Oxford team review: pericoronary fat imaging turns coronary inflammation into a quantifiable metric, with FAI Score falling after lipid-lowering, biologics and radiotherapy

The review presents PCAT as a "thermometer" of coronary inflammation: after the vessel wall releases IL-6, TNF-α and lipid peroxidation products, adjacent adipocytes become smaller with less lipid and more water, creating a radial gradient from the vessel wall that appears on CCTA as attenuation shifting toward the less negative end of the adipose range (approximately −190 to −30 HU). It integrates previously scattered anatomical, paracrine-signalling and imaging observations into a single chain from molecular mechanism to a CT-measurable gradient, while stating explicitly that CT attenuation is an indirect surrogate influenced by technical, anatomical and acquisition-related factors. A narrative synthesis of existing translational and mechanistic literature with no new primary data; the authors describe the biological link as plausible while noting attenuation metrics are an indirect surrogate.

The FAI Score corrects FAI values for technical, biological and anatomical factors and is interpreted using age- and sex-specific nomograms, converting raw FAI into a metric comparable across scanner vendors and scanning parameters; the review states that elevated FAI Scores independently predict MACE even after adjusting for high-risk plaque features, coronary calcium score and clinical risk factors. Relative to unadjusted mean attenuation measurements, the FAI Score aims to be a reproducible, platform-independent biomarker of coronary inflammation that supports cross-centre comparison and serial follow-up. The review cites large cohorts including CRISP-CT and ORFAN for external validation, with ORFAN described as including over 100,000 CCTA scans (to reach 250,000) linked to long-term clinical outcomes; the authors nonetheless state that its role in routine clinical practice remains to be fully established.

The review compiles evidence of therapeutic modulation of FAI: in 52 statin-naïve patients with non-obstructive CAD, long-term high-dose statin therapy significantly reduced FAI Score in the LAD at both approximately 1-year and ≥3-year follow-up alongside decreased non-calcified and increased calcified plaque volume; 15 weeks of orticumab significantly reduced FAI in a randomised, double-blind, placebo-controlled psoriasis trial; one year of anti-TNF-α/IL-12/23/IL-17 biologic therapy in 134 psoriasis patients significantly lowered FAI versus matched controls; and in breast cancer patients both baseline FAI Score and AI-predicted inflammatory risk declined two years after thoracic radiotherapy. It advances FAI from a purely prognostic marker toward a dynamic indicator that may monitor anti-inflammatory treatment response across lipid-lowering, immune-modulating, anti-oxLDL and radiotherapy strategies. Evidence comes from several small prospective or observational studies; the authors explicitly classify PCAT effects of colchicine, canakinumab, SGLT2 inhibitors and GLP-1 receptor agonists as exploratory with imaging confirmation pending.

The review describes how AI enables automated PCAT and coronary structure segmentation through U-Net and its three-dimensional variants, attention mechanisms, transformer modules, state-space model components and self-supervised pretraining, and how plaque quantification and FAI can be merged with Framingham, CAD-RADS or ASCVD scores into multi-parametric prognostic models; FAI, the Fat Radiomic Profile (FRP) and radiotranscriptomic signatures such as C19-RS capture dynamic inflammation, long-term structural remodelling and cytokine activity respectively. It positions AI as a bridge between imaging and molecular mechanism, so that routine CCTA could yield both dynamic and long-term dimensions of inflammation rather than average attenuation alone. The review cites deep-learning segmentation studies reporting interclass correlation coefficients above 0.9 for various plaque types and notes commercial platforms such as CaRi-Heart that integrate fully automated PCAT segmentation; it also notes that most studies rely on relatively small, homogeneous cohorts and that black-box explainability and limited annotated data remain constraints.

This is a review aimed at clinical and imaging readers, and its conclusions apply to people who have already undergone CCTA and in whom coronary inflammatory burden needs assessment, particularly those without obstructive CAD but with residual risk. For clinicians, it suggests considering PCAT as an additional dimension when reporting CCTA and weighing the downstream treatment implications; for researchers, it sketches a roadmap using large registries such as ORFAN as validation platforms and AI-based automated segmentation and multi-parametric models as tools, with prospective registry studies, health-economic modelling and guideline incorporation as necessary steps toward clinical adoption.

Several points warrant attention: CT attenuation is an indirect surrogate for inflammation and is affected by tube voltage, scanner type and reconstruction kernel as well as by calcified or fibrotic plaque; details of the FAI adjustments are proprietary and not always fully accessible; deep-learning models have limited explainability and large, well-annotated PCAT segmentation datasets are lacking, with most studies relying on relatively small, homogeneous cohorts. The effects of colchicine, canakinumab, SGLT2 inhibitors and GLP-1 receptor agonists on PCAT remain exploratory, with direct FAI evidence still pending. In addition, although this is a full-text review, the content of Figures 1 and 2 is available only as captions in the text, so the specific imaging details cannot be fully assessed from the written material.