AG-STDF: Anatomy-Guided ECG Spatiotemporal Dynamical Fusion for Screening Coronary Microvascular Dysfunction.

Area of Science:

• Cardiology
• Medical Imaging
• Computational Biology

Background:

• Coronary slow flow (CSF) affects 1-7% of coronary angiograms, presenting a challenge in managing coronary microvascular dysfunction (CMD) despite normal epicardial arteries.
• Current diagnostic methods for CSF often involve invasive procedures, highlighting the need for non-invasive screening tools.

Purpose of the Study:

• To propose and evaluate an anatomy-guided spatiotemporal dynamical fusion (AG-STDF) framework for the non-invasive screening of CSF.
• To assess the capability of AG-STDF in performing vessel-level assessment for CMD management.

Main Methods:

• The AG-STDF framework integrates anatomical electrophysiological territory mapping with 12-lead electrocardiogram (ECG) dynamics.
• ECG signals are decomposed into coronary territories (LAD, LCX, RCA) and fused using joint recurrence quantification and multi-scale decomposition.

Main Results:

• AG-STDF achieved a high AUROC of 0.9467 ± 0.0272 for CSF detection.
• The framework demonstrated strong performance in predicting corrected thrombolysis in myocardial infarction frame counts (CTFC) (R² = 0.5951 ± 0.0666) at the vessel level.
• AG-STDF showed consistent performance in right-dominant coronary patients and processed a 10-second ECG segment in 19.37 seconds.

Conclusions:

• AG-STDF enables accurate, non-invasive CSF detection and vessel-level assessment, particularly in right-dominant populations.
• This framework serves as a cost-effective tool for CMD screening, potentially reducing the need for invasive testing and supporting early diagnosis.