NLP-based automated detection of coronary artery calcification on routine chest CT: unmet need for guideline-directed LDL reduction

Background Coronary artery disease (CAD) remains the leading cause of death worldwide, despite well-established efficacy of pharmacological low-density lipoprotein cholesterol (LDL-C) reduction regarding major adverse cardiovascular events (MACE) (1). As optimal prevention of MACE relies on timely secondary prevention, opportunities for broad and early CAD detection are critical. Chest computed tomography (CT), frequently performed for non-cardiac indications, often incidentally reveals coronary artery calcification (CAC), which is strongly associated with increased cardiovascular risk (2). Purpose To evaluate automated identification of incidentally detected CAC on routine chest CT using NLP (Natural language Processing), and to assess subsequent initiation and adequacy of guideline-directed LDL-C–lowering therapy in a real-world inpatient population. Methods In this retrospective, single-centre observational study, all consecutive chest CT examinations performed over six months (September 2025 to February 2026) were analyzed. Contrast-enhanced (n=1,459) and non-contrast (n=1,489) thoracic CT scans were included, whereas dedicated cardiac CT studies were excluded. Radiology reports were analyzed using NLP to identify CAC terminology. CT findings were linked to serum LDL-C levels measured within six months prior to the CT scan, as well as current lipid-lowering medication. Results In total, 2,948 chest CT examinations were analyzed. Median patient age was 71 [IQR 61-79] years, and 43.3% were female. CAC was identified in 62% of all cases. Among all patients with CAC, only 45% (n=823) were receiving LDL-C–lowering therapy (Figure 1). Current LDL-C levels were available in 39% demonstrating a median LDL-C level of 2.18 mmol/l [IQR 1.45-2.79] (84.4 mg/dl) among patients with coronary calcification and 2.38 mmol/l [IQR 1.68-3.05] (92.1 mg/dl) in patients without CAC on CT. Among patients with CAC and available LDL-C data, 52% received LDL-lowering medication, compared with 40% of patients without a recent LDL-C measurement (Figure 2). Notably, 76% of patients with known LDL-C levels and reported CAC were not at the guideline-recommended LDL-C target (<1.4 mmol/L). Suboptimal lipid management was particularly pronounced among patients <60 years with CAC, of whom only 6% achieved recommended LDL-C targets. Conclusions Although CAC is frequently detected on chest CT, LDL-C lowering therapy is often insufficient, and guideline-recommended targets are not achieved. These findings highlight the unmet need to systematically capture incidental coronary artery calcification (CAC) in routine chest CT reports, as this represents an important opportunity to reduce the risk of major adverse cardiovascular events.