This study demonstrates the feasibility of acquiring DTI data sufficiently robust to permit quantitative fiber tracking, as demonstrated in two regions of the corpus callosum, in the rat brain in vivo on a 3T clinical scanner. Advantages of employing insert gradients on 3T clinical scanners include the use of an identical prescription interface to human studies, and similar pulse sequences and field strength for preclinical animal and clinical human research, thus enhancing the translational nature of such research. Specifically, for a fixed b-value, insert gradients permit shorter echo time (hence, increasing the SNR) and reduced duration echo-planar readout compared with the product whole-body gradients. Although yet higher field strength improves SNR, the present data acquired at 3T had sufficient SNR (in as short a duration as 46 min with an isotropic 0.5-mm resolution) to yield robust fiber tracking and its quantification based on FA and ADC.
The 0.5-mm resolution at 46 min represented a reasonable scan time trade-off for in vivo animal studies. The parameters yielded an SNR of ~6.2, which was low but adequately robust for quantification of fiber tracking. The course of fibers is not orthogonal to acquisition planes, and isotropic acquisition is the optimal approach. We sacrificed in-plane resolution for slice thickness because thicker slices with higher in-plane resolution may have made more attractive FA images but limited fiber tracking. A further consideration involved the trade-off between the number of diffusion directions to use and the number of shots obtainable in an acceptable, in vivo scan session. We used six directions, the minimum needed to define the tensor, thus enhancing single-shot SNRs with multiple averages of the same direction rather than using more directions with fewer averages.
An across-subject ROI FA standard deviation of ~6 is consistent with human DTI clinical reports in disease states, with tractrography producing less variance, as is the case with the rat data herein, and more sensitive disease detection (Kanaan, et al., 2006
). Further, considering the group standard deviations for each DTI measure, we estimate that we would have 80 to 95% power with samples of only 10 to 12 subjects to detect differences with 2-tailed tests ranging from 5% of the mean genu FA to 15% of the mean splenium FA and ADC.
High-field ( ≥ 4.7 T) animal MR scanners provide higher SNR permitting the acquisition of higher resolution DTI data for a given scan time relative to lower field systems. At the same time, however, off-resonance effects also increase with field strength leading to severe image distortions in EPI-DTI on high-field systems. Whereas spin-echo based DTI sequences are less prone to susceptibility artifacts, as multi-shot techniques they are inherently more susceptible to motion artifacts, which can reduce the effective resolution. All of these trade-offs have to be taken into account when choosing a DTI acquisition protocol and field strength for acquisition. Our proposed method adds another alternative.
In agreement with other rodent (Harsan, et al., 2006
) and human (Huisman, et al., 2006
; Nakayama, et al., 2006
; Persson, et al., 2006
; Pfefferbaum & Sullivan, 2005
) studies, FA values observed herein were higher in the splenium than the genu, suggesting across-species consistency in white matter microstructural quality. The FA values for the corpus callosum reported here from both ROI and fiber tracking analyses fall within the range of published values in rodents from ROI analysis (range: 32 to 75%) (Guilfoyle, et al., 2003
; Harsan, et al., 2006
; Nair, et al., 2005
; Tyszka, et al., 2006
; Verma, et al., 2005
). The few studies in humans that have compared diffusion parameters calculated from fiber tracking and ROI analysis reveal that FA values obtained using the former method tend to be lower (Dubois, et al., 2006
; Thomas, et al., 2005
). Indeed, the only previous rodent study to date calculating FA and ADC using fiber tracking, achieved on a 1.5T scanner, reports lower FA values than most published results (e.g., FA= 30% for the corpus callosum in one of their acquisitions) (Lee, et al., 2006
). Although the fiber tracking process follows areas of maximum FA to generate a tract, FA in an identified tract is also determined by the minimum FA set in the lower limit parameter. Thus, it can be the case that fiber tracking FA can be lower than FA in a brain region with little partial voluming, and indeed we have reported such a difference in our human studies (Pfefferbaum, et al., 2006d
Our ADC values from both fiber tracking and ROI analysis fall within the range of reported ADC values for the corpus callosum in rodents (range: 620 to 740 × 10-6
/s) (Harsan, et al., 2006
; Nair, et al., 2005
; Tyszka, et al., 2006
). However, whereas rodent FA values fall more closely within the range of human infant FA values, ADC values from these rats were lower than in human infants (range: 930 to 1830 × 10-6
/s) (Arzoumanian, et al., 2003
; Dubois, et al., 2006
; Partridge, et al., 2004
; Thomas, et al., 2005
). In fact, rodent ADC values for the corpus callosum corresponded better with adult human ADC values (range: 380 to 850 × 10-6
/s) (Huisman, et al., 2006
; Lim, et al., 2002a
; Zhai, et al., 2003
Small animal imaging has become an invaluable tool for modeling brain damage incurred from a variety of causes, from spontaneously-occurring to genetically-linked neurodegenerative diseases to environmental insults from trauma and neurotoxins. DTI protocols have been particularly useful for developing longitudinal animal models of human conditions that follow a dynamic course, such as stroke, because of the microstructural alteration of white matter that can occur below the detection of conventional MR imaging of macrostructure (c.f., Adami, et al., 2002
; Lansberg, et al., 2000
). In parallel to findings in human alcoholism, a recent postmortem study of a rat model of the combined effects of chronic, voluntary alcohol consumption coupled with severe thiamine deficiency revealed thinning of myelin with electron microscopy (He, et al., 2006
) that was not visible with structural MRI (Pfefferbaum, et al., 2006a
) but may have been detected with DTI (c.f., Pfefferbaum & Sullivan, 2002
Finally, there are many research-dedicated human 3T systems in universities and medical schools. Often these are stand-alone systems, frequently acquired for human cognitive neuroscience studies. These systems can also be used for animal imaging with the advantage of product pulse sequence availability. With addition of insert gradients, the system's performance can be enhanced without the need to purchase a separate animal magnet system. The high strength of the insert gradients allows for thinner slices and much faster read-out, the latter being of particular value in echo-planar DTI acquisition, where B0 inhomogeneity distortion is a major problem.