Optional — Plaque and Calcium Context (for display only)
Custom Radiomic Signature (from a publication) — Logistic Model Use when you have coefficients from an original paper (intercept and feature weights). Calculator will output linear predictor and probability.
Feature 1
Feature 2
Feature 3
Feature 4
Cardiac CT Radiomic Signature (AI Feature Set): Explanation and Clinical Context Radiomics translates imaging textures and intensities into quantitative features that can be combined into signatures predicting pathology or outcomes. In coronary CT angiography (CCTA), two commonly studied AI-driven phenotypes are perivascular adipose tissue (PVAT) inflammation quantified by the Fat Attenuation Index (FAI) and plaque or PVAT radiomic signatures derived from high-dimensional features. FAI is defined as the mean attenuation of voxels consistent with adipose tissue (approximately −190 to −30 HU) within a radial distance from the outer vessel wall equal to the vessel’s diameter; greater (less negative) values indicate tissue remodeling due to vascular inflammation. Thresholds around −70 HU have frequently been used to flag high inflammatory burden, and higher FAI around the RCA and LAD has been associated with increased risk of cardiac death or non-fatal MI in longitudinal cohorts. Beyond single-metric FAI, several studies propose multifeature radiomic signatures of plaques or PVAT to predict vulnerable plaque, rapid plaque progression (RPP), ischemia, or major adverse cardiac events (MACE). Because original coefficients and intercepts differ by publication and endpoint, this calculator includes a custom logistic model section: when you enter the exact intercept and feature weights from a chosen paper along with your measured feature values, it returns the linear predictor and predicted probability, enabling faithful, study-specific risk estimation.
Clinical Significance Elevated FAI or a high radiomic signature has been independently associated with adverse outcomes in multiple cohorts, often providing incremental prognostic value over traditional factors and plaque metrics. Clinically, these AI features may help reclassify risk and prioritize preventive or anti-inflammatory strategies; however, absolute risk estimates are study-specific and should only be interpreted against the definitions, thresholds, and calibration reported in the source paper used to derive the model.
Clinical Interpretation Summary If either LAD or RCA FAI is above your chosen cut-off (e.g., −70 HU), classify as “high perivascular inflammation,” which indicates higher relative near-term risk; combine this with plaque features (e.g., napkin-ring sign, low-attenuation plaque, positive remodeling, spotty calcification) and clinical risk to guide management. For precise probabilities from multifeature radiomic signatures, use the custom section with the original coefficients and intercept from the publication you follow.
References
Oikonomou EK, et al. The Lancet 2018: non-invasive detection of coronary inflammation using perivascular FAI.
Oikonomou EK, et al. Eur Heart J 2019: machine-learning radiotranscriptomic signature of PVAT improves risk prediction.
Coerkamp CF, et al. 2024: FAI methodology and risk reclassification on CCTA.
Chen Q, et al. 2023: CCTA-based plaque radiomics signature predicting rapid plaque progression.
Radiomics in Cardiac CT — comprehensive overview of plaque and PVAT radiomics.