Quantitative MRI Reveals Grey Matter Abnormalities in Children With Drug-Naïve Attention- De cit/Hyperactivity Disorder

Shu Su Sun Yat-sen University First A liated Hospital Yingqian Chen Sun Yat-sen University First A liated Hospital Yan Dai Sun Yat-sen University First A liated Hospital Liping Lin Sun Yat-sen University First A liated Hospital Long Qian GE Healthcare Beijing Qin Zhou Sun Yat-sen University First A liated Hospital Mengsha Zou Sun Yat-sen University First A liated Hospital Hongyu Zhang Sun Yat-sen University First A liated Hospital Meina Liu Sun Yat-sen University First A liated Hospital Xianhong Xiang Sun Yat-sen University First A liated Hospital Zhiyun Yang (  yzhyun@mail.sysu.edu.cn ) Sun Yat-sen University First A liated Hospital https://orcid.org/0000-0002-6139-9049


Introduction
Attention-de cit/hyperactivity disorder (ADHD) is a neurodevelopmental disorder that commonly occurs in children and adolescents. ADHD is usually characterized by age-inappropriate inattention, hyperactivity, and impulsivity [1,2]. So far, several neuropsychological, genetic testing, and neurochemical studies have attempted to explain the pathophysiology of ADHD [3]. Nevertheless, the neural dysfunction leading to ADHD still remains elusive. Increasing neuroimaging ndings have suggested that ADHD is associated with abnormal grey matter (GM) anatomy and aberrant neural circuit activity. Structural magnetic resonance imaging (MRI) indicated structural alterations of cortical and subcortical regions in ADHD patients, including decreased cortical and subcortical volume, reduced cortical thickness, reduced cortical surface area, and gyri cation [4][5][6]. According to task-or resting-functional MRI studies, aberrant neural circuit activity is mainly located in line with the frontostriatal pathway dysfunction and default mode network dysfunction [6][7][8]. Though structure, MRI allows macromorphological measurements of the brain. The functional MRI can reveal neural cortical activity. Yet, the microscopic biophysical processes underlying the macroscopic changes in ADHD remain unclear.
Recent studies indicated that the characterization of these processes could be realized using quantitative MRI biomarkers of brain tissue microstructure [9,10]. Quantitative MR techniques have been used to study multiple central nervous system conditions by providing useful information for assessing diseases' microscopic biophysical processes [11][12][13]. However, so far, no studies have explored T1, T2 relaxation times of the brain pathophysiology in ADHD, where brain microstructure abnormalities could potentially be uncovered.
It has been hypothesized that brain differences observed in ADHD are due to a delay in maturational processes. Myelin has a key role in brain microarchitectural changes. Quantitative MRI parameters provide well-de ned physical parameters and can be applied to estimate the human brain's development [13,14]. Speci cally, the T1 relaxation time is affected by myelin, water, and iron concentrations in brain tissue; myelin content affects the T1 value, and iron de ciency prolongs the T1 value [14]. T2 is a common water proton relaxometry measure, which can estimate the myelin water fraction (MWF), de ned as the ratio between the short T2 components of myelin water (10-40 milliseconds) and the total area of the T2 distribution [15]. Due to the brain's different physical properties and microstructural environments, water between myelin layers has a shorter T2 time than nonmyelin water [15]. In general, T2 relaxometry is mainly in uenced by tissue myelin content, while T1 relaxation times are related mainly (but not only) to the microstructural tissue integrity, including myelin content and iron content. In the other neurodevelopmental disorders on the autism spectrum, increased T2 of the brain was found to re ect reduced myelin content and correspondingly increased water content [11].
Many different relaxometry techniques have been applied to evaluate brain microstructure in children. In most cases, T1, T2 values were obtained using separate methods, which resulted in excessive acquisition time. Each sequence of T1 mapping or T2 mapping usually takes minutes, while SyMRI offers a B1corrected T1 map and a T2 map in 5-6 min [16], which makes it possible to estimate the T1 and T2 values during a single scanning session. Previous studies have demonstrated high repeatability and reproducibility of the SyMRI parametric maps across multiple sites and time points [16] and a high comparability of their constructed contrast-weighted images to conventional MR images in the brain examination of children [17] and normal subjects, as well in the assessment of different neurologic conditions [17,18]. In this study, the synthetic MRI was applied to investigate brain microstructure alterations of grey matter associated with ADHD using a whole-brain approach. In addition, we also correlated these signi cantly altered MR parameters with the disease symptom severity (Test of Variables of Attention) to evaluate if these changes could re ect the disease level. Based on the above literature, we assumed that the ADHD patients would display cortical GM microstructure alterations. Furthermore, we intended to investigate whether the signi cantly altered brain regions were correlated with clinical measurements.

Materials And Method
Participants A total of 37 pediatric ADHD patients and 27 healthy pediatric healthy volunteers were enrolled in this study. The 37 ADHD patients included 14 cases of the inattention subtype, 23 cases of the combined subtype, but no case of the hyperactivity-impulsivity subtype. The age range of the enrolled participants was 6-13y, with an average age of 9y. Age, gender, and handedness were comparable between the two groups. The male to female ratio in both groups was 6:1 (males: females in the ADHD group and healthy control group). All of the ADHD patients were drug-naïve. The typical healthy volunteers were excluded from the diagnosis of ADHD based on the same reports, with both parent and teacher ADHD indices <75th percentile. A face-to-face interview was conducted with all the participants to con rm diagnostic status.
Image Acquisition and processing MRI examinations of all participants were performed on a 3.0T scanner (SIGNA Pioneer GE Healthcare, WI, USA) using 32-channel head coils. Following localizers, a coronal T2-weighted sequence was collected to rule out any cranial organic lesion. For each subject, the sagittal three-dimensional T1weighted fast spoiled gradient echo-based sequence (T1-FSPGR) with 1.00 mm isotropic resolution and SyMRI scan were performed. With regard to SyMRI, a two-dimensional multiple-dynamic multiple-echo (MDME) pulse sequence, comprising four automatically calculated saturation delay times and two echo times, was applied to acquire the axial sections. The parameters were as follows: TR = 10205.0 ms, TE = 11.3 ms, ip angle = 20°, thickness = 2 mm/no gap, NEX = 1.00, ETL = 16, pixel size = 2.0 mm × 2.0 mm, scanning time = 5.5 min.
After MR scanning, the quantitative maps (T1, T2) were estimated from the SyMRI data using the vendorprovided postprocessing software (SyntheticMR, v11.2.2). To obtain the brain regional level T1, T2 values, the T1-FSPGR image was rst co-registered to the T1 map for each subject. Then, the co-registered T1 images were normalized to Montreal Neurological Institute (MNI) space using FSL (www.fmrib.ox.ac.uk/fsl, v6.0). Thereafter, all the T1, T2 images could be transformed to MNI space. Finally, the AAL atlas was applied to all the normalized images to extract the values of all 116 regions, including cerebrum and cerebellum.
Statistical analysis SPSS v21.0 (IBM Corp., Armonk, New York) was used to perform statistical analysis. The normality of the data was assessed via the Shapiro-Wilk test. We compared the demographic data using a chi-square test for all categorical variables. Between-group differences (ADHD vs. HC) of regional quantitative values were performed using an analysis of covariance (ANCOVA) with age and gender as nuisance covariates.
Pearson correlation analysis was performed between the signi cantly altered regional quantitative indices and clinical symptom severity measures within the ADHD group. For all tests, a P value < 0.05 was considered to be statistically signi cant. Importantly, a false-discovery rate (FDR) was performed to control the error of multiple comparisons.

Demographic Data
There were no signi cant differences in age (P = 0.390) or gender (P = 0.772) between the two groups ( Table 1). Cortical and subcortical regions of interest are shown in Table 2.  Note: The regions are listed according to a prior template obtained from an AAL atlas; the odd number represents the corresponding brain regions in the left hemisphere; the even number denotes the speci c brain regions in the right hemisphere. Note: The regions are listed according to a prior template obtained from an AAL atlas; the odd number represents the corresponding brain regions in the left hemisphere; the even number denotes the speci c brain regions in the right hemisphere.

Statistical Analysis And Correlation With Clinical Variables
There were signi cant differences between GM brain regions in T1, T2 relaxation times between the ADHD and control groups. Most brain GM regions showed the longer T1 and T2 relaxation times of all GM brain microstructures in ADHD groups.
As shown in Fig. 1A and Table 3, when analyzing T1 relaxation times of individual brain GM regions, left cerebellum 7b/8 and right cerebellum 7b/8 had longer relaxation times in ADHD than in healthy controls. T1 relaxation times of the left superior frontal gyrus, dorsolateral part was shorter in ADHD than in healthy controls. Moreover, T1 values of the left cerebellum 8 were found to have negative correlation with attention concentration level in ADHD patients (R = 0.1401, P = 0.0225, Fig. 2A).  Table  2.
With regards to T2 maps, GM regions with higher T2 relaxation times were primarily distributed in multiple regions, such as frontal, cingulate, temporal, parietal, occipital, cuneus, and cerebellum of ADHD than healthy controls, as in the following cortical and sub-cortical regions: right middle frontal gyrus, left the supplementary motor area, left superior frontal gyrus, medial, and left superior frontal gyrus, medial orbital, etc. (Fig. 1B and Table 3). Furthermore, we observed correlation relationships between T2 relaxation times of left fusiform gyrus (R = 0.2509, P = 0.0016, Fig. 2B) and right cerebellum crus2 (R = 0.1422, P = 0.0214, Fig. 2C) with attention level in ADHD patients.

Discussion
In the current study, the MR quantitative indices of GM regions were compared to show substantial differences between individuals with drug-naïve ADHD and typical healthy controls. The main differences between ADHD patients and control individuals were: 1) widespread microstructure changes in cortical and subcortical regions were observed in ADHD patients; 2) GM regions, including cerebellum, default mode network (DMN), attention and execution control network (EAN), limbic areas, etc., have been implicated in the disorder; 3) focal GM regions signal changes correlated to clinical attention concentration level were also found in our study, like left cerebellum 8, left fusiform gyrus, and right cerebellum crus2.

T1, T2 values charting cortical microstructure
As a kind of developmental disorder, ADHD has been reported to correlate with brain structure development retardation and abnormal neural network interplay [6,7,19]. Grey matter volume reductions in the ventrolateral prefrontal/insular-striatal regions, such as the right insula, putamen, globus pallidus, and caudate nucleus, have been found in ADHD patients [20,21]. Voxel-based morphometry (VBM) analyses of unaffected siblings of ADHD cases have identi ed smaller GM volume in the prefrontal cortex, medial and orbitofrontal cortex, fronto-occipital regions, and cingulate regions in ADHD patients compared to healthy controls [22]. In our study, changes in MRI relaxometry times of these regions have also been observed in ADHD patients, which may re ect macrostructure changes and microstructure alteration in these regions.
The development of the neuron ber myelination process has a critical role in brain development.
Although myelin is more abundant in white matter, a considerable quantity of myelinated bers can be found in cortical GM [9,14]. In our observation, widespread microstructure changes in cortical and subcortical regions were detected via T1 and T2 map, especially the T2 map. Increased T1value of left cerebellum 7b/8 and right cerebellum 7b/8 in ADHD patients can be explained by low myelin content and iron de ciency. Moreover, we also found a weak, signi cant negative correlation between signi cantly altered T1 relaxation times in the left cerebellum 8 and attention concentration level, which provide support for our assumption. For further study, more speci c iron-target sequence, for example, R2*, susceptibility weighted imaging (SWI) or post-processing quantitative susceptibility mapping (QSM) may use to explore the between-group iron content differences.
The longer T2 relaxation times of GM found in ADHD could also be explained by low concentration of myelin; water trapped between the myelin layers has a shorter T2 relaxation time than water in the intraand extra-cellular compartments [12,15]. Thus, the increased T2 relaxation times of the cerebellum, EAN, DMN, and limbic areas found in our study may suggest a relatively low concentration of myelin content and increased water content. Furthermore, the signi cant correlation relationships between T2 relaxation times of the left fusiform gyrus and right cerebellum crus2 and the symptom severity indicated that these regions might have critical roles in ADHD pathogenesis.
Apart from GM's lower myelination degree and correspondingly increased water content, the increased vascular interstitial space due to developmental retardation may lead to signal change. In previous study, perivascular spaces had been well realized in the neurodevelopmental degeneration and neurological diseases [23], but regards to the developmental disorder, we have poor knowledge of its role related to the pathology and physiology in ADHD. Thus, further studies are needed to clarify the related pathology occurred in ADHD. Moreover, the interference caused by individual age and potential brain development bias is hard to be diminished. In summary, the quantitative MR imaging could re ect the brain tissue microstructure changes, thus furthering our knowledge of the brain changes in ADHD.

The varied brain regions and the network involved
The altered brain GM regions have been mainly classi ed as located in some speci c functional regions [24,25], including cerebellum, default mode network (DMN), attention and execution control network (EAN), limbic areas, etc. In our study, higher T2 values of right cerebellum crus II/7b, left fusiform, and higher T1 values of left cerebellum 7b/8, right cerebellum 7b/8 have been found in ADHD compared to healthy controls, which greatly overlapped with recent ndings suggesting that signi cant gray matter network reductions occurred in bilateral crus I, left lingual gyrus, left crus II, and left fusiform in ADHD [7]. Longer T2 belongs to DMN and EAN regions, including prefrontal, parietal, cingulate, precuneus, and rectus lobes, which were also found in our study. A previous study suggested that the inverse correlation of DMN and the cognitive control networks were diminished or absent in ADHD-related children and adults [26,27]. In addition, a more diffuse pattern of resting-state network connectivity and delayed functional network development in children with ADHD have been reported [19].Signi cant increase in T2 values of limbic areas, including cingulate, insula, putamen, and temporal lobe, was also observed in ADHD, which was consistent with previous studies (meta-and mega-analysis of subcortical structural imaging studies) that found additional volume reductions in basal ganglia, insula [19,28], amygdala and hippocampus [5], as well as entire cingulate cortex structure/function abnormalities in patients with ADHD [29]. More than structure/function abnormalities, our ndings indicated microstructure alterations in these regions. In general, the microstructure alteration located in the speci c GM regions may be the pathophysiology basis for the structure/function abnormalities.
In the present study, we propose that apart from GM structure/function abnormalities in ADHD, GM microstructure abnormalities may also have an important role in the underlying mechanism of ADHD. Besides, some other networks were found to be involved in having signal values alternation in our study, including the visual network (VN), sensorimotor network (SMN), and subcortical areas. However, these other regions were rather limited and thus are not discussed in detail.
This study has a few limitations. First, the sample size was relatively small. Besides, though ADHD can be divided into inattention, hyperactive and combined ADHD subtype, the difference between subtypes was not analyzed in the present study. Therefore, extensive studies with larger populations and long-term follow-up analysis are needed to further con rm the effectiveness of T1, T2 mapping in understanding ADHD GM microstructure changes.

Conclusion
This study characterized the T1, T2 values of whole-brain GM in drug-naïve pediatric ADHD patients. Alterations of T1, T2 values were found in ADHD children, in particular GM microstructure changes, possibly concerning iron de ciency, low myelin content, and enlarge vascular interstitial space. We believe these ndings could further the understanding of the pathophysiology and clinical manifestations of ADHD, thus enabling a more precise and quantitative analysis of this condition. Future longitudinal studies with follow-up of ADHD patients are needed to further verify the partial or complete explanation of these brain changes. The brain regions with group difference in T1 (A) and T2 (B) values between ADHD and healthy controls (P < 0.05, FDR-corrected). The detailed information could be found in Table 3.