The in vivo distribution of gadolinium closely localizes with myocardial fibrosis as proven by histologic correlations with ex vivo LGE images obtained at an image resolution where voxel volumes approached that of approximately 3 cardiomyocytes. Optimal ex vivo LGE acquisition requires limiting the temporal window to about 2.5 hours to avoid the dispersion of Gd-DTPA. Thus, post-mortem dispersion of Gd-DTPA effectively places a limit on the achievable resolution and signal to noise ratio. Both quantitatively and qualitatively, agreement between histology and LGE was excellent, and we observed unprecedented levels of structural detail in LGE images of chronic MI. LGE images were able to resolve narrow bands of collagen separated by a few cardiomyocytes. We confirmed that LGE can detect portions of the peri-infarct border zone where disorganized mixtures of fibrosis and viable myocytes are intermingled resulting in intermediate signal intensity on LGE. However with post-acquisition degradation of image resolution comparable to a clinical transmural resolution, LGE may have difficulty differentiating the peri-infarct border zone from partial volume averaging where voxels straddle the borders of sharply demarcated infarcts, since both types of tissue result in intermediate signal intensity.
The high image resolution in our study can be better appreciated when considering voxel volumes relative to the volume of a cardiomyocyte (). Human imaging typically uses an image resolution of about 1.5 × 1.8 × 6 mm resolution, which represents a volume equivalent to about 405,000 cardiomyocytes. Improving the image resolution to an isotropic (uniform in all directions) 1 × 1 x1 mm voxel still includes 25,000 cardiomyocytes per image voxel. The highest resolution used by prior validation studies23
was about 3125 cardiomyocytes, a resolution nearly 1000 times worse than the current study. The current high resolution images demonstrate the concept of viability approaching a cell-by-cell basis. Clinical viability assessment reflects the proportion of living cells in the myocardium as a continuous variable within a given voxel, rather than a binary ‘yes’ or ‘no’ variable. While clinical imaging can quantify viability measured on a scale relative to the wall thickness,3
standard gadolinium contrast agents appear to track viability down to a cellular level.
That Gd-DTPA localizes thin strands of collagen represents an important extension of prior knowledge.23
Prior validation studies at lower resolution were limited by images of relatively large areas of MI with large, dense accumulations of collagen thus providing less information regarding the margins of the infarct. Imaging the edges of infarcts presents unique challenges. Since LGE uses larger and often non-isotropic imaging voxels, the edges of infarcts may be blurred to a greater extent than one might predict from the in-plane image resolution.23
Our high resolution data confirm the ability of Gd-DTPA to identify the heterogeneous peri-infarct border zone. Yet, our analyses of the significant relation between peri-infarct border zone size and image resolution (i.e., voxel size) also indicate that measurement of intermediate signal intensity voxels at lower resolution should be problematic. The apparent peri-infarct border zone on LGE images varies inversely with image resolution. Thus, while the in vivo
distribution of Gd-DTPA accurately depicts chronic MI and fibrosis, clinicians need to be cautious about over-interpreting the significance of the peri-infarct border zone given the limited resolution currently available clinical CMR scans, especially since artifacts related to motion and temporal segmentation pose additional potential pitfalls. Furthermore, changes in image resolution can change the measured ratio of infarct to border zone, a serious problem for future standardization.
The accurate identification of fibrosis in the peri-infarct border zone and elsewhere in the myocardium is important clinically because fibrosis impedes wave fronts of depolarization and leads to anisotropic conduction which is believed to be the substrate for arrhythmia and sudden death.13, 15, 16, 22, 24, 25
Despite the theoretical inability to resolve the edges of an MI, Yan and colleagues found that the peri-infarct border zone regions had prognostic value.9
Similarly, others have also found that quantification of tissue heterogeneity at the infarct periphery was associated with ventricular arrhythmia.12, 13
A critical mass of collagen does not appear to be necessary for the accumulation of Gd-DTPA around collagen; rather, Gd-DTPA accumulation parallels fibrosis extent even with scant degrees of collagen accumulation. This characteristic is essential for LGE detection of fibrosis in other cardiomyopathies26–28
where fibrosis may be present but scattered diffusely through the myocardium in a density far less than the scars typical of chronic MI. Therefore, Gd-DTPA may track the deposition of collagen in any cardiomyopathy characterized by myocardial fibrosis.
Conceptually, extracellular contrast agents are a marker of viability just as a intact cardiomyocytes membranes are a marker of viable cells – a condition lost in acute MI and also a condition lost when fibrous/collagenous tissue replaces cardiomyocytes. Low molecular weight extracellular contrast agents (i.e.,
~0.8 kDa for Gd-DTPA29
) rely on altered volume of distribution and delayed washout kinetics to generate tissue contrast between viable and either acutely infarcted myocardium of chronically scarred. Iodinated xray contrast agents that distribute in the extracellular space behave similarly.30, 31
Thus, our data could have implications for viability/fibrosis imaging with multidetector computed tomography32
and future extracellular contrast agents
It is interesting that gadolinium dispersion ex vivo
does not follow a simple temporal course. The sudden drop off in signal intensity and blurring of borders following 2–3 hours of stability suggests that some post-mortem event facilitates gadolinium dispersion, such as loss of membrane integrity, intra-cellular digestion of organelles, or destruction of tissue planes.33
Changes in gadolinium dispersion may be a method of monitoring onset of cell death and loss of membrane integrity.
Our study has limitations. First, myocardial fibrosis and collagen deposition in rats may differ from humans. Second, the generalizability of our findings to clinical CMR studies is highly conceptual since there are several orders of magnitude differences in resolution between the high resolution ex vivo rat images and anything currently possible in humans. We also note the absence of a histologic definition for the border zone. A histologic definition should address issues related to optimal magnification, staining techniques for collagen quantification, 2D vs. 3D assessment, and identification of the threshold of collagen deposition required for electrical disturbance. Nonetheless, our data show the ability of Gd-DTPA to mark subtle myocardial fibrosis including the potentially arrhythmogenic peri-infarct border zone defined by LGE. Third, the acute infarct model was used in this study to primarily define a temporal window for scanning and averaging of image volumes. Thus, histochemical staining is not available in those samples. Moreover, this approach ignored potential differences in the dispersion of Gd-DTPA between the acutely and chronically infarcted myocardium. Yet, the time window was confirmed qualitatively in each chronic infarct prior to averaging, and the conservative estimates from the acute infarcts worked well for the chronic infarcts. Fourth, we used the term gadolinium “dispersion” to describe the temporal course of loss of contrast localization since temporal course observed is not compatible with simple random diffusion but additional experiments will be required to better understand that process. Finally, we did not study how reperfusion of acute infarcts affects the border zone size on histologic or LGE images; further study is needed.
In conclusion, Gd-DTPA differentiates myocardial fibrosis following MI at nearly the cellular level but imaging with these agents at a clinical resolution is up against serious issues related to partial volume problems when considering subtle issues like the peri-infarct border zone.
Myocardial fibrosis represents one of the pathologic hallmarks of electrical and mechanical remodeling. For chronicmyocardial infarction (MI), measures of fibrosis in the peri-infarctborder zone by cardiac magnetic resonancelate gadolinium enhancement (LGE) are characterized by intermediate signal intensityand appear to stratify risk. Yet,the current spatial resolution of clinical LGE images imposes important constraints on understanding what type of tissue is concealed within the border zone. Voxels of intermediate signal intensity in clinicalLGE imagesmay not differentiatethe disorganized mixtures of fibrosis and viable myocytes in the peri-infarct border zone from partial volume averaging where voxels simply straddle the borders of sharply demarcated MI. Both types of tissue result in intermediate signal intensity. In our study, high resolution ex vivo LGE data confirm the ability of standard gadolinium contrastto identify the peri-infarct border zoneandtrack viability or fibrosis to nearly a cellular level. Theapparent size peri-infarct border zone on LGE images varies inversely with image resolution. Thus, clinicians need to be cautious about over-interpreting the significance of the peri-infarct border zone given the limited resolution currently availableonclinicalLGEscans. Changes in image resolution can change the size of the border zone, complicatingfuture standardization.