Ethics statement
The Ethics Committee of the University of Fukui, Japan (approval nos. 20210004, 20200047, and 20220039) approved this study’s protocol. This study was conducted in accordance with the Declaration of Helsinki and the Ethical Guidelines for Clinical Studies of the Ministry of Health, Labour, and Welfare of Japan. All children and the parents or director of the child welfare facility provided written informed assent and consent, respectively.
Participants
This study was conducted on a subset who participated in a longitudinal study. Twenty-three Japanese children and adolescents aged 9–18 years (14 boys, and 9 girls; mean age ± SD: 13.9 ± 2.8 years) with maltreatment experiences (CM) were recruited from the University of Fukui Hospital and the local child welfare facilities (Table 1). All children had experienced physical, emotional, and sexual abuse and/or neglect early in life prior to coming into care by the Child Protection Service (CPS). Twenty-two children lived in a stable environment in child welfare facilities. One child lived with his parent who was not the perpetrator after receiving temporary shelter for CPS. Three children were diagnosed with neurodevelopmental disorders according to the DSM-5 criteria: attention-deficit/ hyperactivity disorder (ADHD) (1), ASD and ADHD (1), and intellectual disability (1). Most participants with CM (87.0%) were on medication naïve, except for three (13.0%) who underwent a washout 72 hours before scanning (methylphenidate, atomoxetine, and risperidone). The non-maltreated group consisted of 23 typically developing (TD) Japanese children (11 boys, and 12 girls; mean age ± SD: 14.3 ± 2.6 years) with no maltreatment history, who were recruited from the local community via advertisements, matched on age, gender, and served as controls. The exclusion criteria for TD group were: If children had other psychiatric disease diagnoses (e.g., mood-related disorders, anxiety disorders, and stress disorders), neurodevelopmental disorders (ASD, ADHD, and learning disorders), FSIQ < 70 on the Wechsler Intelligence Scale for Children-Fourth Edition, or the Wechsler Adult Intelligence Scale-Third Edition; for all participants: If children had a history of head trauma with loss of consciousness, perinatal or neonatal complications, neurological disorders, sleep disturbances, or medical conditions that might adversely affect their growth and development.
Ophthalmologic examinations
All participants underwent comprehensive ophthalmic examinations, including best-corrected visual acuity (BCVA) testing, slit-lamp biomicroscopy, intraocular pressure measurement using Goldmann applanation tonometry (Luneau Technology Operations, Normandie, France), critical fusion frequency, the Ishihara color blind test, and dilated fundoscopy.
Visual Activities Questionnaire (VAQ)[31]
The VAQ is a self-report questionnaire consisting of 33 items with eight subscales (color discrimination, glare disability, light/dark adaptation, acuity/spatial vision, depth perception, peripheral vision, visual search, and visual processing speed). Seven items (13, 14, 22, 24, 29, 30, and 31) pertaining to driving were excluded in accordance with an ADHD study in children [54]. The average score of each subscale was used in the analysis.
Farnsworth-Munsell 100-hue Test (FMT100)[32]
The FMT100 was examined to provide an objective assessment of color discrimination ability under standard light conditions (D65 daylight, 6,500 K) in the same room and place. The FMT100 is a widely used color vision test that requires participants to sequence color reference caps in the order of incremental hue variations spanning the visible spectrum. In addition to the total error score, error scores reflecting the number of misplacements were calculated separately for protanopes, deuteranopes, and tritanopes.
Optical coherence tomography (OCT), and OCT angiography (OCTA)
The OCT map, including the central retinal thickness (CRT) measurement, was captured using Triton OCT (Topcon Medical Systems, Inc., Oakland, NJ, USA). Fluorescein angiography (FA) images were captured using Spectralis Heidelberg Retinal angiography (Heidelberg Engineering, Heidelberg, Germany). All imaging tests were conducted by experienced orthoptists blinded to the group status. To examine the geographic pattern of retinal thickness, we used the 3D-map mode. The macular area (6 × 6 mm square) centred on the fovea was divided into 100 sections (10 × 10) and the layers were segmented according to the manufacturer’s instructions. The defined segments were as follows: RNFL, retinal nerve fiber layer; GCL+, ganglion cell layer with inner plexiform layer; GCL++ (RNFL and GCL+); Choroid, from the branch membrane to choroidal scleral junction, and Retina, from the internal limiting membrane to the outer segment/retinal pigment layer. The average thickness of each segment was used for subsequent statistical analyses.
Visual cognitive tasks
The experiments were conducted in a quiet room at the university. The participants were seated on a small chair approximately 50 cm in front of a 19-inch monitor. Visual cognitive tasks were implemented and presented using the PsyToolkit platform[55, 56]. All tasks were conducted using the default settings of PsyToolkit, with instructions translated into Japanese. For “Visual search[57, 58]”, participants were requested to press the button (space) to the letter “T” among presented 5–20 items, but only if it was in its regular upright position and orange, and not to do anything if no “T” was present. This procedure was repeated for 50 trials. Reaction times (RT) and error rates were recorded for each case where the target “T” was displayed (target) and not (non-target). For “Navon task[59, 60]”, participants were requested to press the B button if they saw the letter “H” or “O” and N button if they saw neither an “H” nor “O.” However, the presentation of “H” and “O” was randomly assigned to enhance either global or local features. There were 50 trials. In each trial, they took up to 4s to decide whether they saw a target letter at the global or local level. The RT and error rates were recorded for each case (Global, Local, or Non-target). For “Cueing, Posner Task[61]”, participants were requested to press the A button if they saw a go signal in the left box and L button if they saw a go signal in the right box. However, after the stimulus was presented, the cue (indicated by the X mark) appeared over the left or right box, which may or may not anticipate the correct cue, and the button on the side where the go signal appeared must be pressed to avoid being misled by it. There were 100 trials, and 75% of the cues represented valid directions, but not all trials had cues. The RT and error rates were recorded for each case (Valid, or Invalid). For “Mental rotation[62, 63]”, participants were requested to classify by mouse click which figures presented at both the bottom left and right side were matched with the 2-dimensional stimulus figure presented on the top. However, the bottom figures were displayed in a rotated state. The participants were required to mentally rotate one or both figures. There were 10 trials, and the RT and error rates were recorded.
Structural MRI acquisition
Images of 46 participants were acquired using a 3.0 Tesla General Electric SIGNA MRI system (Signa PET/MR, GE Healthcare, Milwaukee, WI, USA) equipped with an 8-channel head coil. A T1-weighted anatomical scan was obtained using a fast spoiled-gradient recalled imaging sequence with the following parameters: Repetition Time (TR) = 8.488 ms, Echo Time (TE) = 3.248 ms, Flip angle (FA) = 11°, Field of view (FOV) = 256 mm, matrix size = 256 × 256, volume dimensions = 1.0 × 1.0 × 1.0 mm³, slice thickness = 1.0 mm, and a total of 172 slices.
Preprocessing of structural images for VBM analysis
The scan data, originally in the Digital Imaging and Communication in Medicine (DICOM) image format, were converted to the Neuroimaging Informatics Technology Initiative (NIfTI) image format using MRIcron software (https://www.nitrc.org/projects/mricron). Preprocessing and statistical analysis of the structural brain image data were conducted using the Statistical Parametric Mapping 12 software (SPM12) developed by the Wellcome Department of Cognitive Neurology in London, UK. SPM12 was implemented using MATLAB (version 9.0; MathWorks Inc. Natick, MA, USA). Appropriate pre-processing steps were applied to the children’s data. First, the T1-weighted structural images of each individual were segmented into gray matter, white matter, and cerebrospinal fluid, using the segmentation algorithm available in SPM12. Second, the gray matter tissue probability map (TPM) used in the analysis was adjusted based on the model fit and tailored to the demographics of the specific pediatric population of interest. Third, the segmented gray matter and white matter tissues from all subjects were utilized to generate a customized template using Diffeomorphic Anatomical Registration through the Exponentiated Lie Algebra (DARTEL) algorithm. This step ensured accurate intersubject registration, particularly in improving the smaller inner structures’ alignment. During the segmentation process, default parameters were used except for affine regularization, which utilized the International Consortium for Brain Mapping (ICBM) template specifically designed for East Asian brains. The resulting images were further normalized to the MNI space through affine transformation, resulting in the creation of the DARTEL template. Subsequently, the segmented images of each subject were nonlinearly transformed to align them with the DARTEL template. Gaussian smoothing with a full width at half-maximum (FWHM) of 6mm was applied to gray matter images during the normalization process.
The imaging data were analyzed using SPM12 software. First, a whole-brain two-sample model was employed to explore potential differences in regional gray matter volume (GMV) changes between the CM and TD groups. Age, sex, FSIQ, dominant hand, and GMV were included as covariates in the model, and the variances associated with these factors were excluded from the analyses. Second, ROI analysis was conducted, specifically targeting the bilateral V1 as ROIs based on prior hypotheses of atypical V1 development, as observed in a previous study[3, 14]. The ROIs were defined using the automated anatomical labeling (aal) method implemented in the WFU Pick-Atlas toolbox, Version 3.0.5[64–66]. Both whole-brain and ROI analyses were conducted, and corrections for multiple comparisons at the cluster level were applied to examine the GMV differences between groups. The statistical threshold was set at the voxel level P < 0.001 at the cluster level, with a FWE correction with a threshold set to P < 0.05. The Neuromorphometrics Atlas provided by SPM12 (Neuromorphometrics, Inc.; http://www.neuromorphometrics.com/) was used to determine the anatomical localization of the significant clusters.
Statistical analysis
For group comparisons of clinical and psychological variables, Welch’s t-test and linear multiple regression analysis using heteroscedasticity-robust standard errors with age and FSIQ as covariates were used. Pearson correlation analysis was conducted to investigate the potential association between retinal thickness and GMV. The analysis focused on the residuals obtained from group comparisons of the GMV, which were adjusted for control variables (β values). The adjusted eigenvariates, which represent linearly transformed estimates of GMV, were used for correlation analysis. All statistical analyses were conducted using R software (version 4.2.1)[67].