Correlating MRI-Based Brain Volumetry and Cognitive Assessment in Down Syndrome Patients

INTRODUCTION: Down syndrome (DS) is the most common genetic cause of intellectual disability. Here, we use magnetic resonance imaging (MRI) on children and adults with DS to characterize changes in the volume of specic brain structures involved in memory and language and their relationship to features of cognitive-behavioral phenotypes. METHODS: Thirteen children and adults with the DS phenotype and 12 age- and gender-matched healthy controls were analyzed by MRI and underwent a psychological evaluation for language and cognitive abilities. RESULTS: The neuropsychological prole of DS patients showed decits in different cognition and language domains in correlation with reduced volumes of specic regional and subregional brain structures. CONCLUSIONS: The memory functions and language skills affected in our DS patients correlate signicantly with the reduced volume of specic brain regions, allowing us to understand DS's cognitive-behavioral phenotype. Our results provide an essential basis for early intervention and the design of rehabilitation management protocols.


Introduction
As the most common genetic cause of mental retardation, DS affects approximately one in every 700 live births (Parker, Mai et al. 2010). Although intellectual disability is the most well-known feature of the DS cognitive-behavioral phenotype (Gibson 1978), research has indicated a pro le of relative strengths and weaknesses that characterizes children with DS compared to younger, typically developing peers matched on the developmental level and same-age peers with other forms of intellectual disability (Fidler, Most et al. 2008). Those with DS have de ciencies in several domains of language functioning that outweigh their overall cognitive limitations. Furthermore, di culties with explicit memory are common and outnumber global impairments (Jarrold, Baddeley et al. 2007, Hamner, Udhnani et al. 2018).
To date, neuroimaging studies of children with Down syndrome development have limits due to low resolution, small group size, and limited age range (Pinter, Eliez et al. 2001).
A better understanding of the developing brain with DS using a developmental approach will illuminate important neurological principles of DS in children and shed light on the foundations of adult phenotypes, particularly the increased risk of early Alzheimer's disease (Zigman and Lott 2007).
Thus, the current study seeks to ll this gap and provide new data on structural neuroimaging of DS children compared to children in the context of normal brain development and cognitive-behavioral phenotype. These de cits in DS-related research are probably due to problems attributable to the successful execution of development-oriented neuroimaging.
However, it is still unclear why previous reports did not adopt a similar method for the neuroimaging of children with DS as in other cognitive impairment-related neurodevelopmental disorders (e.g., Fragile X syndrome, Williams's syndrome, microdeletion syndrome 22q11), although such studies used modern imaging technologies and analyses and could expect common di culties with imaging such groups (Hamner, Udhnani et al. 2018). To make progress in this eld, there is a need for a boost in increased neuroimaging research, especially with a focus on early childhood, using advanced neuroimaging techniques and best practice guidelines for neuroimaging in children (Raschle, Zuk et al. 2012). The study depends on the hypothesis that children and adults with DS have language and memory problems as they age. These problems are related to brain development, particularly neuroanatomical changes associated with these behavioral changes. To test these hypotheses, we will examine the relationship between different brain regions and cognitive scores for memory and language in children and adults with DS. Therefore, our study will analyze the volumes of the hippocampal areas involved in explicit memory de cits (dentate gyrus, Ammon's horn, and subiculum).
Furthermore, we will consider structures such as the superior temporal gyrus and temporoparietal junction structures underlying impaired speech performance (e.g., angular gyrus, supramarginal gyrus, and occipitotemporal structures (e.g., fusiform gyrus)) (Hickok andPoeppel 2004, Friederici andGierhan 2013) and characterize them in more detail using neuroimaging with volume analysis. We used highresolution MRI acquisition techniques and advanced segmentation and image processing protocols to obtain more accurate quantitative image data in children with DS. We studied brain areas that previous studies have not considered, such as some regions in the temporal and parietal lobes, or not adequately investigated as hippocampal and parahippocampal gyrus areas. By studying these areas of interest radiologically and comparing their volumes with the psychological results, we can approach language and memory problems in people with DS and provide a basis for further research, including early interventions. Furthermore, the study of the early stage of AD as DS can be a model for understanding AD pathology and clinical features at an early age.
In general, we compared the imaging data on brain areas between different ages of DS and between the different genders in this group. We also compared DS samples and normal controls. We correlated neuroanatomical data and neuropsychological scores of DS samples based on age and sex between DS and normal controls. We described and analyzed the volumes of different brain regions in detail, focusing on areas associated with language and memory. We examined areas such as the temporal lobes, parietal lobes, and hippocampus entirely and subregionally. The focus has also been on some regions, such as the superior temporal gyrus, angular gyrus, supramarginal gyrus, fusiform gyrus, and hippocampal subregions (Dentate gyrus, Ammon's horn, and Subiculum).
Some regions, such as the corpus callosum, are also of interest but will not be explored here and will be considered in future studies by the authors.

Subjects
We recruited our DS participants from Down syndrome care centers in Sudan and selected control samples from our DS sample family members and hospital records. Thirteen children and adults with DS (eight males and ve females, mean age=15 years, SD=5.9, range=6.0-25) and twelve healthy controls (eight males and four females, mean age=14 years, SD=6.8, range=4.0-25) underwent MRI scans at Aliaa Specialist Hospital, Khartoum, Sudan. All DS samples and four control samples were assessed for IQ (intelligence quotient), including working memory by applying the IQ test (The Stanford Binet Scale, the fth picture of intelligence) and for language by using a language test (Luttas Language Development Test Scale) by a consultant medical psychologist and speech rehabilitation specialist at the Women & Child Health Development Organization, Khartoum, Sudan. Additionally, we designed parents' questionnaires to assess the psychosocial status and related clinical problems of the family and the participant and evaluated the cognitive level and the psychosocial risk factors that affect the participant's cognition. DS subjects were assessed clinically by a pediatrician to exclude cardiac defects and any contraindications for sedation. The diagnosis of DS was established at birth or in early infancy and childhood by clinical examination with one karyotype diagnosis case. We explained all the procedures in detail to all subjects and parents before enrollment in the study. Written informed consent was obtained from all parents and, when possible, orally by subjects before participation. This study was reviewed and accepted by the Ethics Committee of the Faculty of Medicine -University of Heidelberg, Heidelberg, Germany and National Health Research Ethics Committee, Ministry of Health, Khartoum, Sudan.

Imaging
Magnetic resonance imaging (MRI) is a radiographic sectional imaging technique that allows the imaging of different levels of the body with high soft-tissue contrast. In contrast to computed tomography or conventional X-ray imaging, MRI is based on a different magnetization of the body by the magnetics of MR tomography. MRI does not include X-rays so that children can obtain a safe examination. MRI is a very patient-friendly method that generally requires no special preparation. With the latest generation devices using a magnetic eld strength of 1.5 Tesla (T) and 3.0 T, the entire body can receive the examination in less than 20 minutes. The most used MR scanners are either 1.5 T (Tesla) or 3 T systems. Both systems allow the quanti cation of global and regional brain structures. However, 3 T systems offer an increased resolution of the contrast agent between the tissues (i.e., increased visualization of the boundaries between gray matter, white matter, and cerebrospinal uid (CSF)). MR scans on 1.5 T systems are su cient to quantify relatively small brain structures such as the hippocampus (Keller and Roberts 2009).
In Sudan, a 3D structure MRI technique with a 1.5 T Siemens scanner is common in Khartoum hospitals, 3D Slicer 2.6 is available for volume measurement.

MRI Protocol
Subjects had MRI scans using a 1.5 Tesla Siemens, Syngo, MR system at the Radiology and Medical Imaging Department, Aliaa Specialist Hospital, Khartoum, Sudan. Some children have sedation to be stable and restrained from movement. The whole brain has a scan with a 3d T1 space sequence in a sagittal plane, with slice thickness: 1mm, TR: 550, TE: 8.5, AVERAGE: 1, matrix: 256×256×204.8, time: 4:02 min.
Measurement and segmentation of total brain volume and hippocampal volume To measure the volumes of different brain regions, we used the volBrain online system (http://volbrain.upv.es), an online MRI brain volumetry system that provides free automated brain analysis and segmentation for different brain structures in a short time with accurate and detailed results. To use this system, users should rst register by providing personal information such as the email address, name, and name of the institution to which they belong. The user submits a single anonymized compressed MRI T1w Nifti le in a web interface to analyze the imaging data, and the webserver accepts requests. After approximately 12 minutes, the results will be ready and downloaded as a pdf le and received via email (Manjón and Coupé 2016). We used two pipelines available on the online system to measure our data: 1) the HIPS pipeline (see gure 1a and 1b), which is a pipeline for automatic hippocampal sub eld segmentation from monospectral (T1) images using the Kulaga-Yoskovitz segmentation protocol (Kulaga-Yoskovitz, Bernhardt et al. 2015), and 2) the vol2Brain pipeline (see gure 2a and 2b), which provides automatic brain segmentation dividing the volume into 135 structures. It also provides tissue, macrostructure, and lobe segmentations as well as cortical thickness. Researchers have compared this brain volumetry system with other software packages that provide subcortical brain segmentation. They found it more reproducible and accurate. Therefore, it can be considered one of the rst few platforms that offer hippocampal segmentation, which will help diagnose and study AD (Manjón and Coupé 2016 (Roid and Pomplun 2012). It is an IQ test that measures ve cognitive abilities in both nonverbal and verbal formats: uid reasoning, knowledge, quantitative reasoning, visuospatial processing, and working memory. Because it is effective with so many diverse groups independent of gender, race, culture, religion, area, or socioeconomic level, the SB-5 is considered one of the most extensively utilized intelligence tests.

Luttas Language Development Test Scale
The goal of the test is to assess the child's language development level and extract the child's expressive linguistic age and the child's receptive linguistic age. It can de ne the child's language weaknesses and strengths and develop an appropriate rehabilitation program for language development for each child separately. This test is available for use on an individual basis by assessing the child's capacity to recognize the internal language, name and identify implicit groups, understand, and express object functions, understand and express linguistic context, and identify expression only for the melodic and pragmatic structure. The test is available in Arabic and developed in Egypt but has been found to be suitable for use in other countries, such as Sudan. The test measures expressive language, receptive language, and overall language scores.

Data Analyses
We carried out this investigation as a pilot study. Therefore, we conducted a purely exploratory data evaluation, and the p-values obtained are interpreted purely descriptively and have no con rmatory value.
All data collected were analyzed using descriptive statistics (indicating absolute and relative frequencies or mean and standard deviation). All volumetric data met the requirement for different parametric tests, and normality was con rmed with Kolmogorov-Smirnov and Shapiro-Wilk tests and variance homogeneity with the Levene test. We performed descriptive statistics and analysis of variance (ANOVA) to compare total brain volumes (TBVs) between our study groups and separately to compare related neuropsychological test scores between groups. Analyses of covariance (ANCOVA) were performed for all comparisons while controlling for total brain volume. We employed paired t-tests and repeated measures analysis of variance to observe any signi cant changes in structural brain region volumes over time. To explore the correlation between age and brain region volumes (TBV, hippocampus, parietal lobe, temporal lobe, superior temporal gyrus, parahippocampal gyrus, angular gyrus, supramarginal gyrus, and fusiform gyrus) and between the same regions and related neuropsychological test scores, we used the Spearman correlation test because the scores were not normally distributed. We used partial correlations to assess this relationship while controlling for age and total brain volume. We carried out regression analyses for each brain region volume with the scores as predictors. We de ned a signi cant difference to have a P value less than 0.05 for two-tailed tests. We conducted statistical calculations with IBM SPSS STATISTICS for Windows Version 28.0.0.0 (190).

Neuroanatomy of DS
As shown in Table 1 and Figure 1, the DS samples' mean total brain volume was 20% smaller than that of the control samples (F=3.  Tables 2  and 4) were observable. However, some areas, such as the parietal lobe (r= -0,54, p= 0.06) and parahippocampal gyrus (r = -0,52, p= 0.07), approached signi cance and correlated negatively with age in the DS group when partially correlated while controlling for total brain volume (see table 5). As presented in Table 3, we compared regional brain volumes in males and females of both groups. We found more signi cant brain region volumes in males with DS than females in regions such as the total brain, cerebrum, frontal lobe, parietal lobe, temporal lobe, occipital lobe, and parahippocampal gyrus. The control group indicated no signi cant difference.

Neuropsychology of DS
Language test scores are shown in Table 6, which indicates that the total language score in the DS sample was 40% lower than that in the control group (F=8.2, df=16, p <0.001). The expressive language score in the DS group was 50% lower than that in the control group (F=7.6, df=16, p <0.001). The receptive language score in the DS group was 35% lower than that in the control group and approached signi cance in the difference between the two groups (F=3.3, df=16, p =0.08). Table 7 presents Stanford Binet Intelligence Scale (Fifth Edition) scores in the DS and control groups. These scores included total IQ, which was 36% lower in the DS group than in the control group (F=51.7, df=16, p <0.001). The total IQ has two divisions: Non-Verbal IQ, which is low in DS by 32 % than the control group (F=28.1, df=16, p <0.001), and verbal IQ, which is also low in DS by 40 % than the control group (F=55.0, df=16, p <0.001). Other test scores such as uid reasoning is low in DS by 25 % than control (F=12.3, df=16, p <0.001), knowledge is low in DS by 35 % than control group (F=20.1, df=16, p <0.001), quantitative reasoning is low in DS by 36 % than control group (F=39.1, df=16, p <0.001), visuospatial processing is low in DS by 39 % than control group (F=47.5, df=16, p <0.001) and working memory which is also low in DS by 37 % than control group (F=35.7, df=16, p <0.001). We compared language test scores in males and females of the DS group and control group, as shown in Table 10, and found that the mean total language score in males of DS was lower than that in females compared to the control group. The mean scores of expressive and receptive language were also lower in males with DS than in females but showed no difference from the control group. The Stanford Binet Intelligence Scale (Fifth Edition) scores in males and females of the DS group and control group, as shown in Table 11, indicated no signi cant difference between males and females of DS. However, the DS males have low scores than females. There was a considerable difference between males and females in the control group, and males had lower scores than females.

Association between neuroanatomy and neuropsychology of DS
Using the Spearman correlation test, we correlated DS samples' total brain volume and IQ, working memory, and total language scores. We observed a signi cant correlation between the whole brain and working memory (r = 0.68, p< 0.001), and the correlation between the total brain and total IQ approached signi cance (r = 0.53, p = 0.06). The correlation between the whole brain and total language was not signi cant.
We also correlated parietal lobe volume with total IQ, working memory, total language, and visuospatial processing scores for DS samples. We discovered a signi cant correlation between the parietal lobe and working memory (r = 0.62, p< 0.001) and visuospatial processing (r = 0.55, p< 0.001). The correlation between the parietal lobe and total IQ approached signi cance (r = 0.49, p = 0.08). We observed no signi cant correlation between the parietal lobe and total language.
We correlated temporal lobe volume and total IQ, working memory, whole language, and visuospatial processing scores for DS samples. We found a signi cant correlation between the temporal lobe and working memory (r = 0.68, p< 0.001), the correlation between the temporal lobe and total IQ approached signi cance (r = 0.48, p = 0.09) and no signi cant correlation between the temporal lobe and whole language and visuospatial processing.
We correlated hippocampal volume and whole IQ, working memory, and whole language scores for DS samples. We discovered a signi cant correlation between the hippocampus and whole language (r = 0.59, p< 0.001), and the correlation between the hippocampus and working memory approached signi cance (r = 0.51, p = 0.07). The correlation between the hippocampus and total IQ was not signi cant.
We correlated parahippocampal gyrus volume with whole IQ, working memory, and global language scores for DS samples. We observed a signi cant correlation between the parahippocampal gyrus and working memory (r = 0.57, p< 0.001) and no signi cant correlation between the parahippocampal gyrus and total IQ and total language.
We correlated superior temporal gyrus volume and global IQ, working memory, and whole language scores for DS samples. We observed a signi cant correlation between the superior temporal gyrus and working memory (r = 0.57, p< 0.001) and no signi cant correlation between the superior temporal gyrus and total IQ and total language.
We applied partial correlations between superior temporal gyrus volume and nonverbal and verbal IQ and expressive and receptive language scores for DS samples while controlling for age and total brain volume. We found a negative correlation between the superior temporal gyrus and expressive language that approached signi cance (r = -0.58, p = 0.07), while the correlation between the superior temporal gyrus and other scores was insigni cant.
We correlated angular gyrus volume with total IQ, working memory, and whole language scores for DS samples. We observed that the correlation between the angular gyrus and working memory approached signi cance (r = 0.55, p = 0.05) and no signi cant correlation between the angular gyrus and total IQ and total language.
We correlated DS samples' supramarginal gyrus volume with global IQ, working memory, and whole language scores. We observed no signi cant correlation between the supramarginal gyrus and any of these scores.
We applied partial correlations between supramarginal gyrus volume and nonverbal and verbal IQ and expressive and receptive language scores for DS samples while controlling for age and total brain volume. There was a signi cant negative correlation between the supramarginal gyrus and expressive language (r = -0.71, p< 0.001). The correlation between the supramarginal gyrus and other scores was not signi cant.
We correlated fusiform gyrus volume and total IQ, working memory, and whole language scores for DS samples. We observed a signi cant correlation between the fusiform gyrus and global IQ (r = 0.60, p< 0.001) and working memory (r = 0.73, p< 0.001) and no signi cant correlation between the fusiform gyrus and total language.
We correlated fusiform gyrus volume and uid reasoning, knowledge, quantitative reasoning, and visuospatial processing scores for DS samples. We observed a signi cant correlation between the fusiform gyrus and visuospatial processing (r = 0.65, p< 0.001). The correlation between the fusiform gyrus and other scores was not signi cant.
We correlated DS samples' hippocampal subregion (dentate gyrus) volume and total IQ, working memory, and whole language scores. We discovered a signi cant correlation between the dentate gyrus and whole language (r = 0.64, p< 0.001) and no signi cant correlation between the dentate gyrus and working memory and total IQ.
We correlated dentate gyrus volume with uid reasoning, knowledge, quantitative reasoning, and visuospatial processing scores for DS samples. We found no signi cant correlations between the dentate gyrus and any of these scores.
We correlated DS samples' hippocampal subregion (Ammon's horn) volume and total IQ, working memory, and whole language scores. We observed that the correlation between Ammon's horn and global IQ approached signi cance (r = 0.50, p = 0.07) and that there was no signi cant correlation between Ammon's horn and other scores.
We correlated Ammon's horn volume and uid reasoning, knowledge, quantitative reasoning, and visuospatial processing scores for DS samples. We found no signi cant correlation between Ammon's horn and any of these scores.
We correlated hippocampal subregion (subiculum) volume and total IQ, working memory, and whole language. We found a signi cant correlation between the subiculum and global IQ (r = 0.59, p< 0.001) and working memory (r = 0.61, p< 0.001) and no signi cant correlation between the subiculum and total language.
We correlated subiculum and uid reasoning, knowledge, quantitative reasoning, and visuospatial processing. We found a signi cant correlation between the subiculum and visuospatial processing (r = 0.62, p< 0.001), that the correlation between the subiculum and quantitative reasoning approached signi cance (r = 0.49, p = 0.08) and that there was no correlation between the subiculum and other scores.

Discussion
We investigated regional brain volumes in children and adults with DS to identify the pathophysiological mechanisms related to their cognitive features. We collected imaging data from the participants with DS and the controls and correlated the MRI data of the DS group with cognitive functions evaluated using neuropsychological battery assessments. We applied different neuropsychological tasks to investigate cognitive domains, such as global cognition, memory, and language, and we correlated the results to various regional brain volumes.
Our results con rmed that DS subjects had reduced total brain, cerebrum, cerebellum, brainstem, hippocampus, frontal lobe, parietal lobe, temporal lobe, occipital lobe, and parahippocampal gyrus volumes compared to controls.
These ndings, except for parahippocampal gyrus volume, are consistent with the results of previous neuropathological and neuroimaging studies (Jernigan, Bellugi et al. 1993, Kesslak, Nagata et al. 1994, Raz, Torres et al. 1995, Aylward, Habbak et al. 1997, Pinter, Eliez et al. 2001, White, Alkire et al. 2003, Menghini, Costanzo et al. 2011, Mullins, Daly et al. 2013, Hamner, Udhnani et al. 2018. In contrast to the ndings by Kesslak, Nagata et al. (1994) and Raz, Torres et al. (1995) of larger parahippocampal gyrus volume in DS subjects compared to controls, our study showed smaller parahippocampal gyrus volume in DS than controls. Furthermore, our results indicated reduced white matter and grey matter of the total brain, hippocampal subregions (Dentate gyrus, Ammon's horn, and Subiculum), angular gyrus, supramarginal gyrus, fusiform gyrus, and superior temporal gyrus volumes in DS participants compared to controls. Our study did not indicate any age-related changes in brain areas in either the DS or control groups. According to our results, DS males have more signi cant volumes of different brain regions than females compared to controls.
The neuropsychological pro le of DS patients showed de cits in different cognition and language domains in our study group. Our study results con rmed the ndings of previous studies that impairments in expressive language are more remarkable than de cits in receptive language (Abbeduto, Pavetto et al. 2001, Pulina, Vianello et al. 2019).
Our study group showed a mean IQ of 65, which is consistent with previous studies that con rmed that most people with DS have an IQ between 30 and 70 (Abbeduto, Warren et al. 2007). Our DS group showed a higher nonverbal IQ than verbal IQ compared to the Evans and Uljarević (2018) study. This study described that children and adolescents with DS have a higher verbal IQ than nonverbal IQ (assessed with the Stanford-Binet Intelligence Scale fourth edition).
We also con rmed de cits in working memory, which is consistent with the nding by Couzens, Haynes et al. (2012). Although the cognitive pro le of DS shows relative strength in visuospatial processing skills (Pinter, Eliez et al. 2001), our study group showed impairment in visuospatial processing compared to controls. Additionally, this de cit is remarkable when compared with other verbal and nonverbal abilities. A review by Yang, Conners et al. (2014) also described visuospatial working memory as a weak area in DS. Other domains, such as uid reasoning, knowledge, and quantitative reasoning, show impairment in our study group. Raz, Torres et al. (1995) found no relationship between total brain volume and cognitive variables. Nevertheless, our results showed an association between total brain volume reduction and de cits in whole IQ and working memory. This nding con rms previous reports that found a positive association between brain volume and intelligence in the general population (McDaniel 2005, Ritchie, Booth et al. 2015. There is evidence for the association between the parietal lobe and visuospatial processing skills (Pinter, Eliez et al. 2001). This relationship depends on the nding of preserved parietal lobe volume associated with the relative strength in visuospatial processing by previous reports (Pinter, Eliez et al. 2001). However, our results of reduced parietal lobe volume and impaired visuospatial processing with a positive correlation contrast with these reports.
Additionally, we observed a correlation between the reduction in parietal lobe volume and de cits in working memory, which Menghini, Costanzo et al. (2011) con rmed. Our study could not con rm the association between reduction in the parietal lobe volume and de cits in linguistic abilities. No similar research has reported the relationship between reduced parietal lobe volume and de cits in language skills. The relationship between the decrease in parietal lobe volume and low total IQ approached signi cance. Nevertheless, there is evidence of the correlation between the parietal lobe and intelligence in the general population (Yoon, Shin et al. 2017).
Our study results con rmed the association between temporal lobe volume reduction and de cits in working memory, ensuring the temporal lobe's role in memory function (Galaburda and Schmitt 2003, Pennington, Moon et al. 2003, Vicari 2006, Menghini, Costanzo et al. 2011).
Our results could not con rm the involvement of the temporal lobe in language de cits. Nevertheless, in contrast to a study by Pinter, Eliez et al. (2001), which reported larger corrected volumes of temporal lobe volume, our results showed the reduced volume of the temporal lobe in DS participants.
This result provides neuroimaging evidence for the hypothesis of disproportionately smaller temporal lobe volumes associated with language de cits in DS. The relationship between the reduced temporal lobe volume and low total IQ approached signi cance. Nevertheless, there is evidence of the correlation between the temporal lobe and intelligence in the general population (Yoon, Shin et al. 2017).
Our study con rmed the link between hippocampal volume reduction and de cits in language and working memory, and this has been reported widely by previous studies (Raz, Torres et al. 1995, Krasuski, Alexander et al. 2002, Pennington, Moon et al. 2003. These ndings re ect the role of the hippocampus as an essential biomarker for AD and one of the regions that are severely affected by the neuropathological changes of AD (Aylward, Li et al. 1999).
We found no signi cant correlation between age and hippocampal volume. This nding is consistent with previous studies by Raz, Torres et al. (1995) and Aylward, Li et al. (1999), who failed to nd a correlation between age and hippocampal volume, but contrasts with the study by Kesslak, Nagata et al. (1994), who found a signi cant correlation between age and hippocampal volume. The age range of these reported studies is between 22-50 years, and our study group's age range is between 6-25 years. This comparison con rms the suggestion that the signi cant decrease in hippocampal volume before age 30 remains stable and then decreases later when dementia occurs in DS subjects (Aylward, Li et al. 1999). This decrease in hippocampal volume with increased age is related to changes in the neural pathway associated with memory and learning problems that start at infancy and continue throughout childhood (Kates, Kaufmann et al. 1997).
The most exciting nding in our study is the reduced volume of the parahippocampal gyrus, which contrasts with the results of Raz, Torres et al. (1995), who reported enlargement of this structure that is severely affected by AD.
We suppose that the parahippocampal gyrus volume follows DS's known neuroanatomical, neurodevelopmental, and pathological pathways.
Another interesting nding related to parahippocampal gyrus volume is the association between this structural volume reduction and de cits in working memory. There was no association between parahippocampal gyrus volume reduction and de cits in total IQ and total language. This result contrasts with that of Raz, Torres et al. (1995), who found a negative correlation between parahippocampal gyrus volume and IQ in DS subjects.
Our results signi cantly support the suggestion of narrowness of the superior temporal gyrus (Nadel 1999), although some related studies could not con rm this (Kesslak, Nagata et al. 1994, Pinter, Eliez et al. 2001. The superior temporal gyrus is part of the language network (Friederici and Gierhan 2013). It contributes to the perceptual analysis of the speech signal during auditory word processing and production and comprehension of spoken words (Zevin 2009). Our statistical nding of a negative correlation between superior temporal gyrus volume and expressive language when applying partial correlation while controlling for age and total brain volume, which approached signi cance, con rms superior temporal gyrus function. There was a correlation between a reduction in superior temporal gyrus volume and de cits in working memory. We did not nd an association between reduction in superior temporal gyrus volume and scores of total IQ and total language. We tried to include other speci c brain regions related to de cits in language and memory in DS, such as the temporoparietal junction (e.g., angular and supramarginal gyri) and occipitotemporal structures (e.g., fusiform gyrus), which are parts of the language network (Hickok and Poeppel 2004, Friederici and Gierhan 2013, Hamner, Udhnani et al. 2018). We could not con rm the relationship between angular gyrus volume and language de cits. We found a correlation that approached signi cance between angular gyrus volume and de cits in working memory. This result con rms the role of the angular gyrus in verbal working memory and other complex cognitive functions (Seghier 2013). We observed a signi cant negative correlation when we applied partial correlation between supramarginal gyrus volume and expressive language while controlling for age and total brain volume.
This nding means that the supramarginal gyrus plays a similar role in language processing skills as the superior temporal gyrus.
The reduced volume fusiform gyrus, also known as the occipitotemporal gyrus, is related to de cits in total IQ and working memory.
An interesting nding is an association between reduced fusiform gyrus volume, a parietal lobe subregion, and impairment in visuospatial processing skills. This nding supports our link between reduced parietal lobe volume and de cits in visuospatial processing skills.
As part of our study of hippocampal formation, we studied three hippocampal subregions (Dentate gyrus, Ammon's horn, and Subiculum) to understand their role in the cognitive and language skills of DS.
We observed an association between reduced dentate gyrus volume and de cits in total language skills. Dentate gyrus function in the production of long-term memory is evident by studying impaired neurogenesis in DS fetuses and Ts65Dn DS mouse models (Contestabile, Fila et al. 2007).
We found an association between reduced Ammon's horn volume and de cits in total IQ, which con rms reports indicating this hippocampal subregion's role in cognition (Pang, Kiecker et al. 2018).
Interestingly, reduced subiculum volume is associated with de cits in total IQ, working memory, visuospatial processing, and quantitative reasoning skills.
We suppose that the subiculum plays a signi cant role in cognition and memory processing in DS compared to other hippocampal subregions.
The subiculum plays an essential role in the hippocampal circuit. Nevertheless, little is known about its function, although some reports indicate a critical but ill-de ned role in spatial navigation and mnemonic processing (O'Mara, Commins et al. 2001) This study is the rst to study and assess the neuroanatomy and neuropsychology of DS in detail using high-resolution neuroimaging techniques, considering the limitations of previous related studies. Our results con rm earlier reports regarding overall patterns of brain volumes in individuals with DS and provide new evidence for abnormal volumes of speci c regional and subregional brain volumes associated with language and memory domains. Our sample's small size dampens con dence in the observed pattern of neuroanatomic abnormalities. The di culty in recruiting children and adults with DS and convincing their families to participate in the study, the cost, and the time-consuming nature of radiological and psychological examinations limit the number of subjects included. Additionally, we could not have children under ve years of age, which is not due to the rarity of samples but because the skills of the children are not enough to perform the neuropsychological assessment and respond to its content.
Although understanding the neuropathological nature of DS deserves to be pursued, studying the relationship between abnormal neuroanatomy and de cits in memory and language is of greater scienti c and practical importance. The ndings of this study indicate that the brains of subjects with DS show a well-de ned pattern of abnormalities. The correlational analysis presented in the results section of this study provides excellent evidence that represents rm conclusions. Within the studied group of intellectually disabled individuals, the degree of the global reduction in brain volume predicts the general level of intellectual performance and memory function. Additionally, a decrease in the parietal lobe volume may be a signi cant predictor of cognitive disabilities in DS, especially those associated with visuospatial processing skills. Similarly, reduced volumes of the temporal lobe and hippocampus may signi cantly predict cognitive functions in DS, especially those associated with memory and language skills. The parahippocampal gyrus volume was smaller in DS subjects than in normal controls and was related to de cits in working memory function. Therefore, the phenomenon of parahippocampal gyrus enlargement, indicated twice by independent researchers (Kesslak, Nagata et al. 1994, Raz, Torres et al. 1995, and its speci city to DS, when compared with normal aging and AD contrasted by our study and these results, maybe due to bias occurred by manual measurement of brain regions. Other neuroanatomic abnormalities could also be important markers because of their association with cognitive de cits. These markers include the superior temporal gyrus, which is related to expressive language. Additionally, regions such as the angular gyrus, supramarginal gyrus, and fusiform gyrus are interesting to understand the language network and their association with memory functions. Hippocampal subregions (Dentate gyrus, Ammon's horn, and Subiculum) are essential to understand the role of hippocampal formation and its association with the memory domain. A more extensive and longitudinal study is needed to study neuroanatomical and behavioral changes with increasing age while applying interventional rehabilitation programs to observe the effects of these methods to improve cognitive skills or prevent a greater decline with time. From a practical standpoint, these data can provide educational psychologists and teachers invaluable information for developing rationally grounded interventions to understand and alleviate these individuals' learning di culties and social problems. Figure 1 1a: A report generated by the volBrain hippocampal pipeline (HIPS) showing the segmentation of three hippocampal subregions (CA1-3 (Ammon's horn, CA4-DG (Dentate Gyrus), Subiculum) and their expected volumes from the brain of a 22-year-old male with DS. 1b: A report generated by the volBrain hippocampal pipeline (HIPS) showing the segmentation of three hippocampal subregions (color-coded: CA1-3 (Ammon's horn), CA4-DG (Dentate Gyrus, Subiculum) and their volumes from the brain of a 22year-old male with DS. Upper panels left to right: Planes of section (horizontal, transverse, sagittal) used to obtain the corresponding images below. Middle panels left-side hippocampus and its horizontal, transverse, and sagittal planes are from left to right. Lower panels, right side hippocampus, and its parts, horizontal, transverse and sagittal planes from left to right.        year-old female with DS, D) 12-year-old female control sample E) 14-year-old male with DS, F) 14-year-old female with DS for comparison between DS and control samples as well as between DS males and females.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Tables.docx