An evaluation of the time dependence of the anisotropy of the water diffusion tensor in acute human ischemia

Abstract

We have performed MRI examinations to determine the water diffusion tensor in the brain of six patients who were admitted to the hospital within 12 h after the onset of cerebral ischemic symptoms. The examinations have been carried out immediately after admission, and thereafter at varying intervals up to 90 days post admission. Maps of the trace of the diffusion tensor, the fractional anisotropy and the lattice index, as well as maps of cerebral blood perfusion parameters, were generated to quantitatively assess the character of the water diffusion tensor in the infarcted area. In patients with significant perfusion deficits and substantial lesion volume changes, four of six cases, our measurements show a monotonic and significant decrease in the diffusion anisotropy within the ischemic lesion as a function of time. We propose that retrospective analysis of this quantity, in combination with brain tissue segmentation and cerebral perfusion maps, may be used in future studies to assess the severity of the ischemic event.

Introduction

It is well known that the onset of ischemia in human brain has a profound influence on the effective diffusivity of water, when measured using various modalities of MRI experiments.1, 2 This phenomenon has been employed in recent years to identify and assess the extent of ischemic injury in a wide variety of clinical MRI investigations.3, 4, 5, 6, 7, 8 However, only recently it has become evident that, if the long range diffusivity of water is fully characterized by determining the elements of the water diffusion tensor, a wealth of information regarding the structure of the tissue in which the water resides is also made available.9, 10, 11, 12, 13

The fundamental concept that we wish to exploit is that after the onset of an ischemic event, alterations to the integrity of the cell structure occur within the volume of the infarct.14, 15 These changes impart a signature to the character of the water diffusion in such tissue, by virtue of the effect that they have on the dynamics of water transport. Most diffusion sensitive MRI examinations using clinical scanners, are made on a time scale which is long (∼100 ms) when compared with the time that it takes for a water molecule to diffuse across one cell diameter. In this time scale, most water molecules encounter any available barrier to diffusion that the cell structure may present and the spatial properties of water diffusion appear different from what they would be if the cellular structure had not been compromised by the onset of ischemia. We state this explicitly to clarify that our observations reflect changes in addition to those that may be experienced by the intrinsic self diffusion of water molecules in the ‘hindered,’ intracellular regime. In that spatial domain the effects of acute ischemia have been well characterized by other workers.16, 17, 18

The influence of the state of the tissue structure after an ischemic event on the character of the water diffusion is observable in MRI experiments even when the water diffusion in human brain can be represented by a scalar quantity such as the apparent diffusion coefficient (ADC), or in more careful experiments, the average diffusivity (trace/3) of the diffusion tensor.3, 4, 5, 6, 7, 8 However, the geometry of the cell structure imposes a directional dependence to the nature of water diffusion and hence computation of all the elements of the water diffusion tensor yields a richer and more accurate representation of the phenomenon in question.19, 20, 21

As originally suggested by Basser and Pierpaoli,10 the extent of cellular organization and tissue similarity in human brain are reflected in the degree of anisotropy of the water diffusion tensor. Moreover, a pathologic condition such as focal ischemia induces structural changes to the tissue that give rise to a loss of organization and structure at the cellular level. We are aware that these changes may alter the degree of anisotropy of the tensor12 and propose that after an ischemic event, a quantitative measure of the diffusion tensor anisotropy may be used to characterize the progression of the lesion. In parallel with our investigations, preliminary work with cerebral ischemic lesions has established that within the volume of the infarcted tissue the degree of diffusion anisotropy is substantially modified.22, 23, 24, 25 Similar observations have been made in MRI studies of animal models of acute ischemia.26 Most of these publications report measurements made in a single examination and compare the observations made within the infarct, with those in a contralateral ‘normal’ volume or with those made in ‘normal’ control subjects.

In the present work, we have examined the degree of diffusion anisotropy at each pixel within the infarct volume. This was performed up to eight times over the first 5 days following presentation, with a final examination 90 days later. We propose that after retrospective analysis of the temporal evolution of this parameter—in combination with information on possible blood perfusion deficits—it may be possible to use its relative changes as a measure of the severity or potential outcome of the original ischemic injury.

While the analysis above offers a novel approach to the characterization of ischemic phenomena, we also realize that any instantaneous measure of the absolute value of diffusion anisotropy is influenced by the type of tissue into which the infarct is expanding. To include this factor in our analysis, we also examined the evolution of the anisotropy of those tissue voxels which appeared healthy upon first examination, but which developed a diffusion abnormality in subsequent time points as a result of the infarct growth.

The methodology used to achieve this relied on coregistration of the images from all time points to those which were acquired in the first examination (time point 1). We were then able to map the largest infarct volume (at approximately 4 days) onto the data set of time point 1. It is only by these means that one can accurately follow the evolution of the anisotropy of the diffusion tensor throughout the progression of the ischemic event. Implicit in this analysis is the assumption that at least in the first 100 h after the onset of ischemic symptoms, the volume of tissue which exhibits a diffusion abnormality corresponds to that of the infarcted tissue. Since our data has been collected during and beyond this time window, we discuss some of the implications that our observations may have on this and other assumptions about the evolution of ischemia in human brain.