@article{QIMS154957,
author = {Min Wang and Jingyi Zhang and Rong Zhu and Jun Gu and Li Fan and Qizhi Chen},
title = {Quantifying the impact of slice thickness on cardiovascular risk stratification in lung cancer screening: a multi-center “RESCUE” study},
journal = {Quantitative Imaging in Medicine and Surgery},
volume = {16},
number = {7},
year = {2026},
keywords = {},
abstract = {Background: Patients undergoing routine non-gated chest computed tomography (CT) for health checkups or atypical chest discomfort often present with a coronary artery calcium (CAC) score of zero on standard 5.0 mm reconstructions. We hypothesized that these thick slices obscure mild calcification due to partial volume effects (PVEs), which could be recovered by retrospective analysis of native thin-slice images. This study aimed to quantify the rate of unrecognized coronary calcification on standard thick-slice CT by comparing paired thin- and thick-slice reconstructions.Methods: We analyzed data of 2,914 patients across four datasets: Stanford Coronary Calcium and chest CT’s (COCA) (n=651) for reference validation; an internal cohort evaluated by invasive angiography for early-onset coronary artery disease (CAD) (n=766) and National Lung Screening Trial (NLST) (n=852) with paired thin (1.0–2.0 mm) vs. standard (5.0 mm) scans; and TotalSegmentator (n=645) for robustness. A validated deep learning algorithm quantified CAC. The primary outcome was the “RESCUE” rate: patients reclassified from CAC =0 on 5.0 mm to CAC >0 on thinner slices.Results: In the internal cohort, 19.0% were reclassified from CAC =0 on 5.0 mm scans to CAC >0 on 1.0 mm scans. Similarly, 10.2% of NLST participants were reclassified using 2.0 mm scans. Most reclassified patients (91–99%) fell into the mild risk category (Agatston 1–99). Crucially, 31% of symptomatic patients with CAC =0 on standard scans had obstructive CAD (>50% stenosis); many were “rescued” to a positive CAC status by thin-slice analysis. Risk categorization showed strong agreement (weighted kappa 0.705–0.816). Artificial intelligence (AI) correlated strongly with expert annotations (r=0.956).Conclusions: Standard 5 mm reconstructions cause significant false negative CAC assessments. Analyzing routinely available thin slice reconstructions improves sensitivity for early subclinical atherosclerosis without additional radiation, supporting their use in opportunistic screening.},
issn = {2223-4306}, url = {https://qims.amegroups.org/article/view/154957}
}