Introduction

Child neglect, defined as the failure to provide essential care during early development1 has been linked to impaired social and emotional skills2 and long-term adverse outcomes, including increased vulnerability to mental illness3 heightened suicide risk4 and other physical ailments5. Among all forms of maltreatment, child neglect remains the most prevalent, representing 76.0% of substantiated child abuse cases globally6. This form of maltreatment involves the persistent failure to meet a child’s basic needs—including adequate nutrition, clothing, shelter, supervision, and protection—placing the child’s health and development at serious risk. Despite its profound and lasting impact, neglect often receives less attention compared to more overt forms of maltreatment. Understanding the neurodevelopmental consequences of neglect requires a closer examination of its unique influence on brain structure and functional connectivity, utilizing advanced neuroimaging techniques.

Previous studies have explored the effects of neglect on brain structure and functional connectivity7,8,9. Teicher, et al.9 examined magnetic resonance imaging (MRI) data from 28 children with a history of abuse or neglect and 115 healthy controls, reporting reduced corpus callosum volume among maltreated children. Kawata, et al.7 extended these findings by examining the anterior cingulate cortex in 23 neglected children and 140 healthy controls, revealing larger gray matter volume in this region along with disruptions in the salience network’s functional connectivity, which exacerbated depressive symptoms. These studies highlight the critical role of structural abnormalities and functional connectivity disruptions in understanding neglect’s effects on the brain. However, existing research has significant gaps. Although structural and functional imaging studies have provided valuable insights, relatively few studies have utilized diffusion tensor imaging (DTI), a technique capable of detecting subtle microstructural abnormalities in white matter (WM)10.

DTI provides a unique and complementary perspective on WM microstructure compared to voxel-based morphometry (VBM) and functional connectivity analyses. Although VBM focuses on gray matter volume and density, providing structural insights, it lacks the sensitivity to detect subtle WM alterations essential for understanding neural communication11. Similarly, functional connectivity measures the temporal correlation of brain activity across regions, revealing functional integration and network dynamics without addressing the structural pathways that facilitate these connections. DTI excels in characterizing WM integrity by quantifying diffusion properties such as fractional anisotropy (FA) and axial diffusivity (AD), which serve as direct markers of axonal density, myelination, and microstructural organization11. This capability makes DTI particularly valuable for investigating how disruptions in WM pathways contribute to atypical connectivity and functional impairments. Studies employing DTI have identified microstructural changes in maltreated populations, including reduced FA in the thalamic radiation and cingulum, with these reductions being linked to neglect and associated anxiety symptoms12,13. These findings suggest that WM microstructural pathways play a crucial role in mediating functional connectivity alterations in maltreated children. Child neglect, particularly during critical developmental periods, significantly affects WM integrity14. Children who experience neglect have been associated with reduced FA in specific WM fiber tracts, suggesting impaired connectivity between brain regions responsible for emotional regulation, attention, and cognitive control14. The mechanisms underlying these alterations are thought to involve the chronic stress and deprivation characteristic of neglect, which disrupt neurodevelopmental processes such as axonal growth, synaptogenesis, and myelination15. These microstructural changes in WM are believed to contribute to the cognitive, emotional, and social impairments observed in neglected children, including difficulties in executive functioning, emotional regulation, and social interaction.

Identifying and understanding these WM alterations are critical for developing targeted interventions to mitigate the long-term impacts of neglect on brain development. Despite advancements in research, significant statistical and methodological challenges remain. Previous studies have failed to isolate the unique effects of neglect8 often grouping it with other maltreatment types, which obscures its specific impact. In this study, we proposed that the neural mechanisms underlying the psychosocial characteristics of neglected children were directly linked to the integrity of the microstructure of WM, which is associated with the pathways of connectivity between various brain regions. To test this hypothesis, we examined whether children who had primarily undergone social intervention due to neglect, supported by official maltreatment records, displayed distinct WM microstructural features and psychosocial traits compared to the general population. A comprehensive whole-brain analysis was conducted to detect WM microstructural atypicality using the tract-based spatial statistics (TBSS) analysis.

Results

Clinical demographics psychological assessment outcomes

This brain MRI study employed DTI to compare WM fiber tracts in the brains of neglected children without other types of maltreatment (Neglect, n = 21) with typically developing children (TD, n = 106). The study also explored how these differences in WM fiber tracts relate to specific psychosocial problems observed in neglected children. Table 1 summarizes the clinical demographics and psychological assessment outcomes for the participants in the dataset. Significant group differences were detected in the full-scale intelligence quotient (FSIQ), which was incorporated as a covariate in the group comparisons for brain imaging analyses beyond sex, age, and MRI type. As estimated, individuals in the Neglect group exhibited significantly greater difficulties in psychosocial behavior and cognitive function.

Table 1 Demographic and clinical characteristics of the neglect and typically developing (TD) groups.
Full size table

Tract-based Spatial statistics

A TBSS analysis, with family-wise error (FWE) correction at P < 0.5 for cluster-level significance, was performed to examine microstructural differences in WM between the two groups. Significant findings included larger AD in the right cortical spinal tract (R.CST; Montreal Neurological Institute [MNI] coordinates: x = 28, y = − 15, z = 28; cluster size = 584 voxels; FWE corrected P = 0.015), the right superior longitudinal fasciculus (R.SLF; MNI coordinates: x = 32, y = − 23, z = 39; cluster size = 24 voxels; FWE corrected P = 0.045), and left cingulum (L.CG; MNI coordinates: x = − 17, y = 26, z = 27; cluster size = 67 voxels; FWE corrected P = 0.037) in the Neglect group compared to the TD group (Fig. 1). No other areas of increased or decreased FA, medial diffusivity (MD), or radial diffusivity (RD) were found with a corrected cluster probability value that approached significance. The AD values from the identified clusters in the R.CST, R.SLF, and L.CG were extracted and used for correlation analyses.

Fig. 1
figure 1

Structural differences in the white matter (WM) between neglected and typically developing (TD) groups. Larger WM fiber tracts were observed in the Neglect group compared to the TD group. Slices of the mean FA skeleton (green), overlaid with red clusters, depict significantly higher AD values in the Neglect group than in the TD group (Neglect > TD; FWE corrected P < 0.05, red-yellow). FWE, family-wise error; AD, axial diffusivity; FA, fractional anisotropy; L.CG, left cingulum; R.CST, right cortical spinal tract; R.SLF, right superior longitudinal fasciculus

Full size image

AD values and psychological States correlation

The strength and difficulties questionnaire (SDQ) in conduct problems (CP) showed significant correlations with the AD values in the R.CST (r = 0.20, 95% confidence interval [Cl]: 0.05−0.34, false discovery rate [FDR] corrected P = 0.034; Power (1 − β) = 0.98), R.SLF (r = 0.19, 95% Cl: 0.04−0.32, FDR corrected P = 0.034; Power (1 − β) = 0.92), and L.CG (r = 0.24, 95% Cl: 0.10−0.37, FDR corrected P = 0.018; Power (1 − β) = 0.87) (Fig. 2). No other correlations were significant.

Fig. 2
figure 2

Correlations plots between CP and L.CG, CP and R.CST, and CP and R.SLF. Closed circles represent the Neglect group, and open circles represent the typically developing (TD) group. The shaded beige area indicates 95% confidence intervals for the regression line (dark yellow). L.CG, left cingulum; R.CST, right cortical spinal tract; R.SLF, right superior longitudinal fasciculus; CP, conduct problems.

Full size image

Discussion

This study explored potential differences in WM tracts between children who experienced neglect and TD children. The results of the case-control analysis showed significantly larger AD values in the R.CST, R.SLF, and L.CG in neglected children compared to TD children. The larger values in the R.CST, R.SLF, and L.CG were associated with neglected experiences and current behavioral problems observed in children in this study.

Abnormalities in the WM tract maturation are associated with psychiatric and developmental disorders12,16,17,18. AD, which measures the magnitude of diffusion parallel to the fiber tracts, is a key indicator of WM microstructural integrity. Studies have shown that AD and RD typically decrease with age in most brain regions, likely reflecting increased axonal density or caliber19. Conversely, larger AD values in specific disease processes or developmental stages may indicate pathological changes caused by neuronal alterations19. Two studies have examined the impact of neglect on WM microstructure9,14 highlighting alterations in specific regions associated with cognitive and behavioral functions. Hanson, et al.14 utilized DTI to assess WM integrity in the FA of the prefrontal cortex in children raised in institutionalized settings who experienced early neglect. Their findings revealed more diffuse WM organization in the prefrontal cortex, correlating with neurocognitive deficits. Similarly, Teicher, et al.9 found that children who experienced neglect had significant reductions in the corpus callosum area, which is indicative of compromised interhemispheric communication, potentially contributing to cognitive and emotional difficulties. The variability in the MRI analysis techniques and participant backgrounds, including differences in maltreatment type, duration, and onset of neglect, significantly contributes to the inconsistencies observed across studies. No prior research has explored the microstructural characteristics of WM fibers in neglected children using diffusion-weighted imaging (DWI). Although previous studies have emphasized the importance of WM tract integrity in children exposed to neglect, the interpretation of WM microstructural changes in children who have experienced neglect exclusively remains unclear. Our study is the first to identify increased AD values in the R.CST, R.SLF, and L.CG in this context.

This study emphasized the role of larger R.CST and L.CG tract features in children who experienced neglect. The findings align with previous research that identifies atypical brain WM tracts associated with common symptoms in maltreated children, such as post-traumatic stress disorder (PTSD) and WM microstructural abnormalities linked to maltreatment20,21. The CST, primarily originating from the frontoparietal cortices, including the primary motor cortex, secondary motor area, and somatosensory cortex22 plays a critical role in voluntary motor control. Previous studies have also examined the CST in relation to inattention and impulsivity in children with attention-deficit/hyperactivity disorder (ADHD)20. Conversely, WM microstructural abnormalities in the CG, the WM core of the cingulate gyrus23,24 have been implicated in behavioral issues such as emotional dysregulation and aggression. The CG connects the limbic system to dorsal cortical brain systems, including the corticolimbic emotional system, the fronto-parietal attentional system, and the motor-control system23,25. It plays a vital role in emotion regulation, cognitive processing, autonomic nervous system responses, and motor control through its interaction with the CST25. The observed alterations in WM in child neglect align with findings from a recent meta-analysis, which reported consistent structural abnormalities across fronto-limbic and fronto-parietal pathways in individuals with a history of childhood maltreatment26. This study further underscores the importance of the CG in the context of child neglect, suggesting that WM microstructural abnormalities in the CG may contribute to the neuropathological mechanisms associated with neglect. The positive correlations observed between larger R.CST and CP, as well as between L.CG and CP, may reflect an acquired tendency in neglected children to exhibit heightened behavioral challenges. The increased AD values observed in the CST and CG in this study provide neurobiological evidence supporting these manifestations.

Additionally, we found another atypicality in the WM tracts of the Neglect group, specifically in the larger R.SLF. The SLF is a WM tract in the brain that has been implicated in ADHD27. The SLF is a major WM tract that connects the frontal and parietal regions of the brain, playing a crucial role in cognitive functions such as attention, language, and executive functions28. Alterations in the SLF have been associated with various behavioral and cognitive issues in children. In children with ADHD, atypical microstructural organization within the SLF has been associated with reduced fine motor competence29. Specifically, atypical fiber density in the R.SLF correlates with poorer performance on fine motor tasks. Furthermore, the SLF and CG are major WM tracts integral to various cognitive and emotional functions30. A study examining structural brain connectivity in children with behavioral problems found that atypical WM integrity, particularly in the SLF and CG, is associated with childhood conduct problems31. The present study observed positive correlations between CP and the R.SLF and between CP and the L.CG in neglected children, suggesting the potential acquisition of neurobiological features characteristic of ADHD. The observed WM alterations align with DTI-based structural connectome studies, indicating that childhood maltreatment compromises the integrity of large-scale structural networks, potentially affecting both cognitive and emotional functioning32. Although children exposed to maltreatment often do not meet diagnostic criteria for neurodevelopmental disorders, trauma-related experiences and difficulties in forming secure attachments may lead to symptoms that resemble such conditions33.

Although these WM tracts have distinct origins, they interact within the brain’s intricate neural network. The CST, originating in the motor cortex, may be influenced by cognitive and emotional processes mediated through its connections with the SLF and CG.

This study had several limitations. First, identifying the distinct effects of multiple factors within this cross-sectional study posed significant difficulties due to the intricate and multifaceted nature of neurodevelopmental processes. As neglect frequently occurs alongside poverty or prematurity, low birth weight, and malnutrition, distinguishing their individual contributions to brain development remains complex34. In this study, several steps were taken to minimize the potential confounding influence of these early-life biological risks. We conducted a retrospective review of gestational age and birth weight data (where available) and found no systematic group-level differences between the neglected and TD children. We also applied a stricter inclusion criterion by excluding children with FSIQ scores below 70, which may indirectly account for the effects of severe perinatal complications and malnutrition35,36,37. Additionally, to enhance the interpretability of the results and reduce variability related to cognitive impairment, we used a different participant sample from our previous study7. Moreover, we improved the analytical rigor by adjusting for age and sex as covariates in all correlation analyses to account for developmental influences. FDR correction was applied to all statistical comparisons involving brain regions and psychological states; thereby, reducing the risk of Type I errors. These methodological enhancements substantially strengthen the validity and generalizability of the present findings. Nonetheless, some residual limitations remain. For instance, early developmental factors could not be entirely ruled out, and future multimodal and longitudinal studies should assess these risks more directly. Furthermore, the findings may have limited generalizability due to the possible influence of social interventions aimed at assisting neglected children. Second, although significant group differences in behavioral measures such as CP and hyperactivity-inattention (HI) were observed, these were reported in Table 1 for descriptive purposes. No multiple comparison corrections were applied to these demographic and clinical characteristics statistics, in line with common practices in reporting demographic characteristics7,12. However, CP was subsequently analyzed using hypothesis-driven models with full adjustment for covariates (age, sex, MRI type, and FSIQ), and FDR correction was applied to all statistical tests involving brain regions and psychological states. Third, although multiple steps were taken to reduce the risk of Type I error—most notably through FDR correction of correlation analyses—this study did not correct for the number of psychological scales analyzed, which may have introduced residual inflation of false positives. Additionally, correlations were performed on the entire sample (rather than stratified by group), which may have introduced confounding due to group membership. Fourth, sex differences were not examined despite prior evidence suggesting gender-specific effects of neglect; future research should investigate this possibility. Finally, the MRI data were acquired using two different scanner types and protocols. Although MRI type was included as a covariate, the data were not harmonized across acquisition parameters, potentially affecting the consistency of the results. These limitations underscore the need for meticulous study design, careful participant selection, and rigorous data analysis in future investigations on the effects of neglect on brain development.

In conclusion, this study highlighted significant differences in WM tracts between children who experienced neglect and TD children, particularly in the R.CST, R.SLF, and L.CG. These findings provide valuable clinical insights for designing targeted interventions and therapies that address the specific developmental challenges faced by neglected children. Understanding the interplay between these WM tracts enhances our knowledge of the neurobiological impacts of neglect and lays a foundation for developing evidence-based strategies to support affected children in achieving healthier developmental trajectories.

Methods

Ethics statement

The study protocol was approved by the Ethics Committee of the University of Fukui, Japan (assurance no. 20220034, 20210004, 20130157, 20138012 [UMIN000013215], and 20110104). Written informed consent was obtained for all data used beyond the original study purposes, and secondary data analysis was approved by the Ethics Committee of the University of Fukui, Japan (assurance no. 20220039). This study was conducted in accordance with the Declaration of Helsinki38 and the Ethical Guidelines for Clinical Studies of the Ministry of Health, Labor, and Welfare of Japan. All parents or directors of a child welfare facility provided written informed consent. Further details can be found in the study previously published by Kawata, et al.7.

Participants

This study utilized a brain DWI MRI archival dataset of children exposed to maltreatment and TD children collected under several original study protocols (Supplementary Table S1). Twenty-one cases of neglect (ICD-10-CM Code T74.02) were included in the study (mean age = 13.5 years, standard deviation [SD] = 2.6). These cases did not involve any prior exposure to other forms of maltreatment, such as physical, emotional, or sexual abuse, as previously reported by Kawata, et al.7. The maltreated children in this study had experienced continuous or significant maltreatment to the extent that a serious decision was made by the Child Protection Service or equivalent, which involved legal intervention and potentially drastic changes to the child’s life (e.g., removal from biological parents and placement in residential childcare facilities). One hundred and six TD children without a history of maltreatment were extracted from the dataset and matched for age and sex (mean age = 12.5 years, SD = 2.6). These children had no prior diagnoses of autism spectrum disorder (ASD) or ADHD39. Recruitment of the TD children was described previously40,41. Participants with an FSIQ below 70 were excluded from the analysis.

Psychological assessment

The FSIQ was assessed using the Wechsler Intelligence Scale for Children (WISC) Fourth Edition42 or the Wechsler Adult Intelligence Scale (WAIS) Third Edition43 under the supervision of pediatric clinicians (AT). Caregivers provided ratings on the SDQ to assess various psychopathological domains, including emotional symptoms (ES), CP, HI, peer problems (PP), prosocial behavior (PB), and total behavioral difficulties44. Depressive symptoms were measured using the depression self-rating scale for children (DSRS-C)45. Further details can be found in the study previously published by Kawata, et al.7.

Image preprocessing and statistical analyses

This study integrated diverse datasets from original studies, with details on the resonance image acquisition protocols presented in Supplementary Table S1. The DTI data, initially stored in the Digital Imaging and Communication in Medicine (DICOM) image format, were converted into the Neuroimaging Informatics Technology Initiative (NIFTI) image format using MRIcron (version 1.0.10101102, https://www.nitrc.org/projects/mricron). No participants exhibited focal structural abnormalities in DWI scans. Preprocessing of the DTI data was conducted using MRtrix (version 3.0.4, https://mrtrix.org) and the TBSS pipeline at the Oxford Centre for Functional MRI of the Brain Software Library (FSL, version 6.0.7.6, https://fsl.fmrib.ox.ac.uk/). Corrections for the susceptibility-induced off-resonance field, eddy current, and movement-induced distortions were implemented using the eddy tool. Bias field correction was performed using the Advanced Normalization Tools (ANTs, version 2.4.2) bias correction algorithm. Masks were generated from bias-field-corrected images. The FSL toolbox DTIFIT fitted the preprocessed images to derive four parameters of the diffusion tensor model: FA, MD, AD, and RD. MD, representing the average diffusivity of molecular motion, was calculated. AD, indicating diffusivity parallel to the axonal fibers, corresponded to the first eigenvalue (L1). RD, reflecting diffusivity perpendicular to the axonal fibers, was computed as the average of the second (L2) and third (L3) eigenvalues.

We used TBSS, a voxel-wise approach, to investigate whole-brain fiber tracts, prioritizing sensitivity, objectivity, and interpretability in DTI analyses, as described previously. The TBSS tool in FSL facilitated the voxel-wise analysis of preprocessed FA, MD, AD, and RD values. With the nonlinear registration tool of FNIRT (FMRIB’s Nonlinear Registration Tool), the FA, MD, AD, and RD images of each participant were registered to the FMRIB58_FA_1 mm template in the MNI space. A mean FA image was computed with an FA threshold value of 0.2, and this image was thinned to create a mean FA skeleton. Each participant’s aligned FA map was projected onto the skeleton using the maximum FA value detected around the tract. Finally, the AD, RD, and MD skeletons were created by applying the projection parameters derived from the FA skeletonization process. By focusing only on voxels closest to the white matter fiber skeleton, TBSS minimized statistical errors associated with imperfect registration. Additionally, the non-parametric test approach eliminated the need for image smoothing; thus, reducing partial volume effects.

To compare FA, MD, AD, and RD values between the two groups while controlling for potential confounding variables, we included age, sex, MRI type, and FSIQ as covariates of no interest in the standard general linear model (GLM, version 8.6.13) design matrix in FSL. Voxel-wise statistical analyses of FA, MD, AD, and RD data were performed using FSL’s randomize tool (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/randomise/index.html). This tool employed a non-parametric permutation-based inference approach, incorporating the threshold-free cluster enhancement (TFCE) option. Analyses involved 5,000 permutations, with FWE correction applied at a significance threshold of P < 0.05 for each measure. The anatomical locations of significant clusters were identified using the JHU white-matter tractography atlas integrated into the FSL46.

Correlation between regional brain atypicality and psychological assessment outcomes

Demographic and clinical characteristics were compared using chi-squared (χ2 tests and independent sample t-tests. Pearson correlation analyses were conducted to investigate correlations between each AD value (dependent variables) identified in the group comparisons and psychological assessment scores (independent variables), including total SDQ scores, the five SDQ subscales, and DSRS-C outcomes. These analyses incorporated adjustments for relevant covariates (age, sex, FSIQ, and MRI type) consistent with the multiple linear regression models. All statistical analyses were performed using R software (version 2024.09.0, R Core Development Team, Toulouse, France, https://cran.rstudio.com/). Statistical significance was set at FDR corrected P < 0.05.