissue (VAT) density was significantly associated with an increased risk of death or MI (adjusted Hazard Ratio [HR] 1.24).
• Similar associations were observed for intermuscular adipose tissue (IMAT), subcutaneous adipose tissue (SAT), and epicardial adipose tissue (EAT). Patients with elevated VAT density and reduced myocardial flow reserve showed a substantially higher risk (adjusted HR 2.49).

Conclusions:

• Volumetric body tissue composition can be rapidly and automatically derived from standard cardiac PET/CT scans.
• This automated analysis provides valuable, quantitative insights into sarcopenia and cardiometabolic health.
• The findings highlight the potential of integrating body composition analysis into routine cardiac imaging for improved patient risk stratification.