The reproducibility and statistical power of brain T2 mapping at 7 tesla in naïve rats in vivo
Serguei Liachenko1

1National Center for Toxicological Research, US FDA, Jefferson, AR, United States


The baseline behavior of T2 relaxation at 7 tesla in different parts of the rat brain was studied to provide the foundation for possible biomarker performance evaluation.


Quantitative T2 MRI mapping of the rat brain in vivo offers unique opportunities to provide non-invasive biomarkers of neurotoxicity related changes in non-clinical studies [1, 2]. To better understand the potential performance of this type of biomarker baseline characteristics of T2 mapping need to be determined. The aim of this study was to characterize the between- and within-subject reproducibility of T2 mapping in control (untreated) rats over time and to assess the statistical power of such an approach using a relatively large population of research animals.


An animal use protocol was approved by the National Center for Toxicological Research IACUC. Male Sprague-Dawley rats (N = 138, 360 ± 40 g) were used. All animals were untreated and scanned in the MRI one time except for a separate cohort (N = 21), in which the animals were scanned, injected with saline (once, ip, 2 ml/kg) and imaged again at later time points: 2 hr (N = 5), 24 hr (N = 7), 48 hr (N = 5), 144 hr (N = 5), or 288 hr (N = 4). MRI was performed using a 7 tesla Bruker Biospec AV III equipped with a 4-channel array rat brain RF coil. Animals were anesthetized using isoflurane (3% induction, 1-2% maintenance in oxygen) and body temperatures maintained at 37.3 ± 0.6°C. For T2 mapping a multi-echo spin echo sequence was used (MTX 192 x 192 x 24, FOV = 3.84 x 3.84 x 2.4 cm, echo spacing = 15 ms, 16 echoes, TR = 6 s, NA = 1). T2 maps were skull stripped and co-registered to the template image using the surface registration feature in Analyze 12.0. The resultant T2 maps were segmented using an in-house developed rat brain atlas that was co-registered to the same template. The average T2 relaxation values were calculated in all segmented regions. Statistical analysis (one-way ANOVA with Student-Neuman-Keuls post hoc tests, and power statistical analysis) was performed using SigmaStat.


The between-rat reproducibility of the T2 mapping was excellent as can be seen in figure 1 that shows the T2 maps of a representative rat brain (A) as well as the T2 maps averaged for all rats (B) with corresponding CV% maps (C). Note that the variability in T2 values in the areas away from the borders and CSF is below 10%, while in the places close to tissue interfaces the variability increases significantly, which is most likely caused by imperfect co-registration and variations in brain sizes and shapes. Table 1 shows the results of the segmentation of all baseline T2 maps with corresponding estimates of the statistical power (number of animals required to detect a 5% change at 80% power and a significance level of 0.05). Figure 2 shows the changes in mean T2 values in some anatomical regions (striatum, hippocampus, thalamus, and olfactory bulb) over time. These values were significantly elevated at the 24 hr time point in all brain regions except the olfactory bulb and could be a result of the protracted effect of general anesthesia received during the previous imaging session (at the 0 time point). At all other time points (2 hr, 2, 6, and 12 days) no significant differences were observed with respect to baseline (time point 0).


It has been established that T2 relaxation values for water protons in the normal rat brain are very stable and reproducible with high statistical power. This allows for the detection of very small changes using small group sizes.and constitutes the basis for the further development of the quantitative T2 mapping method as a biomarker of neurotoxicity with the prospect of future formal qualification.


This work was funded by US Food and Drug Administration, project #E07418.


1. Hanig J., et al. Regul Toxicol Pharmacol. 2014, 70(3): 641-7.

2. Liachenko S, et al. Toxicol Sci. 2015, 146(1): 183-91.


Figure 1. Representative T2 maps of a single rat (A) and corresponding averaged T2 maps (B) and coefficient of variance maps (CV%, C). Three representative slices from the whole T2 stack are shown.

Figure 2. Longitudinal changes in T2 values in some brain areas. There were no differences compared to baseline at any time except at day 1 day in all areas except the olfactory bulb, which did not show any changes at any time point.

Table 1. T2 values and number of animals needed to obtain referenced statistical power for different anatomical structures. N - number of animals needed to detect 5% change at the power of 0.8 and P < 0.05.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)