PUNCH: Physics-informed uncertainty-aware network for coronary hemodynamics

More than 10 million coronary angiograms are performed globally each year, providing a gold standard for detecting obstructive coronary artery disease. Yet, no obstructive lesions are identified in 70% of patients evaluated for ischemic heart disease. Up to half of these patients have undiagnosed, life-limiting coronary microvascular dysfunction (CMD), which remains under-detected due to the limited availability of invasive wire-based tools (Doppler-wire or thermodilution) used to measure coronary flow reserve (CFR). Here, we introduce PUNCH, a wire-free, uncertainty-aware framework for estimating CFR from paired resting and hyperemic coronary angiographic acquisitions. PUNCH integrates physics-informed neural networks with variational inference to infer coronary blood flow from first-principles models of contrast transport, without requiring ground-truth flow measurements or population-level training. The pipeline runs in approximately three minutes per patient on a single GPU. Evaluated on 1000 synthetic kymographs and a feasibility-scale, single-center cohort of 20 patients with matched invasive bolus thermodilution CFR, PUNCH produces CFR point estimates that correlate strongly with the invasive reference and uncertainty intervals that widen under image degradation; the raw posterior intervals are however under-dispersed relative to nominal coverage, and we discuss the recalibration step that would be required before threshold-based clinical use. As a proof-of-concept feasibility study, this work illustrates how physics-informed inference could expand the physiological information extractable from existing angiographic imaging, pending validation in larger, multi-center, multi-territory cohorts.