Resting angiography encodes hyperaemic coronary physiology: component-level prospective validation

Background

The index of microcirculatory resistance (IMR) combines hyperaemic distal coronary pressure with thermodilution transit time. Both components require invasive pressure wire instrumentation and pharmacological hyperaemia. We assessed whether resting angiographic measurements predict each component of hyperaemic coronary physiology independently, and validated these predictions prospectively.

Methods

We studied 300 patients with unobstructed coronary arteries undergoing invasive coronary physiology at a regional cardiac centre. A multi-layer perceptron regressor predicted hyperaemic distal coronary pressure from resting aortic pressure and fractional flow reserve (FFR). A linear model predicted resting coronary transit time from TIMI frame count, and a second multi-layer perceptron predicted hyperaemic transit time from the predicted resting transit time and TIMI frame count, modelling the non-linear resting-to-hyperaemic transformation. Development performance was assessed by stratified 10-fold cross-validation. The model was then applied to a distinct group of 119 prospective patients undergoing thermodilution. TIMI frame count and thermodilution measurements were obtained from the same coronary artery in all prospective patients.

Results

In the development cohort, TIMI frame count correlated with thermodilution resting transit time (r = 0.53, p < 0.001) and hyperaemic transit time (r = 0.62, p < 0.001). Cross-validated predictions of hyperaemic distal coronary pressure (r = 0.75, mean absolute error [MAE] 6.7 mmHg), hyperaemic transit time (r = 0.61, MAE 0.10 s), and resting transit time (r = 0.52, MAE 0.32 s) were obtained. In the prospective cohort, distal coronary pressure prediction was preserved (r = 0.62, MAE 8.7 mmHg) and resting transit time prediction exceeded development performance (r = 0.59, MAE 0.39 s).

Hyperaemic transit time prediction was attenuated but remained significant (r = 0.39, p < 0.001, MAE 0.16 s), with a corresponding TIMI-to-thermodilution correlation of r = 0.39. All three components showed minimal bias (pressure −2.1 mmHg; resting transit time +0.04 s; hyperaemic transit time −0.01 s) (Figure).

Conclusions

Resting aortic haemodynamics and angiographic contrast transit predict the individual components of invasive hyperaemic coronary physiology. Pressure prediction generalised well to a prospective cohort; transit time prediction was attenuated, reflecting the greater stochastic variability of thermodilution-derived transit measurements relative to deterministic angiographic frame counts. These findings establish the physiological basis for wire-free estimation of coronary microvascular resistance without reliance upon computational fluid dynamics principles.Figure 1