Deep learning identifies robust gender differences in functional brain organization and their dissociable links to clinical symptoms in autism

Robust and replicable functional brain signatures of 22q11.2 deletion syndrome and associated psychosis: a deep neural network-based multi-cohort study

A major genetic risk factor for psychosis is 22q11.2 deletion (22q11.2DS). However, robust and replicable functional brain signatures of 22q11.2DS and 22q11.2DS-associated psychosis remain elusive due to small sample sizes and a focus on small single-site cohorts. Here, we identify functional brain signatures of 22q11.2DS and 22q11.2DS-associated psychosis, and their links with idiopathic early psychosis, using one of the largest multi-cohort data to date. We obtained multi-cohort clinical phenotypic and task-free fMRI data from 856 participants (101 22q11.2DS, 120 idiopathic early psychosis, 101 idiopathic autism, 123 idiopathic ADHD, and 411 healthy controls) in a case-control design. A novel spatiotemporal deep neural network (stDNN)-based analysis was applied to the multi-cohort data to identify functional brain signatures of 22q11.2DS and 22q11.2DS-associated psychosis. Next, stDNN was used to test the hypothesis that the functional brain signatures of 22q11.2DS-associated psychosis overlap with idiopathic early psychosis but not with autism and ADHD. stDNN-derived brain signatures distinguished 22q11.2DS from controls, and 22q11.2DS-associated psychosis with very high accuracies (86-94%) in the primary cohort and two fully independent cohorts without additional training. Robust distinguishing features of 22q11.2DS-associated psychosis emerged in the anterior insula node of the salience network and the striatum node of the dopaminergic reward pathway. These features also distinguished individuals with idiopathic early psychosis from controls, but not idiopathic autism or ADHD. Our results reveal that individuals with 22q11.2DS exhibit a highly distinct functional brain organization compared to controls. Additionally, the brain signatures of 22q11.2DS-associated psychosis overlap with those of idiopathic early psychosis in the salience network and dopaminergic reward pathway, providing substantial empirical support for the theoretical aberrant salience-based model of psychosis. Collectively, our findings, replicated across multiple independent cohorts, advance the understanding of 22q11.2DS and associated psychosis, underscoring the value of 22q11.2DS as a genetic model for probing the neurobiological underpinnings of psychosis and its progression.

A sex-stratified analysis of the genetic architecture of human brain anatomy

Large biobanks have dramatically advanced our understanding of genetic influences on human brain anatomy. However, most studies have combined rather than compared male and female participants. Here we screen for sex differences in the common genetic architecture of over 1000 neuroanatomical phenotypes in the UK Biobank and establish a general concordance between male and female participants in heritability estimates, genetic correlations, and variant-level effects. Notable exceptions include higher mean heritability in the female group for regional volume and surface area phenotypes; between-sex genetic correlations that are significantly below 1 in the insula and parietal cortex; and a common variant with stronger effect in male participants mapping to RBFOX1 - a gene linked to multiple neuropsychiatric disorders more common in men. This work suggests that common variant influences on human brain anatomy are largely consistent between males and females, with a few exceptions that will guide future research in growing datasets.

Frontal and occipital brain glutathione levels are unchanged in autistic adults

BACKGROUND

The neurobiological underpinnings of Autism Spectrum Disorder (ASD) are diverse and likely multifactorial. One possible mechanism is increased oxidative stress leading to altered neurodevelopment and brain function. However, this hypothesis has mostly been tested in post-mortem studies. So far, available in vivo studies in autistic individuals have reported no differences in glutathione (GSH) levels in frontal, occipital, and subcortical regions. However, these studies were limited by the technically challenging quantification of GSH, the main brain antioxidant molecule. This study aimed to overcome previous studies' limitations by using a GSH-tailored spectroscopy sequence and optimised quantification methodology to provide clarity on GSH levels in autistic adults.

METHODS

We used spectral editing proton-magnetic resonance spectroscopy (1H-MRS) combined with linear combination model fitting to quantify GSH in the dorsomedial prefrontal cortex (DMPFC) and medial occipital cortex (mOCC) of autistic and non-autistic adults (male and female). We compared GSH levels between groups. We also examined correlations between GSH and current autism symptoms, measured using the Autism Quotient (AQ).

RESULTS

Data were available from 31 adult autistic participants (24 males, 7 females) and 40 non-autistic participants (21 males, 16 females); the largest sample to date. The GSH levels did not differ between groups in either region. No correlations with AQ were observed.

CONCLUSION

GSH levels as measured using 1H-MRS are unaltered in the DMPFC and mOCC regions of autistic adults, suggesting that oxidative stress in these cortical regions is not a marked neurobiological signature of ASD.

Frontoparietal and salience network synchronizations during nonsymbolic magnitude processing predict brain age and mathematical performance in youth

The development and refinement of functional brain circuits crucial to human cognition is a continuous process that spans from childhood to adulthood. Research increasingly focuses on mapping these evolving configurations, with the aim to identify markers for functional impairments and atypical development. Among human cognitive systems, nonsymbolic magnitude representations serve as a foundational building block for future success in mathematical learning and achievement for individuals. Using task-based frontoparietal (FPN) and salience network (SN) features during nonsymbolic magnitude processing alongside machine learning algorithms, we developed a framework to construct brain age prediction models for participants aged 7-30. Our study revealed differential developmental profiles in the synchronization within and between FPN and SN networks. Specifically, we observed a linear increase in FPN connectivity, concomitant with a decline in SN connectivity across the age span. A nonlinear U-shaped trajectory in the connectivity between the FPN and SN was discerned, revealing reduced FPN-SN synchronization among adolescents compared to both pediatric and adult cohorts. Leveraging the Gradient Boosting machine learning algorithm and nested fivefold stratified cross-validation with independent training datasets, we demonstrated that functional connectivity measures of the FPN and SN nodes predict chronological age, with a correlation coefficient of .727 and a mean absolute error of 2.944 between actual and predicted ages. Notably, connectivity within the FPN emerged as the most contributing feature for age prediction. Critically, a more matured brain age estimate is associated with better arithmetic performance. Our findings shed light on the intricate developmental changes occurring in the neural networks supporting magnitude representations. We emphasize brain age estimation as a potent tool for understanding cognitive development and its relationship to mathematical abilities across the critical developmental period of youth. PRACTITIONER POINTS: This study investigated the prolonged changes in the brain's architecture across childhood, adolescence, and adulthood, with a focus on task-state frontoparietal and salience networks. Distinct developmental pathways were identified: frontoparietal synchronization strengthens consistently throughout development, while salience network connectivity diminishes with age. Furthermore, adolescents show a unique dip in connectivity between these networks. Leveraging advanced machine learning methods, we accurately predicted individuals' ages based on these brain circuits, with a more mature estimated brain age correlating with better math skills.

A multi-frequency approach of the altered functional connectome for autism spectrum disorder identification

Autism spectrum disorder stands as a multifaceted and heterogeneous neurodevelopmental condition. The utilization of functional magnetic resonance imaging to construct functional brain networks proves instrumental in comprehending the intricate interplay between brain activity and autism spectrum disorder, thereby elucidating the underlying pathogenesis at the cerebral level. Traditional functional brain networks, however, typically confine their examination to connectivity effects within a specific frequency band, disregarding potential connections among brain areas that span different frequency bands. To harness the full potential of interregional connections across diverse frequency bands within the brain, our study endeavors to develop a novel multi-frequency analysis method for constructing a comprehensive functional brain networks that incorporates multiple frequencies. Specifically, our approach involves the initial decomposition of functional magnetic resonance imaging into distinct frequency bands through wavelet transform. Subsequently, Pearson correlation is employed to generate corresponding functional brain networks and kernel for each frequency band. Finally, the classification was performed by a multi-kernel support vector machine, to preserve the connectivity effects within each band and the connectivity patterns shared among the different bands. Our proposed multi-frequency functional brain networks method yielded notable results, achieving an accuracy of 89.1%, a sensitivity of 86.67%, and an area under the curve of 0.942 in a publicly available autism spectrum disorder dataset.

Sex-biasing influence of autism-associated Ube3a gene overdosage at connectomic, behavioral, and transcriptomic levels

Genomic mechanisms enhancing risk in males may contribute to sex bias in autism. The ubiquitin protein ligase E3A gene (Ube3a) affects cellular homeostasis via control of protein turnover and by acting as transcriptional coactivator with steroid hormone receptors. Overdosage of Ube3a via duplication or triplication of chromosomal region 15q11-13 causes 1 to 2% of autistic cases. Here, we test the hypothesis that increased dosage of Ube3a may influence autism-relevant phenotypes in a sex-biased manner. We show that mice with extra copies of Ube3a exhibit sex-biasing effects on brain connectomics and autism-relevant behaviors. These effects are associated with transcriptional dysregulation of autism-associated genes, as well as genes differentially expressed in 15q duplication and in autistic people. Increased Ube3a dosage also affects expression of genes on the X chromosome, genes influenced by sex steroid hormone, and genes sex-differentially regulated by transcription factors. These results suggest that Ube3a overdosage can contribute to sex bias in neurodevelopmental conditions via influence on sex-differential mechanisms.

Neurofeedback training of executive function in autism spectrum disorder: distinct effects on brain activity levels and compensatory connectivity changes

BACKGROUND

Deficits in executive function (EF) are consistently reported in autism spectrum disorders (ASD). Tailored cognitive training tools, such as neurofeedback, focused on executive function enhancement might have a significant impact on the daily life functioning of individuals with ASD. We report the first real-time fMRI neurofeedback (rt-fMRI NF) study targeting the left dorsolateral prefrontal cortex (DLPFC) in ASD.

METHODS

Thirteen individuals with autism without intellectual disability and seventeen neurotypical individuals completed a rt-fMRI working memory NF paradigm, consisting of subvocal backward recitation of self-generated numeric sequences. We performed a region-of-interest analysis of the DLPFC, whole-brain comparisons between groups and, DLPFC-based functional connectivity.

RESULTS

The ASD and control groups were able to modulate DLPFC activity in 84% and 98% of the runs. Activity in the target region was persistently lower in the ASD group, particularly in runs without neurofeedback. Moreover, the ASD group showed lower activity in premotor/motor areas during pre-neurofeedback run than controls, but not in transfer runs, where it was seemingly balanced by higher connectivity between the DLPFC and the motor cortex. Group comparison in the transfer run also showed significant differences in DLPFC-based connectivity between groups, including higher connectivity with areas integrated into the multidemand network (MDN) and the visual cortex.

CONCLUSIONS

Neurofeedback seems to induce a higher between-group similarity of the whole-brain activity levels (including the target ROI) which might be promoted by changes in connectivity between the DLPFC and both high and low-level areas, including motor, visual and MDN regions.

Deep learning models reveal replicable, generalizable, and behaviorally relevant sex differences in human functional brain organization

Sex plays a crucial role in human brain development, aging, and the manifestation of psychiatric and neurological disorders. However, our understanding of sex differences in human functional brain organization and their behavioral consequences has been hindered by inconsistent findings and a lack of replication. Here, we address these challenges using a spatiotemporal deep neural network (stDNN) model to uncover latent functional brain dynamics that distinguish male and female brains. Our stDNN model accurately differentiated male and female brains, demonstrating consistently high cross-validation accuracy (>90%), replicability, and generalizability across multisession data from the same individuals and three independent cohorts (N ~ 1,500 young adults aged 20 to 35). Explainable AI (XAI) analysis revealed that brain features associated with the default mode network, striatum, and limbic network consistently exhibited significant sex differences (effect sizes > 1.5) across sessions and independent cohorts. Furthermore, XAI-derived brain features accurately predicted sex-specific cognitive profiles, a finding that was also independently replicated. Our results demonstrate that sex differences in functional brain dynamics are not only highly replicable and generalizable but also behaviorally relevant, challenging the notion of a continuum in male-female brain organization. Our findings underscore the crucial role of sex as a biological determinant in human brain organization, have significant implications for developing personalized sex-specific biomarkers in psychiatric and neurological disorders, and provide innovative AI-based computational tools for future research.

Is There a Bias Towards Males in the Diagnosis of Autism? A Systematic Review and Meta-Analysis

Autism is more frequently diagnosed in males, with evidence suggesting that females are more likely to be misdiagnosed or underdiagnosed. Possibly, the male/female ratio imbalance relates to phenotypic and camouflaging differences between genders. Here, we performed a comprehensive approach to phenotypic and camouflaging research in autism addressed in two studies. First (Study 1 - Phenotypic Differences in Autism), we conducted a systematic review and meta-analysis of gender differences in autism phenotype. The electronic datasets Pubmed, Scopus, Web of Science, and PsychInfo were searched. We included 67 articles that compared females and males in autism core symptoms, and in cognitive, socioemotional, and behavioural phenotypes. Autistic males exhibited more severe symptoms and social interaction difficulties on standard clinical measures than females, who, in turn, exhibited more cognitive and behavioural difficulties. Considering the hypothesis of camouflaging possibly underlying these differences, we then conducted a meta-analysis of gender differences in camouflaging (Study 2 - Camouflaging Differences in Autism). The same datasets as the first study were searched. Ten studies were included. Females used more compensation and masking camouflage strategies than males. The results support the argument of a bias in clinical procedures towards males and the importance of considering a 'female autism phenotype'-potentially involving camouflaging-in the diagnostic process.

Investigating Sex-Based Neural Differences in Autism and Their Extended Reality Intervention Implications

Autism Spectrum Disorder (ASD) affects millions of individuals worldwide, and there is growing interest in the use of extended reality (XR) technologies for intervention. Despite the promising potential of XR interventions, there remain gaps in our understanding of the neurobiological mechanisms underlying ASD, particularly in relation to sex-based differences. This scoping review synthesizes the current research on brain activity patterns in ASD, emphasizing the implications for XR interventions and neurofeedback therapy. We examine the brain regions commonly affected by ASD, the potential benefits and drawbacks of XR technologies, and the implications of sex-specific differences for designing effective interventions. Our findings underscore the need for ongoing research into the neurobiological underpinnings of ASD and sex-based differences, as well as the importance of developing tailored interventions that consider the unique needs and experiences of autistic individuals.

A sex-stratified analysis of the genetic architecture of human brain anatomy

Large biobanks have dramatically advanced our understanding of genetic influences on human brain anatomy. However, most studies have combined rather than compared males and females - despite theoretical grounds for potential sex differences. By systematically screening for sex differences in the common genetic architecture of > 1000 neuroanatomical phenotypes in the UK Biobank, we establish a general concordance between males and females in heritability estimates, genetic correlations and variant-level effects. Notable exceptions include: higher mean h 2 in females for regional volume and surface area phenotypes; between-sex genetic correlations that are significantly below 1 in the insula and parietal cortex; and, a male-specific effect common variant mapping to RBFOX1 - a gene linked to multiple male-biased neuropsychiatric disorders. This work suggests that common variant influences on human brain anatomy are largely consistent between males and females, with a few exceptions that will guide future research as biobanks continue to grow in size.

Altered Spontaneous Brain Activity in Cirrhotic Patients with Minimal Hepatic Encephalopathy: A Meta-Analysis of Resting-State Functional Imaging

(1) Background: Minimal hepatic encephalopathy (MHE) is an important complication of decompensated cirrhosis. Previous studies have demonstrated spontaneous brain activity alterations in cirrhotic patients with MHE. However, the reported results are inconsistent, which has limited our understanding of the potential neural mechanisms. Thus, we conducted a quantitative meta-analysis of resting-state functional imaging studies to identify the regional activity alterations consistently involved in MHE. (2) Methods: We searched six databases to include resting-state functional imaging studies and compared spontaneous brain activity patterns between MHE patients and healthy controls (HCs), and between cirrhotic patients without minimal hepatic encephalopathy (NMHE) and HCs. Then, a separate whole-brain voxel-wise meta-analysis between MHE or NMHE patients and HCs was conducted using seed-based d mapping with permutation of subject images. We further conducted the conjunction analysis to assess the distinct regional activity alterations between MHE and NMHE patients as compared to HCs. (3) Results: Thirteen studies with twenty datasets were included in this meta-analysis. Compared with HCs, MHE patients showed decreased spontaneous brain activity in the left superior frontal gyrus, left median cingulate/paracingulate gyri, and right precuneus. Compared with NMHE patients, MHE patients indicated decreased spontaneous brain activity in the left superior frontal gyrus, left median cingulate/paracingulate gyri, and right precuneus. (4) Conclusions: MHE is associated with spontaneous brain activity alterations involving the left superior frontal gyrus and median cingulate/paracingulate gyri, which may implicate primarily in spatial working memory and emotional disorders. These findings may contribute to a better understanding of the potential neural mechanisms, and guide further research.

Autism diagnosis in females by eating disorder professionals

IMPORTANCE

Autism and eating disorders cooccur at high rates, with autism impacting the efficacy of eating disorder treatments and outcomes. Females are underdiagnosed with autism and diagnosed later in life than their male counterparts.

OBJECTIVE

The purpose of this study was to define the incidence of eating disorder professionals identifying autism in female adolescents and young adults engaged in treatment for an eating disorder.

DESIGN

The research design is a cross-sectional retrospective records review. The charts reviewed were from the medical records of forty assigned-at-birth females between the ages of 13-25 who received treatment for an eating disorder at the partial hospitalization program (PHP) level of care between 2020 and 2022.

MAIN OUTCOMES

Upon entering PHP for an eating disorder, 10% of the study participants had a pre-existing diagnosis of autism. A total of 27.5% of participants had clinical suspicion of autism. The number of autism traits that an individual possessed directly correlated with the number of calendar days in treatment.

CONCLUSION

Eating disorder professionals in this study identified autism in 17.5% of adolescent and young adult females who entered PHP treatment for an eating disorder who had not previously been diagnosed with autism. Eating disorder professionals can anticipate that individuals with autism may be in treatment for a longer duration. Further studies should explore effective treatment measures for those with autism in treatment for an eating disorder.

Abnormal amygdala functional connectivity and deep learning classification in multifrequency bands in autism spectrum disorder: A multisite functional magnetic resonance imaging study

Previous studies have explored resting-state functional connectivity (rs-FC) of the amygdala in patients with autism spectrum disorder (ASD). However, it remains unclear whether there are frequency-specific FC alterations of the amygdala in ASD and whether FC in specific frequency bands can be used to distinguish patients with ASD from typical controls (TCs). Data from 306 patients with ASD and 314 age-matched and sex-matched TCs were collected from 28 sites in the Autism Brain Imaging Data Exchange database. The bilateral amygdala, defined as the seed regions, was used to perform seed-based FC analyses in the conventional, slow-5, and slow-4 frequency bands at each site. Image-based meta-analyses were used to obtain consistent brain regions across 28 sites in the three frequency bands. By combining generative adversarial networks and deep neural networks, a deep learning approach was applied to distinguish patients with ASD from TCs. The meta-analysis results showed frequency band specificity of FC in ASD, which was reflected in the slow-5 frequency band instead of the conventional and slow-4 frequency bands. The deep learning results showed that, compared with the conventional and slow-4 frequency bands, the slow-5 frequency band exhibited a higher accuracy of 74.73%, precision of 74.58%, recall of 75.05%, and area under the curve of 0.811 to distinguish patients with ASD from TCs. These findings may help us to understand the pathological mechanisms of ASD and provide preliminary guidance for the clinical diagnosis of ASD.

Explainable artificial intelligence for mental health through transparency and interpretability for understandability

The literature on artificial intelligence (AI) or machine learning (ML) in mental health and psychiatry lacks consensus on what "explainability" means. In the more general XAI (eXplainable AI) literature, there has been some convergence on explainability meaning model-agnostic techniques that augment a complex model (with internal mechanics intractable for human understanding) with a simpler model argued to deliver results that humans can comprehend. Given the differing usage and intended meaning of the term "explainability" in AI and ML, we propose instead to approximate model/algorithm explainability by understandability defined as a function of transparency and interpretability. These concepts are easier to articulate, to "ground" in our understanding of how algorithms and models operate and are used more consistently in the literature. We describe the TIFU (Transparency and Interpretability For Understandability) framework and examine how this applies to the landscape of AI/ML in mental health research. We argue that the need for understandablity is heightened in psychiatry because data describing the syndromes, outcomes, disorders and signs/symptoms possess probabilistic relationships to each other-as do the tentative aetiologies and multifactorial social- and psychological-determinants of disorders. If we develop and deploy AI/ML models, ensuring human understandability of the inputs, processes and outputs of these models is essential to develop trustworthy systems fit for deployment.

Interpreting mental state decoding with deep learning models

In mental state decoding, researchers aim to identify the set of mental states (e.g., experiencing happiness or fear) that can be reliably identified from the activity patterns of a brain region (or network). Deep learning (DL) models are highly promising for mental state decoding because of their unmatched ability to learn versatile representations of complex data. However, their widespread application in mental state decoding is hindered by their lack of interpretability, difficulties in applying them to small datasets, and in ensuring their reproducibility and robustness. We recommend approaching these challenges by leveraging recent advances in explainable artificial intelligence (XAI) and transfer learning, and also provide recommendations on how to improve the reproducibility and robustness of DL models in mental state decoding.

Deep learning in neuroimaging data analysis: Applications, challenges, and solutions

Methods for the analysis of neuroimaging data have advanced significantly since the beginning of neuroscience as a scientific discipline. Today, sophisticated statistical procedures allow us to examine complex multivariate patterns, however most of them are still constrained by assuming inherent linearity of neural processes. Here, we discuss a group of machine learning methods, called deep learning, which have drawn much attention in and outside the field of neuroscience in recent years and hold the potential to surpass the mentioned limitations. Firstly, we describe and explain the essential concepts in deep learning: the structure and the computational operations that allow deep models to learn. After that, we move to the most common applications of deep learning in neuroimaging data analysis: prediction of outcome, interpretation of internal representations, generation of synthetic data and segmentation. In the next section we present issues that deep learning poses, which concerns multidimensionality and multimodality of data, overfitting and computational cost, and propose possible solutions. Lastly, we discuss the current reach of DL usage in all the common applications in neuroimaging data analysis, where we consider the promise of multimodality, capability of processing raw data, and advanced visualization strategies. We identify research gaps, such as focusing on a limited number of criterion variables and the lack of a well-defined strategy for choosing architecture and hyperparameters. Furthermore, we talk about the possibility of conducting research with constructs that have been ignored so far or/and moving toward frameworks, such as RDoC, the potential of transfer learning and generation of synthetic data.

Sex Differences in Autism Spectrum Disorder: Diagnostic, Neurobiological, and Behavioral Features

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder with a worldwide prevalence of about 1%, characterized by impairments in social interaction, communication, repetitive patterns of behaviors, and can be associated with hyper- or hypo-reactivity of sensory stimulation and cognitive disability. ASD comorbid features include internalizing and externalizing symptoms such as anxiety, depression, hyperactivity, and attention problems. The precise etiology of ASD is still unknown and it is undoubted that the disorder is linked to some extent to both genetic and environmental factors. It is also well-documented and known that one of the most striking and consistent finding in ASD is the higher prevalence in males compared to females, with around 70% of ASD cases described being males. The present review looked into the most significant studies that attempted to investigate differences in ASD males and females thus trying to shade some light on the peculiar characteristics of this prevalence in terms of diagnosis, imaging, major autistic-like behavior and sex-dependent uniqueness. The study also discussed sex differences found in animal models of ASD, to provide a possible explanation of the neurological mechanisms underpinning the different presentation of autistic symptoms in males and females.

Brain Gray Matter Alterations in Hepatic Encephalopathy: A Voxel-Based Meta-Analysis of Whole-Brain Studies

Background

Previous studies on voxel-based morphometry (VBM) have found that there were gray matter alterations in patients with hepatic encephalopathy (HE). However, the reported results were inconsistent and lack a quantitative review. Therefore, this study aims for a quantitative meta-analysis of VBM analysis on patients with HE.

Methods

The studies in our meta-analysis were collected from Pubmed, Web of Science, and Embase, which were published from January 1947 to October 2021. The seed-based d mapping (SDM) method was applied to quantitatively estimate the regional gray matter abnormalities in patients with HE. A meta-regression analysis was applied to evaluate the relationship between plasma ammonia and gray matter alteration.

Results

There were nine studies, with sixteen datasets consisting of 333 participants with HE and 429 healthy controls. The pooled and subgroup meta-analyses showed an increase in gray matter volume (GMV) in the bilateral thalamus and the calcarine fissure but a decrease in the GMV in the bilateral insula, the basal ganglia, the anterior cingulate gyrus, and the cerebellum. The meta-regression showed that plasma ammonia was positively associated with the GMV in the left thalamus but was negatively associated with the GMV in the cerebellum and the bilateral striatum.

Conclusion

Gray matter volume in patients with HE largely varied and could be affected by plasma ammonia. The findings of this study could help us to better understand the pathophysiology of cognitive dysfunction in patients with HE.