Syllabus

MRI is capable of producing three-dimensional images at resolutions from sub-millimeter to a few microns, and has long been recognized for its potential to augment histopathological evaluations.¹ Investigation of sub-cellular structures by MRI provide the best examples of MR histology.^2-4 The MRI signal is also sensitive to many changes seen in pathology, even though some of these signal changes may not be very specific. Improvements in MRI hardware and techniques make MR histology more accessible to investigators.
MRI performed at or below 100-micron resolution is referred to as MR Microscopy (MRM). Nominal resolution of this magnitude is difficult to achieve in-vivo over large volumes, although recent advances in pulse sequences have put it within grasp.^5,6 On the other hand, sub-millimeter to tens of micron isotropic resolution is relatively easy to achieve on fixed tissues, primarily due to the ability to perform signal averaging over many hours or days, without having to contend with motion. Early studies used a combination of high-resolution in-vivo and postmortem localized-MRM to image small brain substructures.^7,8 However, postmortem MRI of whole-organs (e.g., whole brain) can help identify regions of pathology, which can then be efficiently targeted for further research using technique called MRI-guided histopathology.^9,10 Furthermore, improving the resolution by about a factor of 10 in each dimension results in images that approach the resolution required to image individual cells and its substructures, which in itself can help to better understand the origins of MRI contrast.^11,12
Several technical considerations are needed for imaging whole brains at high resolutions. Higher field strengths generate more signal that is beneficial for high-resolution scans. Furthermore, larger peak-gradient amplitudes are necessary to obtain higher image resolutions. MRM is therefore more easily performed on pieces of tissue on small bore (4.7T to 14T animal) systems equipped with more powerful gradient inserts. Nevertheless, clinical scanners have much larger volume over which the gradient is homogenous which can be of great advantage when imaging whole organs, even if they typically have lower peak gradient strengths. A fixed human brain measures about 15x10x12 cm³, and can be imaged on a clinical 7T or 3T scanner at 100s of micron isotropic resolution over few hours.¹³ High-performance gradients installed on large-bore scanners,¹⁴ can go a long way towards efficient MRM of a whole brain in the future. RF coils also play an important role in MRM. While custom coils can be designed to optimize the filling factor,¹³ several studies on clinical systems use the manufacturer’s clinical coils.¹⁵ Parallel transmit coils also improve image homogeneity.^16,17
Tissue relaxivity decrease rapidly in the initial hours of fixation^18,19 causing deterioration of resolution through blurring and loss of signal. Relaxivity and diffusivity of protons is also influenced by the postmortem interval and tissue fixation techniques.²⁰ For a brain that is immersion fixed in formalin for 14 days, the T1 relaxation time is approximately 550 ms and 700 ms from white and grey matter respectively, while T₂* relaxation time is between 15 and 20 ms at 7T. The diffusivity of water reduces to about 2x10^-4 mm²/s post fixation.19 Most MRM sequences use a gradient-echo type sequence (for T2*-images), and inversion-prepared pulse sequences (to more easily achieve T₁-weighting at high fields). Quantitative maps are an easy way of generating uniform images without bias. Steady-state sequences provide better signal to noise ratio. Others have perfused the tissue with MRI contrast agents to shorten T₁-relaxation and increase signal per unit TR.²¹
Image quality is also affected by magnetic field inhomogeneity and air bubbles that can be trapped in the tissue. One effective way to improve shimming and reduce artifacts is by immersing the brain in an inert solution such as Fomblin, a fluorinated oil, and to apply a small vacuum to the container on a rocker to remove air bubbles.^7,9,22 Lastly, the file size of whole brain MRI at a 100-micron isotropic resolution can be more than 2 GB in DICOM format and over 1 GB in compressed NIFTY format. Acquiring the images using a multi-channel receive coil can have its own challenges depending on the RAM available for image reconstruction engine on the scanner. Images at such high resolutions can also be acquired in slabs and stitched together in postprocessing using slice location information.
The following example highlights the utility of high-resolution postmortem imaging. Diagnosis of multiple sclerosis is often aided by visualization of white matter lesions in the brain. It was clear even in the early days that MRI was able to detect white matter lesions that were not readily visible to the naked eye, and that in-vivo and postmortem imaging will play an important role in aiding complete understanding of these signal changes.^10,23 In addition, postmortem MRI detects cortical lesions with higher sensitivity, likely because of the enhanced imaging resolution and SNR, as well as the diverging changes to relaxivity of normal cortex and cortical lesions during fixation process. In order to better understand these MRI signal changes, we performed MRM on small pieces of tissue from normal cortex and one that had a cortical lesion.¹² MRM was performed on the 11.7T (Bruker) system, using a custom-built close-fitting solenoid RF coil. We found that the hyperintensity in the cortical lesion was driven by loss of iron from oligodendrocytes. We used this information to design a T₂*-weighted pulse sequence, with CSF suppression to better visualize cortical lesions in-vivo on the clinical 3T scanners.²⁴

Acknowledgements

Research supported by the Intramural Research Program at the NINDS