Distinguishing Laterality in Brain Injury in Rabbit Fetal Magnetic Resonance Imaging Using Novel Volume Rendering Techniques Open Access

Introduction: Our laboratory has been exploring the MRI detection of fetal brain injury, which previously provided a prognostic biomarker for newborn hypertonia in an animal model of cerebral palsy (CP). The biomarker relies on distinct patterns of diffusion-weighted imaging-defined apparent diffusion coefficient (ADC) in fetal brains during uterine hypoxia-ischemia (H-I). Despite the challenges posed by small brains and tissue acquisition, our objective was to differentiate between left and right brain ADC changes. Methods: A novel aspect involved utilizing three-dimensional rendering techniques to refine ADC measurements within spheroids encompassing fetal brain tissue. 25-day gestation age of rabbit fetuses underwent global hypoxia due to maternal uterine ischemia. Results: Successful differentiation of left and right brain regions was achieved in 28% of the fetal brains. Ordinal analysis revealed predominantly higher ADC on the left side compared to the right at baseline and across the entire time series. During H-I and reperfusion-reoxygenation, the right side exhibited a favored percentage change. Among these fetal brains, 73% exhibited the ADC pattern predictive of hypertonia. No significant differences between left and right sides were observed in patterns predicting hypertonia, except for one timepoint during H-I. This study also highlights a balance between left-sided and right-sided alterations within the population. Conclusion: This study emphasizes the importance of investigating laterality and asymmetric hemispheric lesions for early diagnosis of brain injury, leading to CP. The technological limitations in obtaining a clear picture of the entire fetal brain for every fetus mirror the challenges encountered in human studies.

Intrauterine hypoxic-ischemic (H-I) events are established risk factors that may lead to cerebral palsy (CP) [1‒3]. The origins of CP trace back to perinatal insults and individual vulnerability [4‒6]. There is a long lead time from insult to clinical manifestation as the clinical diagnosis of CP can only be made typically at 12–24 months after birth [7, 8]. In our previous research, we established rabbit fetal H-I models mimicking human CP symptoms in rabbit neonates, inducing hypertonicity like in human children [9‒12]. There is a dire need for early biomarkers in the perinatal period to predict which fetuses are prone to CP. Early detection of CP holds significant importance, as interventions initiated during the early developmental stages have the potential to improve or even prevent the adverse outcomes resulting from perinatal injuries [13].

MRI diffusion-weighted imaging (DWI) holds promise due to its sensitivity in detecting early H-I brain injuries in human neonates [14] and animal models [15]. The apparent diffusion coefficient (ADC), reflecting brain tissue integrity, quantifies impedance to water diffusion within neural tissue. This enables the investigation of neurological aberrations within fetal brain tissue subjected to H-I insult [16, 17]. ADC values exhibit variations over time after H-I insults [18].

Early detection of the insult and monitoring ADC changes throughout this period are crucial for an accurate reflection of the extent of fetal brain damage. This approach helps prevent misinterpretation of ADC values that may recover to normal levels, indicating pseudonormalization [19]. Previous studies relied on two-dimensional MRI images to locate fetal heads and draw polygonal regions of interest (ROIs) over the entire fetal brain for calculating ADC means at different timepoints during the H-I insult [16, 17, 20].

In this study, we utilized three-dimensional rendering techniques to pinpoint fetal positions and drew spheroids extending into three orthogonal planes for ADC mean calculations. Using spheroids for ADC value calculation enabled accurate ADC values of the brain, even when obscured by multiple slices, enhancing measurement precision. Our goal in this study was to explore the perinatal implications by pushing the boundaries of MRI as a biomarker for H-I-induced fetal insults [16, 20].

Having successfully employed spheroids to calculate whole brain ADC values, our subsequent objective was to investigate differences between the left and right brain hemispheres by calculating ADC values within the spheroids of both hemispheres. This approach became important to answer the question if unilateral brain injury, such as strokes, occur with global hypoxia. While a previous study explored correlations between ADC values in different 2D slices of human neonates [19], our primary aim was to distinguish ADC values in the 3D volumes of separate brain hemispheres. Our null hypothesis was that there were no significant differences between left and right resulting from complete H-I insults. To the best of our knowledge, this work provides the first estimates of ADC values for the left and right hemispheres of E25 rabbit fetuses undergoing H-I.

Pregnant New Zealand white rabbit dams on the 25th day of gestation, corresponding to 79% of full term (E25), were subjected to uterine ischemia. Anesthesia induction was achieved using intravenous fentanyl (75 μg/kg/h) and droperidol (3.75 mg/kg/h), followed by the administration of an epidural containing 0.75% bupivacaine [17]. A balloon catheter was utilized to induce H-I. The catheter was inserted into the femoral artery and advanced into the descending aorta proximal to uterine arteries while remaining distal to renal arteries. Subsequently, the balloon was inflated causing H-I. The catheterized animal was then placed within an MRI scanner [17]. An MRI-compatible rectal temperature probe was employed to monitor the core body temperature of the dam, which was maintained at 37 ± 0.3°C, using a water blanket coupled to a temperature-controlled heating pump positioned around the abdomen of the dam.

Fetal MRI scans were performed for an initial baseline period. Following the baseline period, the balloon catheter was inflated within the aorta for 40 min, leading to uterine ischemia and global fetal hypoxia, thereby inducing the desired H-I within the fetal brains [9]. After the H-I period, the balloon catheter was deflated, initiating the reperfusion phase. MRI scans continued for an additional 20 min after the cessation of uterine ischemia, enabling the capture of dynamic changes during uterine reperfusion.

All MRI scans were acquired using a Siemens Magnetom Verio 3-Tesla research magnet using a knee coil (Siemens Healthcare, Erlangen, Germany). Single-shot fast spin-echo T2-weighted images were taken for anatomic reference in axial, coronal, and sagittal planes of the trunk of the rabbit dam, with 24–35 axial slices covering all fetuses inside the dam. Slice thickness was 4 mm, matrix 256 × 192, and field of view was 16 cm. Anatomic scans were followed by diffusion-weighted echoplanar images (DWI) with b = 0 and 0.8 ms/m², TR/TE = 7,400/70 ms, 2 averages, and the same slice geometry in 3 planes, as in reference anatomic images. To track the time course of H-I and the temporal variation in ADC value, we performed continuous DWI in axial plane with b = 0 and 0.8 ms/m², taking 2 averages, during 10 min of the baseline before H-I, 40 min of H-I, and 20 min of reperfusion-reoxygenation (R-R). Each DWI acquisition took approximately 2 min, depending on the body size of individual dam. The in-house scanner software provided all ADC images per each DWI acquisition for the subsequent analysis.

Following the MRI imaging session, the dams underwent one of two distinct procedures. In the first approach, immediate caesarean section (C-section) was performed. Alternatively, the catheter was removed, and the femoral artery was repaired, allowing the dams to recover post-surgery. Subsequent C-sections were performed on the dams at various gestational timepoints (E26, E27, E28, or E30).

The precise positioning of each fetus within the uterine horn and its corresponding fetal brain was achieved using 3D Slicer (version 5.2.2) [21]. To accomplish this, we utilized the “sphere brush” tool in segmentation editor. The diameter of the “sphere brush” was adjusted to between 3% and 5%, depending on the size of the fetal brain, and marked on the two-dimensional adjusted T2-weighted image. A different color was used to mark each fetal brain, and this was done in the axial, coronal, or sagittal plane, depending on the optimal view of the fetal brain. The three-dimensional MRI image was generated using the volume rendering function in Slicer, with the pre-set of “MR default.” Following this, adjustments were made to the “Scalar opacity mapping” under volume properties after synchronization with the “volume module.” These adjustments were made to create a clear three-dimensional image of the uterine horn with the fetuses. After obtaining the three-dimensional image, the fetal brains, marked with the sphere brush tool, were revealed on the three-dimensional MRI image using the “Show 3D” option in the segmentation editor. By cross-referencing the fetal positions in the uterine horn with the surgical records from the C-section procedures, we successfully determined the precise fetal localization.

The identification of fetal heads within the intensity-scaled and tissue contrast-adjusted T2-weighted images was done using an in-house MATLAB software (MathWorks, Natick, MA, USA). Briefly, in each dam, T2-weighted image was spatially registered to the first DWI scan of the DWI time series using a rigid body registration available in SPM 8 toolbox (https://www.fil.ion.ucl.ac.uk/spm/). Each DWI scan was independently processed to calculate its ADC map using Siemens Syngo VB17 Evaluation Dynamic Analysis software. All ADC maps were then registered to the prior ADC map for correcting inter-scan motion using the SPM 8 rigid body registration. To control the homogeneous quality of ADC map across DWI scans, two evaluators (G.A. and J.-W.J.) visually checked all registered ADC map fetus by fetus. Significant motion artifact was defined as x-y-z translation >2 mm which is a half of the voxel thickness existing between each pair of neighboring ADC maps, from the prior ADC to the present ADC. In the cases of significant motion, problematic registration errors were manually corrected by spatially matching three landmarks of individual fetuses (eyes and a center of the cerebellum) between the neighboring ADC maps. These landmarks were also applied to demarcate the fetal brain ROI at the same locations of individual ADC maps. Thus, the effect of motions on the changes in ADC was appropriately minimized in the ADC time series.

A spheroidal ROI was then meticulously defined with a radius to encompass the ADC values of the entire fetal brain in three orthogonal planes (e.g., 3–5 mm, depending on the size of brain). Additionally, a smaller ROI featuring a reduced radius was established to specifically pinpoint the left and right hemispheres of the fetal brain in three orthogonal planes. This enabled the precise calculation of ADC means within these specific regions.

Three-dimensional images were generated employing 3D Slicer (version 5.2.2) [21]. Axial (Fig. 1a), coronal (Fig. 1b), and sagittal (Fig. 1c) adjusted T2-weighted images were imported into the 3D slicer software platform. The “volume rendering” function was employed to construct a comprehensive three-dimensional representation of each T2-weighted image. The “MR default” option for rendering was chosen to ensure optimal visualization. The resulting rendered structure was synchronized with the volume module, and refinements were made to the scalar opacity mapping parameters to enhance overall image quality.

For each individual T2 weighted image, three distinct segments were established within the segmentation editor. These segments corresponded to the axial, coronal, and sagittal planes. The “scissors” tool was then utilized to define a rectangular region on each plane within the two-dimensional slices of interest. This marked the fetus within the volume rendered image, assigning a three-dimensional coordinate to indicate the position of its head (Fig. 1d–f).

To isolate the specific fetus of interest, the “Display ROI” function was implemented. This isolation enabled the generation of an enlarged view focused on the chosen fetus (Fig. 1g–i).

To achieve a comprehensive and detailed visualization of the fetal brain, we leveraged the segmentation editor to create a three-dimensional animation (Fig. 2) specifically focusing on the E25 fetal brain. During this iterative process, brain tissue was visually highlighted in white, while fluid-filled spaces were distinctly delineated in blue. The principal aim of this endeavor was to construct three-dimensional structures that underscore the importance of encompassing fluid-filled regions when designating the ROI for the calculation of ADC values. These fluid-filled spaces could potentially introduce an elevation in ADC values, necessitating their inclusion in accurate ROI delineation.

For subsequent investigations concerning left and right brain analyses, a strategic modification was introduced in the radius of the spheroid utilized to define the boundaries of the left and right cortical tissue. This tailored adaptation enabled the proportionate reduction of fluid-filled spaces contained within the ROI. By employing this nuanced approach, our intention was to achieve a more refined representation of ADC values, which would inherently align more faithfully with the distinct characteristics of the cortical tissue.

Three-D images of ADC MRI images were rendered in a manner akin to the adjusted T2 weighted images to obtain hemispheric ADC (Fig. 3). Of note, the axial section solely exhibited a sufficiently high resolution to enable rendering into a 3D view and facilitate ADC analysis between the left and right hemispheres of the fetal brain. As a result, we can obtain ADC values of the left, right hemisphere as well as the entire brain at each timepoint.

To determine the left or right handedness of naive rabbit kits at postnatal day 1 (P1), rabbit kits were gently placed on their back on a bench surface covered with pad and observed for spontaneous self-righting. A successful righting normally occurs within 30 s. Each kit completed 5–16 rounds of righting. Two independent examiners analyzed the video recordings and determined the number of successful left-sided or right-sided righting. To determine the handedness of newborn kits, we used the percentage of their response as a criterion. A total of 61 P1 naive rabbit kits from 8 litters were analyzed.

In each fetus, three Pearson’s correlation coefficients were obtained between ADC time series of the left hemisphere ROI and the right hemisphere ROI, left hemisphere ROI and whole brain ROI, and right hemisphere ROI and whole brain ROI. Also, we have presented means and 95% confidence intervals wherever possible to help the reader decide about trends and significance. Most of the statistical tests were done in SAS ver 9.4 (SAS, Cary, NC, USA). Both parametric tests including ANOVA [22], one sample proportion test, and z test for two proportions, and non-parametric test including Wilcoxon Rank Sum test [23] were utilized. Proc Univariate was performed at each timepoint to check for normality, kurtosis, skewness, and for performing the Wilcoxon Signed Rank test. For P1 naïve newborn kits, if the percentage of handedness response equaled or exceeded the 99% confidence limits, we identified their handedness as either left or right. Otherwise, we classified the kits as undetermined.

In our previous research [16, 17], we employed DWI-defined ADC patterns as a predictive biomarker for newborn hypertonia after H-I and R-R. By monitoring fetal brain ADC during uterine ischemia, we delineated four distinct patterns: pattern I indicated no alternation in ADC, pattern II revealed an ADC reduction but not surpassing 70% of baseline [20], pattern III exhibited an ADC decline below the 70% threshold, and pattern IV demonstrated a further decrease in ADC in the R-R period. The extent of the decrease of ADC then represents the extent of brain injury in the R-R period, which is abbreviated as RepReOx. The distinction between R-R and RepReOx is that R-R = the time period while RepReOx = extent of brain injury during R-R time period. RepReOx can be big or small depending on the area of ADC below that of the end ADC at end of uterine ischemia, while R-R, as time, does not change [16]. Among the 54 analyzed fetuses, the distribution of these patterns was as follows: pattern I was observed in 7% (n = 4) of cases, pattern II in 13% (n = 7), pattern III in 52% (n = 28), and pattern IV in 28% (n = 15).

We performed a targeted examination to ascertain the ADC time series in both the left and right hemispheres. We achieved success in 15 fetuses, constituting 28% of the initial population of 54 fetuses. Considering the typical litter size (averaging 8 fetuses) and the impact of uterine and intestinal motion, the resolution of ADC MRI faced limitations in acquiring 3D images of satisfactory quality.

We subsequently identified the patterns of ADC injury from the mean ADC of the entire brain, employing preestablished criteria from previous publications [16, 17, 20], which forecast the presence or absence of hypertonia. Examples of these patterns are illustrated in Figure 4.

We analyzed the correlation between ADCs of the left hemisphere versus right hemisphere, left hemisphere versus whole brain, and right hemisphere versus whole brain (see online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000539212). We observed a moderate positive linear correlation between left and right hemispheres (0.8 > r ≥ 0.4) in 53% (n = 8) of the fetuses and a strong positive linear correlation (1 > r ≥ 0.8) in 47% (n = 7) of the fetuses. Moreover, the ADC values of the left and right hemispheres displayed statistically significant correlation (p < 0.01) in all cases. In the relationship between the left hemisphere and the whole brain, 7% (n = 1) of the fetuses exhibited a weak positive linear correlation, 40% (n = 6) displayed a moderate positive linear correlation, and 53% (n = 8) showcased a strong positive linear correlation. All but one fetus demonstrated a significant correlation (p < 0.01) between the ADC values of the left hemisphere and the whole brain. In the relationship between the right hemisphere and the whole brain, 7% (n = 1) of the fetuses manifested a weak positive linear correlation, 67% (n = 10) exhibited a moderate positive linear correlation, and 26% (n = 4) showcased a strong positive linear correlation. Similarly, except for one fetus, all instances demonstrated a significant correlation (p < 0.01) between the ADC values of the right hemisphere and the whole brain.

We first divided the entire timeline into baseline, first half of H-I, second half of H-I, and R-R. When assessing the ordinal distinctions in the delta difference between left and right (L-R), a clear majority of brains exhibited a left-favoring trend. In this analysis, we categorized all positive values (i.e., >0) as left-favoring and all negative values (i.e., <0) as right-favoring, without considering the magnitude of the change from 0. The percentages of left and right favoring were then calculated. Ideally, the expected mean values for all percentages would be 50%, indicating no preference for either left or right. However, we observed a highly significant left-favoring trend both at baseline and throughout the entire timeline (Fig. 5). When examining the direction of change after H-I, there was a shift toward right favoring during this transition (Fig. 5). Thus, our method of spheroid analysis reveals a favoring of left hemisphere through the timeline from baseline to H-I and R-R.

However, when we determined the actual delta difference of left minus right hemispheric ADC in numerical terms, the means of delta ADC in these time epochs was not significantly different from zero (Fig. 6), because the mean of the absolute values of ADC of the left hemisphere was similar to that of the right. There was a tendency of the right-favoring brains to have a higher magnitude than the left-favoring brains (mean 0.0023, UCI 0.0187, LCI −0.015). This would suggest that fetuses favoring the right hemisphere, though fewer in number, manifest the favoring of ADC by a bigger magnitude.

The area under the curve for the mean ADC of the entire brain was then calculated. We then compared different patterns of injury (Fig. 7). We could only compare patterns II, III, and IV, as we only had one fetus in pattern I. Because of the definitions of the patterns of injury, Pattern III had expected decrease in the second half of H-I and pattern IV had a further decrease in R-R (Fig. 7).

We wanted to ask the question if there were any unilateral brain lesions in patterns predictive for hypertonia (i.e., III or IV). The actual delta difference of L-R hemispheres was then converted to area under the curve for each timepoint. There is a trend in pattern III for right-favoring differences (Fig. 8). However, pattern IV shows a slight tendency to favor the left hemisphere.

When we investigated individual delta L-R ADC values, there was no significant difference between the patterns of injury except at one timepoint (Fig. 9). At 33 min after start of H-I, pattern III shows right-favoring change clearly different from patterns II and IV (Fig. 9). This confirms the trend to right-favoring change of pattern III in the area under curve for L-R ADC as seen in Figure 8. These findings would suggest that different patterns of injury may manifest different preferences for the left or right hemisphere.

When we investigated sex differences, our population was found to be consisting of 4 males and 11 females. We excluded one female fetus due to an exceptionally low ADC at baseline and an inability to classify the pattern. Females tended to favor the left hemisphere more than males (Fig. 10).

Using 99% confidence intervals of the percentage of the response, we were able to categorize 26 kits (43%) as left-handed and 24 kits (39%) as right-handed, with 11 kits undetermined.

This study marks the first use of 3D spheroids in fetal MRI. Additionally, it represents the first exploration of laterality and unilateral lesions by defining the left and right hemispheres in fetal MRI within animal models. We successfully identified changes favoring the left and right hemispheres through dynamic ADC measurements. However, this advanced technique achieved success in generating satisfactory images with sufficient resolution in only 28% of the MRI attempts. The low percentage resulted from various technical challenges we faced, such as dealing with multiple fetuses in a confined space, motion artifacts, the limitations of timed ADC measurements, and interference from the intestines. It should be noted that achieving successful DWI and diffusion tensor imaging remains a challenge even in humans, with success rates from a quarter to four fifth of cases [24, 25] in fetuses. Some technological challenges, such as suboptimal coil sensitivity and motion-related artifacts, contribute to issues like low signal-to-noise ratio, signal loss, and spatial distortion in fetal brain situated in the outer region of the field of view, mirroring challenges encountered in rabbit studies.

What caught our attention in our study was that throughout the entire timeline, the left hemisphere’s ADC was significantly higher in the majority, with this being most noticeable during the baseline period (Fig. 5). This disparity could be attributed to the resting state, indicating either a higher ADC on the left or a lower ADC on the right. Factors like the partial volume effect, temperature fluctuations, variations in gestational age, or myelination (which typically occurs after birth) are likely not sufficient to account for this phenomenon. The predominance of females might partly account for the left hemisphere being greater than the right. Nevertheless, the observation of left hemisphere being favored during baseline (Fig. 5) remains unexplained, as males were not significantly different from females at baseline. It is noteworthy that the actual ADC did not exhibit any bias toward either hemisphere, but the ordinal score indicating left hemisphere favoring revealed a distinction. Not much is known about laterality in fetal life, and it was a bit unexpected to observe this at 79% gestation.

In normal Beagle dogs, a similar left hemisphere favoring was observed in actual ADC values [26]. On the other hand, other researchers have noted a trend for higher ADC on the right in cats [27] and dogs [28], though not reaching statistical significance. Intriguingly, H-I events seem to cause rabbit brains to favor the right side as a dynamic change (Fig. 5, 8,,9). In a comprehensive study involving 321 dogs, with 77% displaying neurological signs, no significant difference was, however, found between the left and right hemispheres [29]. The biological significance of these rabbit findings remains uncertain. In the baseline, they likely mirror subtle disparities in brain development between the left and right hemispheres. It is well established that increasing cellularity and neuronal maturation can influence ADC measurements [30]. There are various explanations for ADC differences [31]. Considering that ADC typically decreases with gestational age [30, 32], it is speculated that in rabbit fetuses, the right side might be more normally developed at that gestational age. Perhaps with H-I injury, there is now damage that reverses the ADC favoring among hemispheres.

A more intricate examination of the ADC patterns unveiled a consistent trend wherein the ADC mean values of the left and right hemispheres tended to be lower than those of the whole brain (data not shown for brevity). It is known that ADC values of cerebrospinal fluid in humans are greater than white and gray matter values [33]. It is thus possible that fluid-filled spaces, such as brain ventricles or subarachnoid spaces could explain the discrepancy (Fig. 2), as there is more cortical tissue within the defined ROI in the hemispheres contributing to the relatively lower ADC values observed in these regions. If that was the case, one would expect the variance of serial ADC measurements to be significantly higher in when calculating the mean of the entire brain compared to left or right hemispheres, but that is not the case (average, UCI, LCI of variance of mean entire brain 0.042, 0.051, 0.033 vs. left 0.037, 0.046, 0.028 vs. right 0.037, 0.045, 0.030).

Thus, analyzing hemispheric data could enhance the accuracy of predicting changes in cortical tissue and potentially provide insights into the development of handedness-related alterations. Determining the ADC in both left and right hemispheres opens the possibility of detecting the onset of unilateral strokes. If a stroke happens, a substantial reversal of ADC from low to high should be observed with complete cell lysis, as seen in adult strokes. However, the occurrence of this phenomenon was not noted in the fetal brains of rabbits, implicating the absence of unilateral strokes in the global hypoxia model. In humans, newborn strokes are uncommon, with an incidence of 1 out of 4,000 newborns [34]. One study suggests 76% of these cases stem from arterial ischemia [35]. It is plausible that we did not observe evidence of unilateral strokes because it was a rare event given the possibility that global hypoxia induces unilateral strokes at a comparable low rate in animal models. It is noteworthy that statistical analysis of population changes may fail to reveal an equal frequency of left and right infarctions. Some fetuses exhibited unilateral ADC favoring in pattern IV during both H-I and R-R, but not in pattern III, with an equal distribution between right and left (2 each). In conjunction with the suggestive trends depicted in Figures 8 and 9, these findings prompt the intriguing notion that RepReOx injury in R-R might preferentially affect a distinct hemisphere compared to sole H-I injury. If this hypothesis holds true, the implications are substantial, suggesting, for instance, that the etiology of free-radical injury in the fetal brain, forming the basis of pattern IV, may have an asymmetric component to the injury. It is crucial to highlight that serial ADC measurements during H-I provide a unique advantage in tracking dynamic changes in cell injury progression and provide a clearer picture of the dynamic changes instead of a single timepoint evaluation. Fetal brains may follow trajectories of either cell lysis or complete recovery and offering the advantage of identifying early injury patterns for further investigation into molecular mechanisms, as was done in a previous study [17].

Spastic cerebral diplegia is the most common clinical phenotype of CP [34]. In 111 diplegic children, 40.5% showed left hand dominance compared to 2.9% in 444 normal children [35]. It is unknown whether hemispheric asymmetries diagnosed by DWI may predict handedness. In the pilot study testing handedness in 61 P1 naïve rabbit newborn kits at P1, 43% were deemed to be left-handed and 39% right-handed. It would be interesting to further investigate the dynamic changes of MRI with the behavioral determination of handedness.

We would certainly have liked to conduct a much larger study, especially to include eventual survival to term delivery and to study correlation with handedness, but given the constraints of the costs of MRI and limited funding, we were unable to obtain more than 54 fetal MRIs or perform survival endpoints of handedness. This study has encouraged us to modify our DWI protocols or use of contrast reagents to correct some of the technological problems [36]. Improvements in automatic segmentation may obviate the necessity of manual placement of ROI in each MRI image, which at present takes a lot of time.

In conclusion, it is important to differentiate right and left-favoring brains. We introduce 3D spheroid analysis in rabbit fetal MRI which can differentiate between left and right hemispheres. This toolkit allows the detection of a favoring of the left hemisphere at baseline and through the timeline of H-I and R-R. In certain patterns of brain injury, there is a change favoring the right during H-I, which is influenced possibly by the sex of the fetus. This study underscores the importance of looking at the dynamic change from baseline to R-R to see the dynamic change of laterality instead of a single timepoint. There is a need for a much larger study to investigate the occurrence of laterality and handedness in fetal brain injury. Future research will explore whether improvements to the current toolkit, specifically in the automatic segmentation of the fetal hemisphere on an ADC image to further enhance the precision in placing a 3D sphere at the center of the segmented hemisphere, can successfully reduce the probability of type-II errors in calculating average ADC values from a given DWI, without relying on the user’s demarcation experience.

The authors gratefully thank Alexandra Lim, who helped in the video analysis of handedness.

This study protocol was reviewed and approved by the IACUC of Wayne State University, approval number (IACUC -19-06-1161).

The authors have no conflicts of interest to declare.

S.T. received NIH funding support (NINDS R01NS081936, R01NS114972, R01NS117146, and R01NS130258), and J.-W.J. received NIH funding support (R01NS089659). These funding sources were not involved in the study design, execution and analysis, and manuscript conception, planning, writing, or decision to publish.

Gaurav Ambwani: acquisition and analysis of data for the work, drafting the work, final approval of the version to be published, and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Zhongjie Shi: designed the study, acquisition of data for the work, reviewed the manuscript critically for important intellectual content, final approval of the version to be published, and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Kehuan Luo and Jeong-Won Jeong: acquisition of data for the work, reviewed the manuscript critically for important intellectual content, final approval of the version to be published, and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Sidhartha Tan: designed the study, drafting the work, final approval of the version to be published, and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

The data that support the findings of this study are not publicly available due to IACUC regulations but are available from the corresponding author upon reasonable request.