Postmortem MR imaging of formalin-fixed human brain

Abstract

High-resolution postmortem neuroimaging of the brain can play a role in research programs by providing archival and reslicable images of brain specimens before permanent sectioning. These images can supplement evidence attained from both traditional neuropathological observations and in vivo neuroimaging. Differential brain tissue conspicuity, detectable with MRI, is determined by the density and mobility of water protons. Water content is about 70% in white matter, 80% in gray matter, and 99% in cerebrospinal fluid (CSF). To the extent that brain tissue contrast is determined by the number and microenvironment of water protons, timing parameters of MR image acquisition can interrogate this environment. Because the chemical environment of protons is different in living from dead tissue, optimal temporal imaging parameters, for example, for spin-echo imaging, commonly used for in vivo clinical and research study are different from those best for postmortem imaging. Here, we present a series of observations to identify relaxation times and optimal parameters for high-resolution structural imaging of formalin-fixed postmortem brain tissue using commercially available clinical scanners and protocols. Examples of high-resolution images and results from attempts at diffusion imaging are presented.

Introduction

MRI has been used to acquire images of the postmortem human brain for the past 15 years. These images offer a valuable complement to the work of the neuropathologist by providing a third dimension for histological specimens, documentation of the whole brain before it is sectioned, and a flexible tool for teaching (Boyko et al., 1994). One use of this technique has been to compare MRI sections acquired postmortem with equivalent physical sections prepared for neuropathological assessment. This approach has been applied to brains of people with diagnoses of AIDS (Grafe et al., 1990) and Alzheimer's disease Bobinski et al., 2000, Bronge et al., 2002, with a focus on identifying false positives, true positives, and false negatives in the MR images relative to histologically identified lesions such as infarction, necrosis, diffuse HIV-associated microglial nodules in AIDS cases, and various white matter changes in AD. Cross-modality comparison has also been made of qualitatively assessed atrophy in AIDS patients (Everall et al., 1997) and quantitatively assessed hippocampus volume in normal and AD patients (Bobinski et al., 2000). While these comparative studies have highlighted many similarities, they have also noted discrepancies attributable to the relative sensitivity of each technique to the characteristics of formalin-fixed postmortem tissue.

MR imaging of a postmortem brain specimen needs to accommodate the significant differences between a living and a dead brain. Ideally, the postmortem brain is intact and has been minimally damaged by removal from the skull; as such, it is drained of cerebrospinal fluid (CSF), blood no longer flows through the capillaries and it is devoid of its encasing skull. Even with formalin fixation, brain tissue requires support to remain stable and not collapse on itself. Furthermore, formalin fixation changes the microstructure of tissue and affects water mobility. Differences in water mobility between different tissue types are key to tissue discriminability by MRI. T1 (longitudinal) and T2 (transverse) relaxation times of water protons rely on mobility of water within the tissue. Formalin fixation reduces difference in water mobility between gray and white matter. Early declines in T1 and T2 after fixation lead to convergence of T1 values in gray and white (Tovi and Ericsson, 1992) and limit gray and white matter discriminability on T1-weighted postmortem MR images. Some imaging protocols yield a reversal of relative signal intensities between gray matter and white matter in fixed brains compared with in vivo images Boyko et al., 1994, Schumann et al., 2001. T2 also declines over time but not so much as T1. Changes in both T1 and T2 occur rapidly over the short term but stabilize after 3–4 weeks (Tovi and Ericsson, 1992), such that length of fixation has little impact on gray–white conspicuity after 3 weeks (Boyko et al., 1994). By contrast, differences in the relative density of protons in gray and white matter increase during the first weeks of fixation (Blamire et al., 1999). Thus, a spin (proton) density-weighted protocol is preferred for postmortem imaging as it provides excellent gray–white contrast (Schumann et al., 2001).

Initial postmortem MRI studies used conventional acquisition protocols, producing images with relatively thick slices in fixed orientations. Comparisons between MRI and pathological sections were made by matching pathological slices to acquired MR slices. More recent imaging developments (e.g., Schumann et al., 2001) allow acquisition of three-dimensional high-resolution MRI images with in-plane voxel dimensions on the order of 100 μm, allowing the image to be rotated in any orientation and thinly resliced with minimal distortion from interpolation. Improvement in resolution of in vivo imaging is limited by the living subject's tolerance for lengthy scan sessions, whereas the postmortem brain is ideally suited for time-dependent improvements in resolution limited only by computer storage space and monetary cost.

Consistent with postmortem histological findings of age-related degradation of white matter microstructure (Meier-Ruge et al., 1992) are recent results based on in vivo diffusion tensor imaging, which is particularly useful for visualizing white matter bundles and microstructure (Basser and Pierpaoli, 1996). The in vivo studies have revealed age-related disruption of brain white matter microstructural integrity in men and women that varies by region (e.g., Pfefferbaum and Sullivan, 2003; for review, Sullivan and Pfefferbaum, 2003). The mechanism for alcoholism-related white matter volume loss, restoration with alcohol abstinence, and disruption of microstructural integrity remains unclear but probably involves changes in both myelination and axonal integrity, which have been identified postmortem.

Postmortem imaging provides a technique to explore and quantify discrepancies between in vivo results and postmortem neuropathology as well as to provide a new tool to complement traditional and emerging neuropathological investigations. The goals of this investigation were to use standard clinical 1.5- and 3.0-T MR scanners and product sequences to establish MR imaging parameters for formalin-fixed human brain specimens that would yield high isotropic resolution, whole-brain coverage, and volumetric images. The target imaging modalities were conventional structural MRI and MR diffusion tensor imaging. Among the issues to address were effects of death and fixation on the microenvironment of brain tissue water protons and its effect on temporal imaging parameters, specimen movement, signal-to-noise ratio (SNR), and cost in terms of time and money.