An area-based imaging biomarker for characterizing coronary artery stenosis with myocardial BOLD MRI

Sotirios A. Tsaftaris, Richard Tang, Xiangzhi Zhou, Debiao Li, Rohan DharmakumarJournal of Cardiovascular Magnetic Resonance 2011 · DOI

Abstract: BOLD MRI may be used for detecting myocardial oxygenation changes secondary to coronary artery stenosis. However, current approaches for analyzing BOLD changes are suboptimal for detecting critical stenosis. This paper introduces ARREAS, a statistical framework that identifies stenosis by measuring the spatial extent of oxygen-deprived tissue rather than average signal intensity drops.

The Challenge of BOLD MRI Analysis

Myocardial Blood Oxygen Level Dependent (BOLD) MRI offers a non-invasive, contrast-free method for monitoring oxygenation changes in the heart. By utilizing deoxygenated hemoglobin as an endogenous contrast agent, clinicians can theoretically identify areas of ischemia. However, BOLD signals are inherently subtle and prone to noise.

The traditional "gold standard" for analysis involves averaging signal intensity across predefined segments of the heart (the AHA 17-segment model). This approach, known as Segment-Intensity-Response (SIR), often fails to detect critical stenosis. If a blockage only affects a portion of a segment, the averaging process "dilutes" the pathological signal with data from healthy tissue, masking the severity of the disease.

ARREAS: A Spatial Approach to Ischemia

To overcome the limitations of intensity averaging, researchers developed ARREAS (Area-based biomaRker for chaRactErizing coronAry Stenosis). Instead of asking "how much darker is this segment?", ARREAS asks "how large is the contiguous area of oxygen deprivation?"

The ARREAS workflow involves three key steps:

  1. Statistical Baselining: Pixel intensities from resting images are fitted to a location-scaled t-distribution to establish a healthy baseline.
  2. Affected-Fraction (AF): A "hypointense" region is defined as any contiguous cluster of pixels falling below a statistical threshold. The AF is the ratio of this area to the total myocardium.
  3. Ischemic-Extent (IE): This primary biomarker is the ratio of the Affected-Fraction during stress (induced by adenosine) to the Affected-Fraction at rest ($IE = AF_{STRESS} / AF_{REST}$).
BOLD MRI image showing the identified ischemic area (yellow) using the ARREAS statistical framework.

Experimental Validation

The study utilized a canine model (9 subjects, 26 distinct studies) to compare the ARREAS metric (Ischemic-Extent) against the conventional SIR metric. To ensure accuracy, the researchers used microsphere analysis—the gold standard for measuring absolute blood flow—to verify the actual Myocardial Flow Ratio (MFR) and the severity of the induced stenosis.

Performance and Correlation

The results demonstrated that the size of the ischemic territory is a far more robust indicator of disease than the magnitude of the signal drop. The ARREAS method significantly outperformed conventional techniques in several key areas:

  • Diagnostic Accuracy: At end-systole, the Area Under the Curve (AUC) for detecting critical stenosis was 0.97 for IE, compared to only 0.83 for SIR.
  • Sensitivity and Specificity: The IE metric achieved 100% specificity and 88% sensitivity, whereas SIR lagged behind at 88% specificity and 77% sensitivity.
  • Physiological Correlation: The IE metric showed a strong exponential correlation ($R^2 = 0.8$) with actual microsphere-measured flow, while the conventional SIR showed a much weaker linear relationship ($R^2 = 0.5$).
Comparison of Ischemic Extent (IE) and Segment-Intensity Response (SIR) against the ground-truth Microsphere Flow Ratio (MFR) at end-systole (A) and end-diastole (B).

Key Insights and Future Directions

A critical finding of the study was the superiority of measurements taken at end-systole. When the heart muscle is most contracted and thickest, the BOLD signal appears more stable, likely due to the increased tissue volume within the imaging slice. Furthermore, by focusing on "contiguous" regions, ARREAS effectively filters out random image noise that typically plagues BOLD MRI.

While the results are promising, the study's reliance on manual myocardial border tracing highlights a need for future automation via deep learning. Additionally, while the canine model provided excellent controlled data, further validation in human subjects with complex multi-vessel disease is necessary to confirm the clinical utility of the ARREAS framework.

Summary updated

Abstract

BOLD MRI may be used for detecting myocardial oxygenation changes secondary to coronary artery stenosis. However, current approaches for analyzing BOLD changes are suboptimal for detecting critical stenosis.
BibTeX
@article{tsaftaris2011area,
  author = {Sotirios A. Tsaftaris and Richard Tang and Xiangzhi Zhou and Debiao Li and Rohan Dharmakumar},
  title = {{An area-based imaging biomarker for characterizing coronary artery stenosis with myocardial BOLD MRI}},
  journal = {Journal of Cardiovascular Magnetic Resonance},
  year = {2011},
  doi = {10.1186/1532-429x-13-s1-o22},
}

Loading related content...

Figure