Deep learning identifies morphological determinants of sex differences in the pre-adolescent brain

Deep learning model integrating radiologic and clinical data to predict mortality after ischemic stroke

Objective

Most prognostic indexes for ischemic stroke mortality lack radiologic information. We aimed to create and validate a deep learning-based mortality prediction model using brain diffusion weighted imaging (DWI), apparent diffusion coefficient (ADC), and clinical factors.

Methods

Data from patients with ischemic stroke who admitted to tertiary hospital during acute periods from 2013 to 2019 were collected and split into training (n = 1109), validation (n = 437), and internal test (n = 654). Data from patients from secondary cardiovascular center was used for external test set (n = 507). The algorithm for predicting mortality, based on DWI and ADC (DLP_DWI), was initially trained. Subsequently, important clinical factors were integrated into this model to create the integrated model (DLP_INTG). The performance of DLP_DWI and DLP_INTG was evaluated by using time-dependent area under the receiver operating characteristic curves (TD AUCs) and Harrell concordance index (C-index) at one-year mortality.

Results

The TD AUC of DLP_DWI was 0.643 in internal test set, and 0.785 in the external dataset. DLP_INTG had a higher performance at predicting one-year mortality than premise score in internal dataset (TD- AUC: 0.859 vs. 0.746; p = 0.046), and in external dataset (TD- AUC: 0.876 vs. 0.808; p = 0.007). DLP_DWI and DLP_INTG exhibited strong discrimination for the high-risk group for one-year mortality.

Interpretation

A deep learning model using brain DWI, ADC and the clinical factors was capable of predicting mortality in patients with ischemic stroke.

Deep learning with diffusion MRI as in vivo microscope reveals sex-related differences in human white matter microstructure

Biological sex is a crucial variable in neuroscience studies where sex differences have been documented across cognitive functions and neuropsychiatric disorders. While gross statistical differences have been previously documented in macroscopic brain structure such as cortical thickness or region size, less is understood about sex-related cellular-level microstructural differences which could provide insight into brain health and disease. Studying these microstructural differences between men and women paves the way for understanding brain disorders and diseases that manifest differently in different sexes. Diffusion MRI is an important in vivo, non-invasive methodology that provides a window into brain tissue microstructure. Our study develops multiple end-to-end classification models that accurately estimates the sex of a subject using volumetric diffusion MRI data and uses these models to identify white matter regions that differ the most between men and women. 471 male and 560 female healthy subjects (age range, 22-37 years) from the Human Connectome Project are included. Fractional anisotropy, mean diffusivity and mean kurtosis are used to capture brain tissue microstructure characteristics. Diffusion parametric maps are registered to a standard template to reduce bias that can arise from macroscopic anatomical differences like brain size and contour. This study employ three major model architectures: 2D convolutional neural networks, 3D convolutional neural networks and Vision Transformer (with self-supervised pretraining). Our results show that all 3 models achieve high sex classification performance (test AUC 0.92-0.98) across all diffusion metrics indicating definitive differences in white matter tissue microstructure between males and females. We further use complementary model architectures to inform about the pattern of detected microstructural differences and the influence of short-range versus long-range interactions. Occlusion analysis together with Wilcoxon signed-rank test is used to determine which white matter regions contribute most to sex classification. The results indicate that sex-related differences manifest in both local features as well as global features / longer-distance interactions of tissue microstructure. Our highly consistent findings across models provides new insight supporting differences between male and female brain cellular-level tissue organization particularly in the central white matter.

Sex, gender diversity, and brain structure in early adolescence

There remains little consensus about the relationship between sex and brain structure, particularly in early adolescence. Moreover, few pediatric neuroimaging studies have analyzed both sex and gender as variables of interest-many of which included small sample sizes and relied on binary definitions of gender. The current study examined gender diversity with a continuous felt-gender score and categorized sex based on X and Y allele frequency in a large sample of children ages 9-11 years old (N = 7195). Then, a statistical model-building approach was employed to determine whether gender diversity and sex independently or jointly relate to brain morphology, including subcortical volume, cortical thickness, gyrification, and white matter microstructure. Additional sensitivity analyses found that male versus female differences in gyrification and white matter were largely accounted for by total brain volume, rather than sex per se. The model with sex, but not gender diversity, was the best-fitting model in 60.1% of gray matter regions and 61.9% of white matter regions after adjusting for brain volume. The proportion of variance accounted for by sex was negligible to small in all cases. While models including felt-gender explained a greater amount of variance in a few regions, the felt-gender score alone was not a significant predictor on its own for any white or gray matter regions examined. Overall, these findings demonstrate that at ages 9-11 years old, sex accounts for a small proportion of variance in brain structure, while gender diversity is not directly associated with neurostructural diversity.

Sex, gender diversity, and brain structure in children ages 9 to 11 years old

There remains little consensus about the relationship between sex and brain structure, particularly in childhood. Moreover, few pediatric neuroimaging studies have analyzed both sex and gender as variables of interest - many of which included small sample sizes and relied on binary definitions of gender. The current study examined gender diversity with a continuous felt-gender score and categorized sex based on X and Y allele frequency in a large sample of children ages 9-11 years-old (N=7693). Then, a statistical model-building approach was employed to determine whether gender diversity and sex independently or jointly relate to brain morphology, including subcortical volume, cortical thickness, gyrification, and white matter microstructure. The model with sex, but not gender diversity, was the best-fitting model in 75% of gray matter regions and 79% of white matter regions examined. The addition of gender to the sex model explained significantly more variance than sex alone with regard to bilateral cerebellum volume, left precentral cortical thickness, as well as gyrification in the right superior frontal gyrus, right parahippocampal gyrus, and several regions in the left parietal lobe. For mean diffusivity in the left uncinate fasciculus, the model with sex, gender, and their interaction captured the most variance. Nonetheless, the magnitude of variance accounted for by sex was small in all cases and felt-gender score was not a significant predictor on its own for any white or gray matter regions examined. Overall, these findings demonstrate that at ages 9-11 years-old, sex accounts for a small proportion of variance in brain structure, while gender diversity is not directly associated with neurostructural diversity.

Prediction of gender from longitudinal MRI data via deep learning on adolescent data reveals unique patterns associated with brain structure and change over a two-year period

Deep learning algorithms for predicting neuroimaging data have shown considerable promise in various applications. Prior work has demonstrated that deep learning models that take advantage of the data's 3D structure can outperform standard machine learning on several learning tasks. However, most prior research in this area has focused on neuroimaging data from adults. Within the Adolescent Brain and Cognitive Development (ABCD) dataset, a large longitudinal development study, we examine structural MRI data to predict gender and identify gender-related changes in brain structure. Results demonstrate that gender prediction accuracy is exceptionally high (>97%) with training epochs > 200 and that this accuracy increases with age. Brain regions identified as the most discriminative in the task under study include predominantly frontal areas and the temporal lobe. When evaluating gender predictive changes specific to a two-year increase in age, a broader set of visual, cingulate, and insular regions are revealed. Our findings show a robust gender-related structural brain change pattern, even over a small age range. This suggests that it might be possible to study how the brain changes during adolescence by looking at how these changes are related to different behavioral and environmental factors.

The association of prenatal cocaine exposure with expressive and receptive language skills, phonological processing and reading ability at age 17

OBJECTIVE

Prenatal cocaine exposure (PCE) has been associated with small but significant effects on language development in childhood and early adolescence. This study examined whether this association persists into later adolescence and what relationship language skills may have with reading proficiency in this population.

METHODS

Enrolled were 338 (167 with PCE, 171 with NCE or no cocaine exposure) 17-year-olds recruited at birth who, together with their current caregiver, were seen as part of a 17-year follow-up. Participants were given assessments of reading achievement (WIAT-III), receptive and expressive language (CELF-IV), and phonological processing (CTOPP). Relationships between PCE status and language outcomes were modeled using multiple linear regression controlling for environmental and caregiver factors, and other prenatal substance exposures.

RESULTS

Adolescents with PCE scored lower in areas of phonological processing and reading related skills compared to adolescents with NCE. PCE by sex interactions were identified on language, memory and spoken language comprehension, with lower scores for girls with PCE compared to girls with NCE.

CONCLUSION

These findings suggest the persistence of PCE's relationship with phonological awareness well into adolescence. PCE was also associated with lower scores on measures of skills related to reading ability, which may be a manifestation of the observed deficits in phonological processing.

Fairness-related performance and explainability effects in deep learning models for brain image analysis

Purpose: Explainability and fairness are two key factors for the effective and ethical clinical implementation of deep learning-based machine learning models in healthcare settings. However, there has been limited work on investigating how unfair performance manifests in explainable artificial intelligence (XAI) methods, and how XAI can be used to investigate potential reasons for unfairness. Thus, the aim of this work was to analyze the effects of previously established sociodemographic-related confounders on classifier performance and explainability methods. Approach: A convolutional neural network (CNN) was trained to predict biological sex from T1-weighted brain MRI datasets of 4547 9- to 10-year-old adolescents from the Adolescent Brain Cognitive Development study. Performance disparities of the trained CNN between White and Black subjects were analyzed and saliency maps were generated for each subgroup at the intersection of sex and race. Results: The classification model demonstrated a significant difference in the percentage of correctly classified White male ( 90.3 % ± 1.7 % ) and Black male ( 81.1 % ± 4.5 % ) children. Conversely, slightly higher performance was found for Black female ( 89.3 % ± 4.8 % ) compared with White female ( 86.5 % ± 2.0 % ) children. Saliency maps showed subgroup-specific differences, corresponding to brain regions previously associated with pubertal development. In line with this finding, average pubertal development scores of subjects used in this study were significantly different between Black and White females ( p < 0.001 ) and males ( p < 0.001 ). Conclusions: We demonstrate that a CNN with significantly different sex classification performance between Black and White adolescents can identify different important brain regions when comparing subgroup saliency maps. Importance scores vary substantially between subgroups within brain structures associated with pubertal development, a race-associated confounder for predicting sex. We illustrate that unfair models can produce different XAI results between subgroups and that these results may explain potential reasons for biased performance.

Interpretable deep learning-based hippocampal sclerosis classification

OBJECTIVE

To evaluate the performance of a deep learning model for hippocampal sclerosis classification on the clinical dataset and suggest plausible visual interpretation for the model prediction.

METHODS

T2-weighted oblique coronal images of the brain MRI epilepsy protocol performed on patients were used. The training set included 320 participants with 160 no, 100 left and 60 right hippocampal sclerosis, and cross-validation was implemented. The test set consisted of 302 participants with 252 no, 25 left and 25 right hippocampal sclerosis. As the test set was imbalanced, we took an average of the accuracy achieved within each group to measure a balanced accuracy for multiclass and binary classifications. The dataset was composed to include not only healthy participants but also participants with abnormalities besides hippocampal sclerosis in the control group. We visualized the reasons for the model prediction using the layer-wise relevance propagation method.

RESULTS

When evaluated on the validation of the training set, we achieved multiclass and binary classification accuracy of 87.5% and 88.8% from the voting ensemble of six models. Evaluated on the test sets, we achieved multiclass and binary classification accuracy of 91.5% and 89.76%. The distinctly sparse visual interpretations were provided for each individual participant and group to suggest the contribution of each input voxel to the prediction on the MRI.

SIGNIFICANCE

The current interpretable deep learning-based model is promising for adapting effectively to clinical settings by utilizing commonly used data, such as MRI, with realistic abnormalities faced by neurologists to support the diagnosis of hippocampal sclerosis with plausible visual interpretation.

A Penalty Approach for Normalizing Feature Distributions to Build Confounder-Free Models

Translating the use of modern machine learning algorithms into clinical applications requires settling challenges related to explain-ability and management of nuanced confounding factors. To suitably interpret the results, removing or explaining the effect of confounding variables (or metadata) is essential. Confounding variables affect the relationship between input training data and target outputs. Accordingly, when we train a model on such data, confounding variables will bias the distribution of the learned features. A recent promising solution, Meta-Data Normalization (MDN), estimates the linear relationship between the metadata and each feature based on a non-trainable closed-form solution. However, this estimation is confined by the sample size of a mini-batch and thereby may result in an oscillating performance. In this paper, we extend the MDN method by applying a Penalty approach (referred to as PDMN). We cast the problem into a bi-level nested optimization problem. We then approximate that objective using a penalty method so that the linear parameters within the MDN layer are trainable and learned on all samples. This enables PMDN to be plugged into any architectures, even those unfit to run batch-level operations such as transformers and recurrent models. We show improvement in model accuracy and independence from the confounders using PMDN over MDN in a synthetic experiment and a multi-label, multi-site classification of magnetic resonance images.

Multiple Instance Neuroimage Transformer

For the first time, we propose using a multiple instance learning based convolution-free transformer model, called Multiple Instance Neuroimage Transformer (MINiT), for the classification of T1-weighted (T1w) MRIs. We first present several variants of transformer models adopted for neuroimages. These models extract non-overlapping 3D blocks from the input volume and perform multi-headed self-attention on a sequence of their linear projections. MINiT, on the other hand, treats each of the non-overlapping 3D blocks of the input MRI as its own instance, splitting it further into non-overlapping 3D patches, on which multi-headed self-attention is computed. As a proof-of-concept, we evaluate the efficacy of our model by training it to identify sex from T1w-MRIs of two public datasets: Adolescent Brain Cognitive Development (ABCD) and the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA). The learned attention maps highlight voxels contributing to identifying sex differences in brain morphometry. The code is available at https://github.com/singlaayush/MINIT.

Deep Learning Prediction and Visualization of Gender Related Brain Changes from Longitudinal Structural MRI Data in the ABCD Study

Deep learning algorithms for predicting from neuroimaging data have shown considerable promise. Deep learning models that take advantage of the data's 3D structure have been proven to outperform ordinary machine learning on a number of learning tasks[1]. The majority of past research in this area, however, has focused on data from adults. Within the Adolescent Brain and Cognitive Development (ABCD) dataset, a major longitudinal development research, we examine the use of structural MRI data to predict gender and to identify gender related changes in brain structure. The results demonstrate that gender prediction accuracy is extremely high (>94%), and that this accuracy increases with age. Brain regions identified as the most discriminative in the task under study include predominantly frontal regions in addition to temporal lobe. When evaluating gender predictive changes specific to a two year increase in age, a broader set of visual, cingulate, and insular regions are revealed. Overall, our findings show a robust pattern of gender related structural brain changes, even over a small age range. This suggests the potential for evaluating the relationship of these changes to various behavioral and environmental factors to further study how the brain develops during adolescence. Clinical relevance- These results are not focused on clinical relevance currently, but in the future may be useful to characterize interactions between gender and potentially clinically relevant measures in adolescents.

The sexual brain, genes, and cognition: A machine-predicted brain sex score explains individual differences in cognitive intelligence and genetic influence in young children

Sex impacts the development of the brain and cognition differently across individuals. However, the literature on brain sex dimorphism in humans is mixed. We aim to investigate the biological underpinnings of the individual variability of sexual dimorphism in the brain and its impact on cognitive performance. To this end, we tested whether the individual difference in brain sex would be linked to that in cognitive performance that is influenced by genetic factors in prepubertal children (N = 9,658, ages 9-10 years old; the Adolescent Brain Cognitive Development study). To capture the interindividual variability of the brain, we estimated the probability of being male or female based on the brain morphometry and connectivity features using machine learning (herein called a brain sex score). The models accurately classified the biological sex with a test ROC-AUC of 93.32%. As a result, a greater brain sex score correlated significantly with greater intelligence (p_fdr  < .001, η p 2 $$ {\eta}_p^2 $$  = .011-.034; adjusted for covariates) and higher cognitive genome-wide polygenic scores (GPSs) (p_fdr  < .001, η p 2 $$ {\eta}_p^2 $$  < .005). Structural equation models revealed that the GPS-intelligence association was significantly modulated by the brain sex score, such that a brain with a higher maleness score (or a lower femaleness score) mediated a positive GPS effect on intelligence (indirect effects = .006-.009; p = .002-.022; sex-stratified analysis). The finding of the sex modulatory effect on the gene-brain-cognition relationship presents a likely biological pathway to the individual and sex differences in the brain and cognitive performance in preadolescence.

A Novel Explainability Approach for Technology-Driven Translational Research on Brain Aging

Brain aging leads to difficulties in functional independence. Mitigating these difficulties can benefit from technology that predicts, monitors, and modifies brain aging. Translational research prioritizes solutions that can be causally linked to specific pathophysiologies at the same time as demonstrating improvements in impactful real-world outcome measures. This poses a challenge for brain aging technology that needs to address the tension between mechanism-driven precision and clinical relevance. In the current opinion, by synthesizing emerging mechanistic, translational, and clinical research-related frameworks, and our own development of technology-driven brain aging research, we suggest incorporating the appreciation of four desiderata (causality, informativeness, transferability, and fairness) of explainability into early-stage research that designs and tests brain aging technology. We apply a series of work on electrocardiography-based "peripheral" neuroplasticity markers from our work as an illustration of our proposed approach. We believe this novel approach will promote the development and adoption of brain aging technology that links and addresses brain pathophysiology and functional independence in the field of translational research.

Multi-label, multi-domain learning identifies compounding effects of HIV and cognitive impairment

Older individuals infected by Human Immunodeficiency Virus (HIV) are at risk for developing HIV-Associated Neurocognitive Disorder (HAND), i.e., from reduced cognitive functioning similar to HIV-negative individuals with Mild Cognitive Impairment (MCI) or to Alzheimer's Disease (AD) if more severely affected. Incompletely understood is how brain structure can serve to differentiate cognitive impairment (CI) in the HIV-positive (i.e., HAND) from the HIV-negative cohort (i.e., MCI and AD). To that end, we designed a multi-label classifier that labels the structural magnetic resonance images (MRI) of individuals by their HIV and CI status via two binary variables. Proper training of such an approach traditionally requires well-curated datasets containing large number of samples for each of the corresponding four cohorts (healthy controls, CI HIV-negative adults a.k.a. CI-only, HIV-positive patients without CI a.k.a. HIV-only, and HAND). Because of the rarity of such datasets, we proposed to improve training of the multi-label classifier via a multi-domain learning scheme that also incorporates domain-specific classifiers on auxiliary single-label datasets specific to either binary label. Specifically, we complement the training dataset of MRIs of the four cohorts (Control: 156, CI-only: 335, HIV-only: 37, HAND: 145) acquired by the Memory and Aging Center at the University of California - San Francisco with a CI-specific dataset only containing MRIs of HIV-negative subjects (Controls: 229, CI-only: 397) from the Alzheimer's Disease Neuroimaging Initiative and an HIV-specific dataset (Controls: 75, HIV-only: 75) provided by SRI International. Based on cross-validation on the UCSF dataset, the multi-domain and multi-label learning strategy leads to superior classification accuracy compared with one-domain or multi-class learning approaches, specifically for the undersampled HIV-only cohort. The 'prediction logits' of CI computed by the multi-label formulation also successfully stratify motor performance among the HIV-positive subjects (including HAND). Finally, brain patterns driving the subject-level predictions across all four cohorts characterize the independent and compounding effects of HIV and CI in the HAND cohort.

BrainGNN: Interpretable Brain Graph Neural Network for fMRI Analysis

Understanding which brain regions are related to a specific neurological disorder or cognitive stimuli has been an important area of neuroimaging research. We propose BrainGNN, a graph neural network (GNN) framework to analyze functional magnetic resonance images (fMRI) and discover neurological biomarkers. Considering the special property of brain graphs, we design novel ROI-aware graph convolutional (Ra-GConv) layers that leverage the topological and functional information of fMRI. Motivated by the need for transparency in medical image analysis, our BrainGNN contains ROI-selection pooling layers (R-pool) that highlight salient ROIs (nodes in the graph), so that we can infer which ROIs are important for prediction. Furthermore, we propose regularization terms-unit loss, topK pooling (TPK) loss and group-level consistency (GLC) loss-on pooling results to encourage reasonable ROI-selection and provide flexibility to encourage either fully individual- or patterns that agree with group-level data. We apply the BrainGNN framework on two independent fMRI datasets: an Autism Spectrum Disorder (ASD) fMRI dataset and data from the Human Connectome Project (HCP) 900 Subject Release. We investigate different choices of the hyper-parameters and show that BrainGNN outperforms the alternative fMRI image analysis methods in terms of four different evaluation metrics. The obtained community clustering and salient ROI detection results show a high correspondence with the previous neuroimaging-derived evidence of biomarkers for ASD and specific task states decoded for HCP. Our code is available at https://github.com/xxlya/BrainGNN_Pytorch.

Machine learning to advance the prediction, prevention and treatment of eating disorders

Machine learning approaches are just emerging in eating disorders research. Promising early results suggest that such approaches may be a particularly promising and fruitful future direction. However, there are several challenges related to the nature of eating disorders in building robust, reliable and clinically meaningful prediction models. This article aims to provide a brief introduction to machine learning and to discuss several such challenges, including issues of sample size, measurement, imbalanced data and bias; I also provide concrete steps and recommendations for each of these issues. Finally, I outline key outstanding questions and directions for future research in building, testing and implementing machine learning models to advance our prediction, prevention, and treatment of eating disorders.

Longitudinal self-supervised learning

Machine learning analysis of longitudinal neuroimaging data is typically based on supervised learning, which requires large number of ground-truth labels to be informative. As ground-truth labels are often missing or expensive to obtain in neuroscience, we avoid them in our analysis by combing factor disentanglement with self-supervised learning to identify changes and consistencies across the multiple MRIs acquired of each individual over time. Specifically, we propose a new definition of disentanglement by formulating a multivariate mapping between factors (e.g., brain age) associated with an MRI and a latent image representation. Then, factors that evolve across acquisitions of longitudinal sequences are disentangled from that mapping by self-supervised learning in such a way that changes in a single factor induce change along one direction in the representation space. We implement this model, named Longitudinal Self-Supervised Learning (LSSL), via a standard autoencoding structure with a cosine loss to disentangle brain age from the image representation. We apply LSSL to two longitudinal neuroimaging studies to highlight its strength in extracting the brain-age information from MRI and revealing informative characteristics associated with neurodegenerative and neuropsychological disorders. Moreover, the representations learned by LSSL facilitate supervised classification by recording faster convergence and higher (or similar) prediction accuracy compared to several other representation learning techniques.

Longitudinal Pooling & Consistency Regularization to Model Disease Progression From MRIs

Many neurological diseases are characterized by gradual deterioration of brain structure andfunction. Large longitudinal MRI datasets have revealed such deterioration, in part, by applying machine and deep learning to predict diagnosis. A popular approach is to apply Convolutional Neural Networks (CNN) to extract informative features from each visit of the longitudinal MRI and then use those features to classify each visit via Recurrent Neural Networks (RNNs). Such modeling neglects the progressive nature of the disease, which may result in clinically implausible classifications across visits. To avoid this issue, we propose to combine features across visits by coupling feature extraction with a novel longitudinal pooling layer and enforce consistency of the classification across visits in line with disease progression. We evaluate the proposed method on the longitudinal structural MRIs from three neuroimaging datasets: Alzheimer's Disease Neuroimaging Initiative (ADNI, N=404), a dataset composed of 274 normal controls and 329 patients with Alcohol Use Disorder (AUD), and 255 youths from the National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA). In allthree experiments our method is superior to other widely used approaches for longitudinal classification thus making a unique contribution towards more accurate tracking of the impact of conditions on the brain. The code is available at https://github.com/ouyangjiahong/longitudinal-pooling.

Metadata Normalization

Batch Normalization (BN) and its variants have delivered tremendous success in combating the covariate shift induced by the training step of deep learning methods. While these techniques normalize feature distributions by standardizing with batch statistics, they do not correct the influence on features from extraneous variables or multiple distributions. Such extra variables, referred to as metadata here, may create bias or confounding effects (e.g., race when classifying gender from face images). We introduce the Metadata Normalization (MDN) layer, a new batch-level operation which can be used end-to-end within the training framework, to correct the influence of metadata on feature distributions. MDN adopts a regression analysis technique traditionally used for preprocessing to remove (regress out) the metadata effects on model features during training. We utilize a metric based on distance correlation to quantify the distribution bias from the metadata and demonstrate that our method successfully removes metadata effects on four diverse settings: one synthetic, one 2D image, one video, and one 3D medical image dataset.

Association between Hippocampal Volume and Working Memory in 10,000+ 9-10-Year-Old Children: Sex Differences

AIM

This study tested sex differences in the association between hippocampal volume and working memory of a national sample of 9-10-year-old children in the US. As the hippocampus is functionally lateralized (especially in task-related activities), we explored the results for the right and the left hippocampus.

METHODS

This is a cross-sectional study using the Adolescent Brain Cognitive Development (ABCD) Study data. This analysis included baseline ABCD data (n = 10,093) of children between ages 9 and 10 years. The predictor variable was right and left hippocampal volume measured by structural magnetic resonance imaging (sMRI). The primary outcome, list sorting working memory, was measured using the NIH toolbox measure. Sex was the moderator. Age, race, ethnicity, household income, parental education, and family structure were the covariates.

RESULTS

In the overall sample, larger right (b = 0.0013; p < 0.001) and left (b = 0.0013; p < 0.001) hippocampal volumes were associated with higher children's working memory. Sex had statistically significant interactions with the right (b = -0.0018; p = 0.001) and left (b = -0.0012; p = 0.022) hippocampal volumes on children's working memory. These interactions indicated stronger positive associations between right and left hippocampal volume and working memory for females compared to males.

CONCLUSION

While right and left hippocampal volumes are determinants of children's list sorting working memory, these effects seem to be more salient for female than male children. Research is needed on the role of socialization, sex hormones, and brain functional connectivity as potential mechanisms that may explain the observed sex differences in the role of hippocampal volume as a correlate of working memory.

Training confounder-free deep learning models for medical applications

The presence of confounding effects (or biases) is one of the most critical challenges in using deep learning to advance discovery in medical imaging studies. Confounders affect the relationship between input data (e.g., brain MRIs) and output variables (e.g., diagnosis). Improper modeling of those relationships often results in spurious and biased associations. Traditional machine learning and statistical models minimize the impact of confounders by, for example, matching data sets, stratifying data, or residualizing imaging measurements. Alternative strategies are needed for state-of-the-art deep learning models that use end-to-end training to automatically extract informative features from large set of images. In this article, we introduce an end-to-end approach for deriving features invariant to confounding factors while accounting for intrinsic correlations between the confounder(s) and prediction outcome. The method does so by exploiting concepts from traditional statistical methods and recent fair machine learning schemes. We evaluate the method on predicting the diagnosis of HIV solely from Magnetic Resonance Images (MRIs), identifying morphological sex differences in adolescence from those of the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA), and determining the bone age from X-ray images of children. The results show that our method can accurately predict while reducing biases associated with confounders. The code is available at https://github.com/qingyuzhao/br-net .