Morphometric Shape Analysis of Corpus Callosum in Children With Down Syndrome

Down syndrome (DS) is characterized by varying degrees of mental retardation and delay in neurocognitive functions. Herein, we analyzed the morphometric shape of the corpus callosum (CC) in children with DS. Twenty-three DS cases underwent magnetic resonance imaging and have grossly normal CC, and 23 control group cases were included in this retrospective study (2012-2020). The CC was obtained from T2-weighted mid-sagittal images, and certain anatomical points were marked on the CC. Statistical geometric shapes and deformations of CC were evaluated for both groups. The age range of patients with DS and control group was 6 to 42 months. A statistically significant difference was found in the shape of CC between the groups (P < .001). Deformation was most evident in the splenium in the DS group.


Introduction
Some of the studies in medicine are related to examining the geometrical properties of an organ or organism.Commonly, quantitative or qualitative data sets in statistical analysis consisted of measuring values; today's organ or organism's appearance or shape began to be used as input data by the development of imaging techniques.In recent years, there has been considerable interest in the statistical analysis of shape.The field of geometric morphometrics is known as statistical shape analysis applications based on landmarks. 1Coordinatebased methods are also referred to as geometric morphometrics because they retain all geometric knowledge from data collection to analysis and visualization. 2 Statistical shape or image analysis, used to analyze 2-dimensional or 3-dimensional (3D) shape data, has become widely attractive in medicine and biology.The key reasons for the growing use of statistical shape analysis in medicine are advances in imaging technology and a desire to learn more about the impact of diseases and environmental influences on organ or organism structure. 1own syndrome (DS; Trisomy 21) is the most common chromosomal condition occurring in 1 out of 800 live births worldwide.It is associated with intellectual disability and is characterized by varying degrees of developmental delays in cognitive functions such as speech, memory, and learning in childhood. 3Corpus callosum (CC), which is the largest white matter pathway connecting the cerebral hemispheres, has both a connection and integration role for neurocognitive function information. 4The CC's primary function is to exchange information and coordination between the hemispheres.It is divided from anterior to posterior into four parts: the rostrum, genu, corpus, and splenium.The genu is at the front, and the splenium, the thickest part, is at the back. 5[10] Although these cognitive dysfunctions exist from birth in individuals with DS, there is only one study on CC morphology in individuals in the pediatric age group, 7 and CC was evaluated volumetrically in this study.However, there is no study examining the relationship between the shape of CC and DS.Shape analysis is more effective than volume measurement methods and can provide more specific information about anatomy.For example, two different geometric shapes may have the same volume but completely different shapes. 11his study aimed to investigate the shape differences in the normal-appearing CC of DS patients compared to that of healthy controls.

Patients
The brain magnetic resonance imaging (MRI) data of 36 DS patients aged 18 years or younger who were admitted to the department of radiology between 2012 and 2020 and underwent cranial imaging for reasons such as headache, seizure, suspected seizure history, and developmental delay were retrospectively examined.Seven patients with concomitant intracranial pathology, atrophic or degenerative changes in the CC, delayed myelination in white matter compared to their peers, and whose CC could not be evaluated due to poor examination quality were excluded.Since the CC reaches optimal myelination and thickness in approximately 4 to 6 months of the neonatal period, 12,13 4 patients younger than 6 months were excluded as well.The karyotype analysis results of all children in the patient group included in the study were confirmed as Trisomy 21 from the medical archive.Accordingly, 25 patients whose intracranial structures and CC were evaluated as relatively normal were included in the study.The age of 23 of the patients was 3.5 years or younger.Two patients, 13 and 17 years old, were excluded from the study, as this would disrupt the homogeneity regarding age.Inclusion and exclusion criteria are shown in Diagram 1.As the control group, 23 children of similar age and gender, who underwent cranial MR imaging for other reasons (headache, seizure, and suspected seizure) and whose examinations were normal, were selected.The age range of the DS group was 6 to 42 months (11 female, 12 male), with a median value of 16 months, and the age range of the control group was 6 to 42 months (10 female, 13 male), with a median value of 15 months.

MRI Examination
MRIs were performed in a 3.0-Tesla MRI Device (Achieva; Philips, Best, the Netherlands) with a standard head coil.The scanner alignment tool and immobilization of the head helped to ensure the patient's standardized position.The sagittal T2-weighted spin-echo sequence images (Field of view [FOV], 230 × 250 mm; section thickness, 5 mm; time to repeat/time to echo (TR/ TE), 5800-6700/95-115 milliseconds; flip angle, 90°) were used for subsequent morphometric analysis.

Two-Dimensional Landmarks Collection
A single radiologist used TpsDig version 2.30 software to identify and mark the anatomical landmarks.On the CC, 16 anatomical landmarks were described and chosen as landmarks for the CC, as stated by Sigirli et al. 14 Figure 1 shows the landmarks used for CC on an MRI image of a subject.

Statistical Analysis
The Shapiro-Wilk test determined if the variables followed a normal distribution.Since the data did not fit Procrustes analysis was used to compare CC shapes between the groups.Box's M test determined if the covariance-variance matrices were homogeneous.Since the variance-covariance matrices were not homogeneous, James's Fj test was used.The root means square of Kendall's Riemannian distance rho to mean shape was used to obtain an overall measure of shape variability.
Thin-plate spline (TPS) analysis evaluated the shape deformations.For TPS analysis, the Procrustes mean shapes were determined.The areas with the greatest reductions or enlargements were labeled in different colors to suggest deformations based on the TPS analysis results.Principal coordinate analysis was applied to tangent coordinate, which was derived from Procrustes analysis.Euclidean distance was used in the principal coordinate analysis.
Relationships between CC shape-size and age were analyzed with the Spearman correlation coefficient.Shape-size was considered centroid.

Landmark Reliability
Based on the generalizability theorem (GT), we measured the interrater reliability coefficient for a 2-facet crossed design (landmark pairs by rater and by subject). 15he generalizability coefficient is used in the GT to describe the reliability of relative (norm-referenced) interpretations. 16A single rater identified the anatomical landmarks in this research.Using repeated landmarks on groups, the rater's reliability was evaluated.A single investigator collected landmarks on CC, and after a month, the same investigator re-marked the same landmarks on the same 46 subjects (23 cases and 23 controls).The study revealed that CC (G = 0.9978) has good repeatability.The landmark reliability was calculated with the following below: http://biostat.home.uludag.edu.tr/landmark_reliability/G_coefficient.html.

Results
There was no statistically significant difference between the control and patient groups regarding age and gender (P = .869and P = 1.000, respectively) (DS group comprised 11 female and 12 male patients, and the control group comprised 10 female and 13 male participants).The mean age in the DS group was 16 months (6-42 months), and 15 months (6-42 months) in the control group.
According to the CC shape, a statistically significant difference between the control and patient groups (P < .001) was observed.Procrustes mean shape graph is presented in Figure 2. The shape variability in CC in the control and patient groups were .099and .103,respectively.The patient and control groups' CC shape variability was similar.
When the change from control to the patient was evaluated, a remarkable deformation (change) in the region of the splenium-level landmarks was observed according to the TPS analysis results compared to the other regions (Figures 2 and 3).
When the units were evaluated in terms of CC shapes according to the first three coordinates obtained from the principal coordinate analysis, there were differences in the CC shapes of patients and controls.Based on the classification according to the convex hull graphs drawn in the first two coordinates with the highest explanation rate for both control and patient cases (Figure 4A), 54.34% (n = 25) of the patients and controls overlapped, while 45.66% (n = 21) did not.It was observed that 58.7% (n = 27) overlapped and 41.3% (n = 19) did not in the first and third coordinates (Figure 4B), and in the second and third coordinates (Figure 4C), 76.09% (n = 35) overlapped and 23.91% (n = 11) did not.
When the relationship between CC shape-size and age was assessed, there was a statistically significant relationship between the centroid size and age in months in the control group (r = .561,P = .005),but no significant relationship was determined in the patient group (r = .275,P = .204).
The expansion and contraction factors at the landmarks were numerically displayed in the TPS analysis (expansion factors larger than 1 and contraction factors smaller than 1).

Discussion
Understanding the neuropathological changes underlying the intellectual disability and neurological features in individuals with DS is critical.Although these changes in the brain begin from the fetal period, 17 neuroimaging studies, especially in early childhood, include a small patient cohort.Our study performed statistical shape analysis for CCs in DS cases during infancy and early childhood that had grossly normal CC morphology.This is the first modern morphometric shape analysis study to evaluate CC in DS patients in this age group.
The CC development commences with the development of the genu at about the 12th week of the intrauterine period, followed by the isthmus and splenium, and finally, the rostrum around the 18th to 20th weeks of gestation. 18,19The maximum number of fibers pass through the CC in the intrauterine period, and myelination of these fibers continues in the postnatal period. 20In the third month of the newborn period, the genu and the splenium are myelinated and thickened in the fourth to sixth month, 12,13 and this causes the CC to increase in size with age.When the size of the CC was examined, we found a statistically significant increase, as expected, with increase in age in the healthy group.However, we could not find a relationship between age and CC size in the DS group.In the literature, the first sign of neuroanatomical abnormality is a characteristic reduction in brain size, and an increase in progression in the last 3 months of pregnancy is observed in fetuses with DS. 21,22 Parallel to this, in our study, CC did not show the expected development with age in patients with DS in the postnatal period and fell behind.
Studies in the literature on children and youth with DS, in which MRI evaluated structural changes in the brain, primarily focused on volume analysis.In these studies, reduction in total brain volume, [23][24][25][26] decrease in cerebellar [23][24][25]27 and hippocampal 24,26,27 volumes, and regional reductions in frontal and temporal lobes 25,26 were reported, whereas parietal lobe gray matter, thalamus, and basal ganglia volumes 24 were maintained.However, there is no evaluation of CC in these studies. In the pdiatric age group, only in the study of Gunbey et al, 7 CC volume decrease in individuals with early childhood DS was reported, which reveals the loss of neocortical neuronal projections involved in the maintenance of higher cognitive processes.In our study, unlike volume studies, we aimed to determine whether the shape of CCs, which were evaluated to have qualitatively normal morphology, is also quantitatively normal and if there is any abnormality and its location.Variations in the morphology of CC can be seen in normal population. It may aear tubular due to the absence or mild narrowing of the body or bulbous with a marked enlargement of the splenium.We excluded individuals with thinning and agenesis in the CC when planning the study.We found a statistically significant difference between the healthy and patient groups regarding the CC shape.A shape difference was quite evident,  especially at the splenium level.The splenium is the bulbous-shaped part of the CC at the most posterior part.The fibers in the splenium are projections from the occipital-parietal and temporal cortex. Although the function of the splenium remains unclear, an increase in the size of the splenium during adolescence could facilitate the maturation of multiple higher functions such as language, reading, computational skills, intelligence quotient, behavior, and consciousness, which require visuospatial information transfer, and can allow these functions to develop.[29][30][31] The retardation in these functions, which may occur in individuals with DS starting from the neonatal period, might be due to dysmorphic changes that are more clearly detected in the splenium part of the CC.In the study by Gunbey et al, 7 while there was no statistically significant decrease in total and segmental brain volumes, hippocampal volume, and white matter volumes, a significant decrease in CC volume may be a finding that may support this observation.That is, while findings such as a decrease in total brain volume, especially diffuse cortical atrophy including parietal lobes, symmetrically sulci, and hippocampus in adults with DS, 32,33 are not yet seen in early childhood, the first deteriorations in the brain may start from the CC, especially from the splenium part of the CC.A theory that could explain this deformity might be that myelination deficiency or disorders in axons can be detected earlier in the CC, which has a more compact structure than the white matter and especially in the splenium, which is the thickest part of the CC.
The study's limitations include the small patient cohort and recruitment from a single center.Also, a technical limitation was performing morphometric analysis on single-plane (sagittal) T2-weighted images.The CC's landmarks and actual size could be better evaluated with a 3D sequence.However, this was not possible due to the retrospective nature of the study.
To the best of our knowledge, this is the first statistical shape analysis study to evaluate CC in individuals with DS in early childhood.We think that the results of our study and other findings in the literature may contribute to the elucidation of how neurocognitive disorders seen in individuals with DS cause changes in the brain in the early stages.Besides, the results we obtained can be a crucial reference source for creating a database of artificial intelligence programs that we will be using more soon.However, the findings should be supported by other studies that are more comprehensive and include a larger number of patients.Also, more comprehensive studies that include other anatomical reference points in the brain will contribute to the field.

Figure 1 .
Figure 1.The corpus callosum landmarks are visible in this mid-sagittal slice.

Figure 3 .
Figure 3.A thin-plate spline demonstrating the average corpus callosum shape deformation from controls to patients.

Figure 2 .
Figure 2. Procrustes mean shapes for the corpus callosum images of Down syndrome cases and controls (controls: □; cases: +).