Cardiac CT Radiomic Signature (AI Feature Set) Calculator

Clinical calculation support · Verify results before clinical use

Enter measured coronary perivascular fat attenuation values and, when available, the exact coefficients from a published radiomic model. FAI classification and custom-model probability are reported separately. A model probability is only meaningful when the original study's population, endpoint, feature definitions, transformations, and calibration are reproduced.

Imaging and model inputs

Use the measurement units and model specification from the original study.

CCTA · HU

1. Perivascular Fat Attenuation Index (FAI)

The default −70 HU threshold is configurable and should not be assumed to be a universal vessel-independent cutoff.

2. Optional plaque and calcium context

Displayed for context only; these values do not alter FAI classification or enter the custom model automatically.

3. Custom radiomic logistic model

Enter the intercept and each complete feature value/coefficient pair from one publication. Leave both fields blank for unused features. At least one complete pair is needed to calculate a model probability.

Feature 1

Feature 2

Feature 3

Feature 4

Clinical context and limitations

Radiomics converts imaging intensity, texture, and spatial patterns into quantitative features. In coronary CT angiography, perivascular adipose tissue attenuation and multifeature plaque/PVAT signatures have been investigated as markers associated with vascular inflammation, plaque phenotype, disease progression, and cardiovascular outcomes.

FAI classification

FAI reflects attenuation characteristics of adipose tissue around a coronary artery. Higher (less negative) values may reflect inflammation-related PVAT changes. This calculator applies the user-selected threshold to both arteries as requested by the original implementation; it does not establish that the same threshold is validated for both vessels.

Custom logistic model

LP = β0 + Σ(βi × xi), and p = 1 / (1 + e−LP). Use the original coefficients and feature values after applying the same transformations, units, normalization, and missing-data rules described in the source publication.

  • FAI depends on acquisition, reconstruction, segmentation, software, and vessel-specific methodology.
  • The −70 HU value is a configurable reference, not a universal cutoff for all vessels and populations.
  • Custom-model probability is not clinically validated by this calculator.
  • CACS and HRP count are contextual only and are not entered into the logistic model.

References

Oikonomou EK, et al.Non-invasive detection of coronary inflammation using computed tomography and the perivascular fat attenuation index. The Lancet. 2018. DOI: 10.1016/S0140-6736(18)31114-0; PMID: 30170852. Verify vessel-specific thresholds and methodology in the original publication.
Oikonomou EK, et al.A novel machine learning-derived radiotranscriptomic signature of perivascular fat improves cardiac risk prediction using coronary CT angiography. European Heart Journal. 2019;40:3529–3543. DOI: 10.1093/eurheartj/ehz592; PMID: 31504423.
Model-specific publicationNot identified in the supplied source. Verify the full citation, coefficients, feature transformations, endpoint, and calibration before clinical use.
Clinical disclaimer: This educational calculator does not diagnose coronary inflammation or provide an independently validated estimate of individual clinical risk. Interpret results with the original study methodology, CCTA findings, clinical history, and established risk assessment. Do not use a custom logistic output for clinical decisions unless the complete source model has been verified.