Volumetric assessment and longitudinal changes of brain structures in formalinized Beagle brains

High field MRI represents an advanced technique both for diagnostic and research purposes on animal models such as the Beagle dog. The increasing interest in non-invasive neuroscience, aging, and neuropathological research led to a need of reference values (in terms of volumetric assessment) for the typical brain structures involved and, nowadays, several canine brain MRI atlases have been provided. Since no reports are available regarding the measurements reproducibility and few information are available about formalin fixation effect on brain structures when applied to MRI segmentation, we assessed the segmentation variability of selected structures as a function of the operator (two operators segmented the same data) and their intrinsic variability within a sample of 11 Beagle dogs (9 females and 2 males, 1.6 ± 0.2 years). Then, we analyzed for one further Beagle dog (2 years old) the longitudinal changes in the brain segmentations of these structures corresponding four conditions: in vivo, post mortem (after euthanasia), ex vivo (brain extracted and studied after 1 month in formalin and after 11 months); all the MRI images were collected with a 3 T MRI scanner. Our findings suggest that the segmentation procedure can be considered overall reproducible since only slight statistical differences were detected, apart from the ventricles. Furthermore, in the post mortem/ ex vivo comparison, the majority of the structures showed a higher contrast leading to more reproducible segmentations across operators and a net increase of volume of the studied structures; this could be justified by the intrinsic relaxation time changes observed as a consequence of formalin fixation, that led to an improvement of brain structures visualization and then segmentation. To conclude, MRI based segmentation seems to be a useful and accurate tool that allows longitudinal studies, especially when applied to formalin fixed brains.


Results
198 199 Brain segmentation after 12 months in formalin 200 The segmentations obtained from two operators after 12 months in formalin are reported in 201  1). 215 The percentage variability of each structure operator O1 (solid line) and O2 (dotted line). It can be 216 noticed that the lowest variability, except for the ventricles, was observed for both operators for the 217 largest and more defined brain structures, thus suggesting that the MRI volume of the segmented 218 structure is influenced by the actual size and its intrinsic contrast with the surrounding parenchyma.

220
The spatial topography of these variations is shown in Fig 2, where the % CV, averaged across 221 operators, is overlaid on a T1w image of a representative dog.

226
As it can be seen in Table 1 and Fig 3, where the whisker plots of the analyzed structures are 227 reported, apart from the ventricles (p <0.01 -dotted box in Fig 3), a t-test evidenced no statistically 228 significant differences among the two users.  245 This might be justified since both operators experienced higher difficulty to segment the smaller 246 structures due to their relatively poor signal contrast.

A longitudinal study on brain structures 248
In this part of the study, for a representative dog, we assessed the longitudinal changes of 249 segmentations performed in vivo, post mortem, and ex vivo_1 (after one month in formalin) and ex 250 vivo_12 (after 12 months in formalin).
251 First, we observed that both T1w and T2w images showed a change of signal contrast between the 252 post mortem and ex-vivo data (Fig 4). 272 whose borders were difficult to define by both operators in vivo and post mortem phases (Fig 4 E-H).  310 (Fig 5B-D). However, we also observed for some structures, a large increase in volume. This was not 311 expected since the formalin fixations are reported to lead to shrinking and reduced tissue volumes 312 (24). This effect was particularly evident for Substantia Nigra (average increase of around 70% Fig   313 5B, C), followed by Globus Pallidus (around 30%) and Geniculate (around 20%). In general, for both 314 ex vivo phases, all the operators described an increased contrast to detect the borders of the various 315 structures, and the smallest structures were described to be sensitively easier to be segmented.
316 Considering the previous findings, this suggests that the increase of contrast seems to be ascribed to 317 the ex-vivo condition and thus the effect of fixation. Now, to study if this change remains stable over 318 time, we compared the two ex vivo conditions (ex_vivo_1 and ex_vivo_12). As it can be seen in Fig   319 5D, the highest changes fluctuate around 20%. This suggests that the increased time spent in formalin 320 did not influence significantly the volumes of the various structures. Basically, in this period the 321 structures' volumes remained stable. This is an important finding suggesting that the segmentation 322 can be performed even after a significant amount of time from the formalin fixation. In order to 323 understand if these volume changes were induced by an overall shrinkage or inflation of the brain, 324 we performed the coregistration of the data acquired at the different time points.
325 As an example, in Fig 6, we report the borders of the brain extracted (for a representative slice) from 326 in-vivo data overlaid to the T1w images obtained ex-vivo (central panel).

327
328 Fig 6. Data coregistration. T1w in vivo data are used as the reference to coregister the ex-vivo data 329 with a 6-parameter coregistration approach (no scaling included). As an example, a representative 330 slice is reported (A). The in vivo brain contours are overlaid to the ex-vivo data after the coregistration 331 (B). It can be noted a good agreement. Analogously, the in-vivo brain contours overlaid on the ex vivo 332 12 months show that the rigid coregistration successfully aligned the data (C). Since no scaling was 333 involved in the data transformation, this suggests that the brain did not experience any significant 334 inflation/shrinkage.

336
It can be noted that a rigid co-registration with 6 parameters, thus excluding any scaling factor, 337 was able to coregister the brain successfully. Moreover, as it can be noted in Fig 6 (left panel), the 338 sample applies to the coregistration of the ex-vivo data after 12 months. The fact that no scaling factor 339 was needed to coregister the data, seems to suggest that the brain did not experiment with any 340 significant inflation/deflation over time. Therefore, the changes observed in the volumes were likely 341 due to a change in the signal contrast.