The abstract is read at the publisher; the summary is JClub's.
A fully automated CT pipeline (sarco) robustly measures L3 skeletal muscle area with high agreement (ICC 0.983; bias +1.35 cm2) despite varied scan ranges, slice order, and contrast phase.
- Why it matters: Manual body-composition quantification is labor-intensive and can fail when CT protocols vary, limiting scalable prognostic use across routine scans.
- What they did: sarco detects orientation, localizes L1/L3 with out-of-range rejection, segments with U-Net, estimates four contrast phases, converts to non-contrast-equivalent values, and outputs muscle/marrow age; validated on 4986 series and prognostic testing in 2006 post–curative cancer surgery patients.
- The result: L3 slice selection on truncated scans was off by ≥0.5 vertebra in 4.7–5.6% (TotalSegmentator 11.3–14.3%); phase accuracy was 0.949 and attenuation RMSE improved 7.96→2.81 HU, while low muscle mass increased Uno’s C by 0.034 (cancer-specific survival) and 0.031 (overall survival).
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.