INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

DIFFUSION OF WATER is sensitive to changes in the structural geometry and organization of its molecular environment and can be noninvasively measured by using nuclear magnetic resonance (MR) techniques via the signal attenuation caused by the loss of spin-phase coherence in the presence of magnetic field gradients (30). The MR apparent diffusion coefficient (ADC) of tissue water has been found to decrease dramatically after acute cerebral ischemia (21). The ensuing research, the subject of several reviews (9, 11, 27, 32), suggests that the rapid decrease and reversal of the ADC change on recovery may offer the potential for early, noninvasive detection of stroke and reliable prediction of ischemic brain injury. The ADC decrease has been associated with the depletion of cellular energy stores (19), and cell swelling resulted from the disruption of membrane electrochemical homeostasis (3). Theoretical models of water diffusion in tissues (15, 29, 31) and direct measurements in single cells undergoing physiological perturbation (12) indicate that the predominant biophysical factor underlying the ADC decrease is the reduction of extracellular water volume fraction during cell swelling (rather than changes in the intrinsic diffusion characteristics of intra- and extracellular water). The specificity of the ADC decrease to cell swelling, which is a common pathophysiological response in distressed tissues, has been extended to diffusion-weighted MR imaging studies of other disorders such as cortical spreading depression (5), status epilepticus (36), and acute renal failure (33).

MR techniques have long been used in the study of ischemic heart diseases and, in general, have included assessments of 1) myocardial metabolic activities using MR spectroscopy (8, 34), 2) contractile function using rapid imaging (23, 28) and MR tagging (2, 18) methods, and 3) MR image intensity changes induced by endogenous relaxation mechanisms (17, 24) or exogenous contrast agents (8, 26). Each of these approaches has its unique advantages and drawbacks. For example, MR spectroscopy may provide better specificity for pathophysiological responses but is limited in spatial and temporal resolution because of the inherently low concentration of tissue metabolites. Although imaging techniques based on mapping the distribution of water molecules offer improved resolution, regional contractility measurements may not be effective in distinguishing viable but dysfunctional (e.g., stunned or hibernating) tissue from nonviable myocardium. On the other hand, tissue viability and perfusion studies using exogenous contrast agents have been complicated by quantitation difficulties in relating image intensity to contrast agent concentration in tissues.

Because of the complexity and heterogeneity of the pathophysiology, proper assessments of ischemic myocardial injury may require not just one but several examinations that are each sensitive to different responses in a comprehensive study. In this regard, tissue diffusion changes may provide new insights regarding the dynamics of myocardial injury and may complement established approaches in the study of ischemic heart diseases. The goal of this study is thus to determine whether tissue ADC changes occur after regional myocardial ischemia in an isolated, perfused rabbit heart model and whether diffusion changes can be used to monitor the nature and evolution of tissue injury.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Isolated, perfused heart preparation. Hearts were obtained from New Zealand White male rabbits and were perfused retrogradely via the aorta using an MR-compatible Langendorff apparatus as described previously (1). Tissue and perfusate temperature were maintained at 37°C via a water-filled heat-exchange circuit. The perfusate was continually equilibrated with a 95% O₂-5% CO₂ gas mixture. To avoid excess hydrostatic pressure accumulation and distension of the left ventricle (LV) resulting from thebesian drainage or aortic valvular insufficiency, a thin (1 mm OD) polyethylene tube was inserted into the LV through the mitral valve to serve as a vent. The beating heart was perfused with a modified Krebs-Henseleit (KH, pH 7.4) bicarbonate buffer containing (in mM) 118.0 NaCl, 25.0 NaHCO₃, 5.0 dextrose, 4.6 KCl, 2.5 CaCl₂, 1.2 KH₂PO₄, and 1.0 MgSO₄. Because of the sensitivity of diffusion-weighted MR images to bulk motion, the heart was arrested before imaging. Cardiac arrest was induced and maintained by perfusing with a cardioplegic solution [modified St. Thomas' Hospital solution (35)] that consisted of (in mM) 110.0 NaCl, 16.0 MgCl₂, 16.0 KCl, 10.0 NaHCO₃, 5.0 dextrose, and 1.2 CaCl₂. Bovine serum albumin (BSA, 3% wt/vol) was added to both perfusates to minimize interstitial edema formation.

MR imaging. MR imaging experiments were conducted using a General Electric 4.7-T Omega CSI instrument equipped with shielded gradients (Accustar). The perfused heart, together with a small sealed tube of perfusate standard (to provide reference for the ADC), was placed inside a 31-mm-diameter loop-gap radio-frequency transmitter-receiver coil. On cardiac arrest (which occurred several minutes after perfusion was switched to the cardioplegic solution), a series of four baseline diffusion-weighted images [256 × 64 zero-filled to 256 × 256 matrix, 40-mm field of view, 4-mm slice thickness, echo time (TE) 50 ms, repetition time (TR) 1 s, and 2 averages] containing a short-axis view of the heart were acquired using a standard spin-echo sequence, with diffusion- weighting levels (or b-values) of 67, 266, 599, and 740 s/mm². In the first heart, diffusion was encoded in the phase-encoding direction. Because this encoding direction had a component parallel to the myocardial fiber orientations in the targeted ischemic myocardium, anisotropic diffusion and the rotation of fibers caused noticeable, systematic transmural variations in image intensity. This dependence of the ADC on the diffusion-encoding direction and the myocardial fiber orientation is schematically explained in Fig. 1. To reduce the nonuniform anisotropic diffusion effects, we encoded diffusion in the next four hearts in the readout direction, which was mostly perpendicular to the regional myocardial fibers.

View larger version (43K): [in this window] [in a new window]   Fig. 1.   Effect of anisotropic myocardial fiber orientation on diffusion measurement. Ventricular myocardium contains parallel fibers that are characterized by transmural rotation of fiber angle. Because diffusion is faster along fibers than across fibers, encoding diffusion parallel to fiber planes (a) would produce a transmural variation in the measured apparent diffusion coefficient (ADC). In contrast, the epicardial-endocardial axis (b) is mostly perpendicular to fibers and would thus yield a relatively uniform ADC.

Data analysis. ADC maps were generated for each series of diffusion-weighted images via pixel-by-pixel nonlinear least-squares fitting according to

(1)

where I is the signal intensity, I₀ is the diffusion-independent signal intensity (including spin density and relaxation effects), and D is the ADC. The acquisition starting time for each image series was assigned to be the time point (in min) of the corresponding ADC map, with t = 0 representing the preocclusion time point. It is noted that the calculated exponential decay constant (i.e., the ADC) based on images acquired with constant TR and TE would be independent of the longitudinal (T₁) and transverse relaxation time (T₂) contrast present in the individual images. Furthermore, the ADC calculation implicitly assumed a monoexponential signal intensity decay as a function of diffusion weighting. Because each signal intensity decay curve was sampled at only four diffusion-weighting b values, no attempt was made to examine changes in the nature of diffusion (e.g., mono-, bi-, or complex exponential decay).

(2)

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Representative images of an isolated, perfused heart are shown in Fig. 2, including least-diffusion-weighted images obtained before (A) and 10 min after LAD occlusion (B), a Gd-DTPA-enhanced image (C), and a corresponding ADC map at t = 150 min (D). The images contain a short-axis view of the heart in which the LV appears as an annular-shaped region. The nonperfused LV appears in the Gd-DTPA-enhanced image (Fig. 2C) as a relatively bright area because the normally perfused myocardium has a shorter T₂ from the presence of high concentration of the contrast agent. The postocclusion image (Fig. 2B) contains an LV region that has reduced signal intensity, and the region is morphologically similar to the hypoperfused zone demarcated in Fig. 2C. On the other hand, the ADC map (Fig. 2D) shows that the ADC within the hypoperfused LV has decreased; however, the area of reduced ADC is smaller than the hypoperfused region.

View larger version (144K): [in this window] [in a new window]   Fig. 2.   Representative images of an ischemic heart. A-D contain short-axis view of heart showing annular left ventricle (LV). Images are least-diffusion-weighted spin-echo images obtained before (A) and 10 min after ischemic insult (B), gadolinium-diethylenetriaminepentaacetic acid-enhanced gradient echo image that delineates hypoperfused LV as area of higher intensity (C), and calculated ADC map (to be distinguished from a diffusion-weighted image) at 150 min after left anterior descending coronary artery (LAD) occlusion, containing a region of reduced ADC within the hypoperfused LV (D). Circular object at top left of cross-sections is perfusate standard. Images have been numerically scaled to enhance contrast.

The calculated slopes of the ischemic and nonischemic ADC as functions of time are tabulated in Table 1. The normalized ADC of the ischemic region shows a statistically significant (P < 0.03) decreasing trend over time, with an average slope of 1.38 ± 0.20 × 10³ min¹ (n = 5, ±SE). In contrast, no significant trend was found in the ADC of the nonischemic region (0.01 ± 0.10 × 10³ min¹ average slope, P > 0.8). The linear regression results are summarized graphically in Fig. 3.

View larger version (25K): [in this window] [in a new window]   Fig. 3.   ADC of ischemic and nonischemic myocardium as functions of time after LAD occlusion. Each data point represents averaged ADC (n = 5) normalized to both mean ADC of water standard and the preocclusion (t = 0) value. Error bars represent ±SEs. Solid lines are reconstructed from averaged parameters of individual linear regression reported in Table 1.

Individual results of the sigmoid logistic regression describing the morphological progression of the ischemic ADC reduction are tabulated in Table 2. Figure 4 shows the average percent areas of reduced ADC and the response reconstructed from averaged curve-fit parameters. The results indicate that the ischemic myocardium showing reduced ADC appears after LAD occlusion and gradually increases in size over time. Extrapolating from the average curve-fit response, the percent area of reduced ADC reaches %P = 50 at approximately t = 130 min after LAD occlusion.

View larger version (22K): [in this window] [in a new window]   Fig. 4.   Progression of area within ischemic myocardium showing decreased ADC. No. of pixels that showed ADC values <50% of perfusate ADC was measured and normalized to area of ischemic myocardium. Data points and error bars represent average (n = 5) percent areas and ±SEs, respectively, at various time points. Response of each heart was curve-fitted to a sigmoid logistic function (Eq. 2). Solid line represents response reconstructed from averaged curve-fit parameters reported in Table 2.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Comparison between the least-diffusion-weighted images before and immediately after LAD occlusion (Fig. 2, A and B, respectively) reveals a region of reduced intensity in the LV of the postocclusion image. The anatomic similarity between the region and the nonperfused area demarcated by Gd-DTPA enhancement (Fig. 2C) indicates that the intensity reduction is linked to the perfusion deficit (i.e., ischemia) in the myocardium. Similar intensity changes after regional myocardial ischemia were reported in a previous study (6). Qualitatively, Fig. 3 indicates that the ADC of the ischemic myocardium appears to remain constant initially, then steadily decreases after 60 min following LAD occlusion. Although the linear regression analysis cannot determine the transition time point in which the ADC reduction begins and is likely to underestimate the slope of the decrease, the decreasing trend of the ADC of the ischemic LV was found to be significant (Table 1). In contrast, the ADC of the nonischemic myocardium remained unchanged over time. The morphological analysis summarized in Table 2 and Fig. 4 indicates that the ADC decrease in ischemic myocardium reflects an expansion of the area that has reduced ADC within the hypoperfused myocardium rather than a uniform ADC decrease in the entire hypoperfused tissue. These results, combined, suggest that the temporal and morphological progression of the ADC reduction in the ischemic myocardium is a delayed and gradual process.

The dynamics of the ADC decrease may reflect the delayed myocyte swelling after myocardial ischemia. Electron microscopic examination of cellular ultrastructure has revealed that little or no cell swelling occurs during the reversible and early stages of irreversible ischemic injury (13). Direct measurement of tissue water and electrolyte content has found that the ischemic canine myocardium is capable of at least partial cell volume regulation through anaerobic glycolysis (14). The ability of myocytes to regulate cell volume is lost after 60 min of ischemia following the onset of irreversible injury (i.e., myocardial infarction). The transition time point between reversible and irreversible cell damage is likely dependent on the severity of ischemia, degree of collateral arterial flow, and the animal species. Studies of dog hearts under total ischemia have found evidence of irreversible damage as early as 20-40 min after ischemic insult (14). However, it has been reported that the jeopardized myocardium can be salvaged by reperfusion as late as 3-6 h after ischemia (25).

The delayed and gradual ADC decrease after myocardial ischemia is markedly different from the dramatic change observed in cerebral ischemia. The ADC decrease in the brain has been found to take place within minutes after the onset of ischemia (7) before any detectable relaxation contrast changes (21). Moreover, close anatomic correlation has been observed between the hypoperfused brain regions delineated by contrast-enhanced MR imaging and areas of reduced ADC in diffusion-weighted images (16, 22). The discrepancy between the MR appearance of myocardial and brain ischemia may originate from basic differences in the tissue type and pathophysiological responses (e.g., timing of cell swelling) of these organs. The intensity decrease in the ischemic LV (Fig. 2B, which contained mixed T₁ and T₂ contrast) may be linked to hemodynamic alterations, although it cannot be determined from the present experimental protocol whether this reflects primarily a T₁ or T₂ change. The T₁ of the myocardium has been found to depend on perfusion in a similar heart model (1). Hemodynamic effects are likely to be more pronounced in the heart than the brain because the heart has a higher vascular volume than the brain. Vascular (blood) volume occupies ~13% of myocardial space (10), compared with 5% of cerebral space (4).

Although the underlying pathophysiology needs to be correlated with histology, the dynamics of the diffusion changes and the sensitivity of tissue diffusion to cell swelling may link the ADC decrease to the onset of irreversible tissue injury, namely myocardial infarction. The morphological progression of myocardial infarction in the rabbit heart was previously examined using histochemical techniques by Miura et al. (20). The size of the infarcted region, as a percentage (%P) of the ischemic myocardium, was found to be a sigmoid function of the logarithm of time postischemia (specifically, %P = 30.5 log t + 3.3), with %P = 50 reached at t = 34 min. In contrast, the present study shows 50% of the ischemic myocardium had decreased ADC (below 50% of the perfusate ADC) at t = 130 min. The results of the two studies cannot be directly compared because of the dissimilar methodology employed. The determined dynamics of the ADC decrease necessarily reflect the choice of the ADC threshold (i.e., 50% of water ADC, which may have biased the area measurements toward regions that had a low initial ADC) and curve-fitting basis function. In addition to these methodological differences, it is possible that the myocardium showing decreased ADC represents only a subset of the infarcted region. Clearly, correlation between MR diffusion changes to tissue histology is needed to identify the exact pathophysiology underlying the myocardial diffusion decrease and to elucidate potential sources of the difference in the MR appearance of ischemic diseases in the brain and the heart.

The present study is limited in that infarct area was measured in only one plane. Because of the branching vasculature of the artery, a coronary occlusion is expected to cause a three-dimensional, inverted cone-shaped ischemic zone that starts at the site of occlusion and occupies progressively larger cross-sectional areas toward the apex of the heart. Interpretation of ADC measurements must also take into account that a relatively large slice thickness (4 mm) was used, hence producing a considerable through-plane volume-averaging effect in the observed diffusion characteristics. Nevertheless, the present study demonstrates the potential to noninvasively visualize morphological progression of acute myocardial tissue injury through alterations in the tissue ADC. The noninvasive approach is expected to have significant implications for determining infarct size in assessing, for example, the effectiveness of pharmacological protective agents in reducing irreversible tissue injury. Conventional histological methods require the heart to be mechanically sectioned and therefore preclude quantitation of the infarct size in the same heart at different time points. A noninvasive technique is advantageous because repeated measurements can be performed on the same hearts at different time points. Not only does the technique dramatically reduce the number of animals that need to be killed [e.g., 29 animals were used by Miura et al. (20) compared with 5 used in the present study], but also repeated measurements provide improved statistical power by controlling for the heterogeneity in the responses among different animals.

In conclusion, the tissue ADC has been shown to decrease after regional myocardial ischemia. The reduction reflects the expansion of a subregion of ischemic myocardium that has decreased ADC. In contrast to acute cerebral ischemia, ischemic ADC reduction in the myocardium is a delayed and gradual process and occurs subsequent to observable T₁ and T₂ contrast changes. The temporal and morphological progression of the ADC reduction suggests that the diffusion change may be related to the onset of myocardial infarction, which is known to involve myocyte swelling. These findings are promising for using the ADC reduction as an MR imaging tissue-contrast mechanism to noninvasively determine the viability of the ischemic myocardium and to assess the dynamics of acute myocardial infarction.