Skip to content

Advertisement

We’re sorry, something doesn't seem to be working properly.

Please try refreshing the page. If that doesn't work, please contact us so we can address the problem.

Automated detection and quantification of microcirculatory oxygenation changes in the heart

  • Sotirios A Tsaftaris1,
  • Xiangzhi Zhou2,
  • Richard Tang2,
  • Aggelos Katsaggelos1,
  • Debiao Li2 and
  • Rohan Dharmakumar2
Journal of Cardiovascular Magnetic Resonance201012(Suppl 1):P216

https://doi.org/10.1186/1532-429X-12-S1-P216

Published: 21 January 2010

Keywords

Coronary Artery StenosisAdenosine StressPerfusion DeficitFluorescent MicrosphereAutomatic Quantification

Introduction

Blood-oxygen-level dependent (BOLD) MRI may be used for detecting myocardial oxygenation (MO) changes secondary to coronary artery stenosis (CAS). Under pharmacological stress, the myocardial territory affected by CAS appears hypointense relative to healthy regions in BOLD images.

Purpose

To test a method for automatic quantification of myocardial signal changes reflecting the regional variations in oxygenation against true microsphere flow measurements obtained from controlled canine studies.

Methods

Data Acquisition: Short-axis 2D cine SSFP-based myocardial BOLD images were acquired in 7 dogs under adenosine stress with and without hydraulically-controlled left-circumflex CAS in a Siemens 1.5 T scanner. Scan parameters: resolution = 1.2 × 1.2 × 6 mm3; flip-angle = 90°; and TR/TE = 5.7/2.9 ms. Fluorescent microspheres were infused to measure true myocardial perfusion. Following imaging studies dogs were euthanized and the myocardial tissue was processed to ascertain perfusion. The flow within each segment was summed to obtain total flow μ F for each slice. Image Processing: End-systolic images were identified and segmented. Baseline images (BA) were used as reference, while stress without (AD) or with various levels of CAS (SS) were used as targets (TRG). BA myocardial intensities were collected and the mean (μ), variance (σ), and degrees-of-freedom of a Student's t-distribution were found. C M , defined as the size of the largest contiguous hypointense region (pixel intensity below μ-σ) divided by the number of pixels in the myocardium, was computed. C M ratios between BA and TRG images, Q M TRG, BA) = C M (TRG)/C M (BA), were also calculated. Statistical tests were used to show that Q M (AD, BA)<Q M (SS, BA). Finally, Q M was correlated against the ratio of microsphere flow ρ = μ F (TRG)/μ F (BA).

Results

Representative end-systolic images with hypointense regions automatically detected are color coded and shown in Fig. 1. Fig. 2 shows a box-plot demonstrating the change in Q M pre- and post-stenosis. Fig. 3 illustrates a scatter plot between Q M and ρ, and a non-linear fit with power of 0.99.
Figure 1
Figure 1

Color-coded end-systolic basline image (A); adenosine induced stress image (B); adenosine induced stress image with moderate LcX stenosis (C); and with severe stenosis (D). Only hypointense regions of the myocardium were color coded. Yellow hues correspond to values close to m-s as described in text, while red is for darker values. The small and minimally contiguous regions of hypointensity (yellow coded) in th BA and AD images may be attributed to signal inhomogeneties due to physiological nose, coil bias, or limitations in shimming. However, regions of hypointensity are significantly larger under stenosis (C and D). The premise of usin baseline images is to isolate the BOLD signal from background deviations. The CM values for the imges A to D are 0.05, 0.02, 0.13, and 0.14, respectively.Total microsphere flow for each case is 19.82, 53.41, 34.67, and 30.44, resppectively.

Figure 2
Figure 2

A box-plot Q M values derived when comparing baseline with adenosine stress images pre and post stnosis. Median is shown in red and notches demonstrate 95% confidence intervals. A non-parametric t-test (Wilcoxon rank-sum test) illustrates a significant difference among the two medians (P = 0.001).

Figure 3
Figure 3

A scatter plot between Q M (ration of the image-derived metric) and r (the ration of the flow). Q M = Q c (TRG)/C M (BA), where BA is without stenosis. Observe that when QM < 1, the ration of the flow is high indicating no apparent stenosis. However, when QM >1, the flow ratio drops significatly. When an exponential decaying function is fit through the data, an R2 of 0.68 is achieved and the regression is deemed significant (P < 0.001) with a power or 0.99.

Conclusion

The method is capable of automatically delineating the perfusion deficit territories (Fig. 1). Observe that C M increases as the total flow decreases, establishing the foundation for utilizing the ratio of C M between rest and stress studies as an image-based metric for detecting microcirculatory oxygenation changes. Fig. 2 supports the fact that Q M provides adequate power (0.8) in detecting CAS. The exponential relationship (Fig. 3) between microsphere flow and Q M has 0.99 statistical power. The work forms an initial step in the development of an objective and automated analysis of BOLD MR images and a metric for quantifying microcirculatory oxygen changes in the heart.

Authors’ Affiliations

(1)
Northwestern University, Evanston, USA
(2)
Northwestern University, Chicago, USA

Copyright

© Tsaftaris et al; licensee BioMed Central Ltd. 2010

This article is published under license to BioMed Central Ltd.

Advertisement