Summary

The objective of this study was to quantify microstructural remodeling in peri-infarcted and infarcted porcine myocardium using diffusion tensor MRI (DT-MRI) for the first time. High resolution ex vivo late gadolinium enhanced (LGE) MRI was used to segment the DT-MRI data into normal, peri-infarct and infarcted myocardium. LGE-MRI based segmentation produces regions with significantly different microstructural remodeling.

Background

T1-weighted late gadolinium enhanced (LGE) magnetic resonance imaging (MRI) is recognized as the “gold standard” for MRI based myocardial infarct mapping [1]. It is unclear how variations in LGE signal intensity relate to microstructural remodeling. Diffusion tensor magnetic resonance imaging (DT-MRI) enables 3D evaluation of soft tissue microstructure. Specifically, DT invariants provide a basis for evaluating changes in the trace (TR, magnitude-of-isotropic-diffusion), fractional anisotropy (FA, magnitude-of-anisotropy), and the tensor mode (MD, type-of-anisotropy) as a consequence of remodeling [2]. Previous DT-MRI studies of microstructural remodeling in post-infarct myocardium have not used LGE to segment the remote, peri-infarcted, and infarcted myocardium [3, 4]. The objective of this study was to use high resolution ex vivo LGE MRI of post-infarct porcine hearts to segment remote, peri-infarcted and infarcted myocardium and subsequently use this segmentation to quantify microstructural remodeling in peri-infarcted and infarcted myocardium using DT-MRI for the first time.

Methods

Antero-septal infarctions in adult female porcine hearts (N=3) were achieved via micro-bead injection distal to the mid left anterior descending coronary artery. After 8-weeks, Gd-DTPA was injected (0.1mmol/kg) and allowed to circulate for 15 minutes before euthanizing. Normal adult porcine hearts (M=3) served as controls. The hearts were excised and LGE MRI (0.33x0.33x0.50mm resolution) began within 2 hours of sacrifice. Immediately afterwards, co-registered DT-MRI (1x1x3mm resolution) was performed. Myocardial voxels were segmented as normal, peri-infarct or infarct based on signal intensity (SI) thresholds of the LGE images for each heart. We defined bootstrapped histograms and medians with 95% confidence intervals (95%-CIs) of each DT invariant in order to make statistical comparisons of non-Gaussian datasets tractable. Remote myocardium in infarcted hearts was also compared to myocardium of control hearts.

Results

DT invariant medians and 95%-CIs for segmented myocardium in each heart are listed in Table 1. Invariants in control hearts were similar to remote myocardium in infarcted hearts (Table 1). Intra-heart statistical differences between segmented myocardium were significant. Figure 1 depicts invariant maps for a short axis slice of one heart accompanied by bootstrapped histograms with 95%-CIs for invariant data for the whole heart. LGE and remote/infarct/peri-infarct segmentation for the same slice are also shown.

Table 1 Tensor invariant medians and 95%-CIs for normal, remote, peri-infarcted, and infarcted myocardium
Full size table
Figure 1
figure 1

(A) Short-axis T1w LGE image depicting enhancement of the infarct. (B) Segmentation of the T1w LGE slice into remote, peri-infarcted and infarcted myocardium using LGE SI thresholds. Trace map (C) and whole heart segmented bootstrapped histograms (D); FA map (E) and whole heart segmented bootstrapped histograms (F); mode map (G) and whole heart segmented bootstrapped histograms (H).

Full size image

Conclusions

LGE segmentation of DT-MRI data identifies regions of statistically significant microstructural remodeling in peri-infarcted and infarcted myocardium. Improved LGE segmentation methods hold promise for identifying regions of microstructural remodeling when DT-MRI is not available.

Funding

Mahajan R01-HL084261; Shivkumar R01-HL084261 and R01-HL067647; Garfinkel P01-HL78931 (Core).