Arm in arm: how in silico modelling and AI tackle the “grey zone” in coronary diagnostics

During the latest Avicenna Alliance webinar, our team presented how combining CCTA, in silico patient‑specific modelling, and AI may help address one of the most discussed areas in coronary diagnostics – the functional “grey zone” between clearly significant and clearly non‑significant lesions. 

In the presentation, CSO Ziemowit Ostrowski and co‑founder Andrzej Kosior described an approach in which CCTA is treated not as an endpoint, but as a data‑rich starting point for broader physiological assessment. CCTA offers fast, non‑invasive, outpatient diagnostics with first‑line guideline status and provides detailed anatomical information that remains partly underused in everyday clinical practice. In this context, CCTA becomes a natural entry point for in silico workflows that aim to transform static anatomy into dynamic physiology, to derivemodel‑based metrics that may support therapeutic decision‑making. 

A key concept discussed was a “unique Hemolens triple engine” architecture that combines AI, biomedical engineering (BME), and computational fluid dynamics (CFD). In this framework, AI is used to accelerate precise coronary artery segmentation. Biomedical engineering modules are used to derive patient‑specific boundary conditions, including microcirculatory resistance models, while CFD is applied to simulate blood flow (velocity fields and pressure drops) along the coronary tree. Together, these components form an integrated workflow that can convert a single CCTA into continuous FFR‑like profiles, pressure maps, and diameter curves, effectively generating a functional map of the coronary circulation rather than a single FFR value. 

The speakers also outlined an approach to boundary‑condition modelling in non‑invasive FFR‑CT. In contrast to methods that rely mainly on geometric assumptions or fixed pressure inputs, the presented CNBP‑based method uses non‑invasive peripheral blood pressure waveforms recorded with a certified device. These signals are transformed into aortic waveforms using an in‑house transfer function designed to preserve both shape and amplitude and then used as inputs for the FFR-CT calculation and lumped‑parameter circulation model. Internal analyses suggest that such patient‑specific boundary conditions may be particularly relevant in the “grey zone” (FFR range 0.75–0.85), where clinical decisions are most challenging. 

Preliminary internal data indicate that, when CNBP‑based boundary conditions are applied, FFR‑CT diagnostic performance appears to remain within clinically relevant ranges in the 0.7–0.9 interval (still under evaluation). These observations point to a potential role for AI‑supported, physiology‑grounded modelling in refining clinical decision‑making in the most equivocal range of coronary stenosis severity. 

Beyond FFR‑CT, the presentation briefly addressed the diagnostic and prognostic value of CCTA‑based plaque assessment. Special attention should be given to high‑risk plaque that, according to the most recent publications, has been associated with myocardial infarction even in the absence of severe luminal narrowing. Future integration of plaque‑level information into in silico models could, in principle, contribute to improved patient stratification, more tailored pharmacotherapy, and more informed procedure planning. The speakers contrasted current cath‑lab workflows – in which coronary anatomy is often delineated only after multiple contrast runs – with a possible future scenario in which CCTA‑derived 3D anatomy and simulated hemodynamics inform interventional strategy before the first injection. 

 Finally, the discussion linked these concepts with international verification and validation initiatives for in silico models, evolving regulatory expectations and practical workforce challenges in cardiology and imaging. In this setting, scalable, non‑invasive, AI‑supported diagnostic workflows are increasingly considered as potential components of future care pathways. At Hemolens, we view CCTA‑based in silico modelling as a promising foundation for end‑to‑end diagnostic and treatment‑planning pathways, designed to complement clinical judgement and potentially support more precise and accessible coronary care in the years ahead. 

Watch the full recording of our webinar here.