AI-enabled body composition biomarkers at post-mortem CT for enriching autopsy: analysis of a large decedent cohort

Abstract

Objective

To correlate fully-automated PMCT-based body composition measures with causes of death and comorbidities.

Materials and methods

Retrospective study of New Mexico Decedent Image Database (NMDID) with non-contrast PMCT scans between 2010 and 2017. Automated pipeline of AI-driven algorithms for quantifying skeletal muscle, subcutaneous/visceral fat, and aortic calcification from the abdominal component of PMCT scans was used. Scans with more than minimal decomposition were excluded. Cause of death was categorized as “acute” or “chronic.” A predetermined model derived CT-based “biological age.”

Results

6638 decedents (mean age, 50±18 [SD]; 74% male) comprised the final cohort. 80% of deaths were classified as “acute,” 10% as “chronic,” and 10% “uncertain.” Muscle density (HU) and area at the L3 lumbar level were higher in the “acute” versus “chronic” group (26 HU vs. 18 HU, p < 0.001; 192 cm² vs. 183 cm², p < 0.001). Muscle density and area at the L3 level were higher among those without cancer (25 HU vs. 16 HU, p < 0.001; 190 cm² vs. 169 cm², p < 0.01). Aortic Agatston scores were higher in those who died of heart disease (5120 vs. 2098, p < 0.001). Diabetic patients had higher L3 visceral fat area (227 cm² vs. 175 cm², p < 0.001) and lower muscle density (17 HU vs. 25 HU, p < 0.001). The deviation between chronological and biological age was significantly higher in the chronic versus acute group (median age deviation, 19 years vs. 10 years; p < 0.001).

Conclusion

Fully-automated quantitative PMCT-based tissue biomarkers correlate with the temporal nature of death and chronic co-morbidities, supporting their use for enhancing autopsies.

Graphical abstract