Interpreting Brain Biomarkers: Challenges and solutions in interpreting machine learning-based predictive neuroimaging

Transfer Learning of Deep Neural Networks Pretrained using the ABCD Dataset for General Psychopathology Prediction in Korean Adolescents

BACKGROUND

This study examines whether a deep neural network (DNN), trained to predict the general psychopathology factor (p-factor) using functional magnetic resonance imaging (fMRI) data from adolescents in the Adolescent Brain Cognitive Development (ABCD) study, generalizes to Korean adolescents.

METHOD

We trained a scanner-generalization neural network (SGNN) to predict p-factor scores from resting-state functional connectivity (RSFC) data of 6,905 ABCD adolescents, controlling for MRI scanner-related confounds. We then transferred the pretrained SGNN to a DNN to predict p-factor scores for 125 adolescents, including healthy individuals and those with major depressive disorder, using data from Seoul National University Hospital (SNUH). We compared the transferred DNN's performance with that of kernel ridge regression (KRR) and a baseline DNN.

RESULTS

The transferred DNN outperformed KRR (0.17 ± 0.16; 0.60 ± 0.07) and the baseline DNN (0.17 ± 0.16; 0.69 ± 0.11), achieving a higher Pearson's correlation coefficient (0.29 ± 0.18) and lower mean absolute error (0.59 ± 0.09; p < 0.005). We identified the default mode network (DMN) and visual network (VIS) as crucial functional networks (FNs) for predicting p-factors across both datasets. The dorsal attention network was specific to ABCD, while the cingulo-opercular and ventral attention networks were specific to SNUH.

CONCLUSION

The transferred SGNN successfully generalized to Korean adolescents. Altered RSFC in the DMN and VIS may serve as promising biomarkers for p-factor prediction across diverse populations, addressing heterogeneity in demographics, diagnoses, and MRI scanner characteristics.

Novel brain biomarkers of obesity in young adult women based on statistical measurements of white matter tracts

OBJECTIVE

Novel brain biomarkers of obesity were sought by studying statistical measurements on fractional anisotropy (FA) images of different white matter (WM) tracts from young adult women.

METHODS

Tract measurements were chosen that showed differences between two groups (normal weight and overweight/obese) and that were correlated with BMI. From these measurements, a simple and novel process was applied to select those that would allow the creation of models to quantify and classify the state of obesity of individuals. The biomarkers were created from the tract measurements used in the models.

RESULTS

Positive correlations were found between WM integrity and BMI, mainly in tracts involved in motor functions. From these results, two models were built to quantify and classify obesity status, whose regression coefficients formed the novel proposed obesity associated brain biomarkers.

CONCLUSION

A process for the selection of tract measurements was proposed, such models were built to determine the obesity status of subjects individually. From these models, novel brain biomarkers associated with obesity were created. These results generate new knowledge in the field, intended to be used in the future in the clinical environment as a prevention and treatment tool for brain changes associated with obesity.

SIGNIFICANCE

After studying young adult women, results opposed some of the previous results reported in literature. These consisted of positive correlations between WM integrity and obesity mainly in tracts involved in motor functions. Novel brain biomarkers of obesity were also proposed, formed by the regression coefficients involved in precise models of quantification and classification of obesity status. All this allows the generation of new knowledge and its probable subsequent clinical application.

Asymmetric Inter-Hemisphere Communication Contributes to Speech Acquisition of Toddlers with Cochlear Implants

How the lateralized language network and its functions emerge with early auditory experiences remains largely unknown. Here, early auditory development is examined using repeated optical imaging for cochlear implanted (CI) toddlers with congenital deafness from onset of restored hearing to around one year of CI hearing experiences. Machine learning models are constructed to resolve how functional organization of the bilateral language network and its sound processing support the CI children's post-implantation development of auditory and verbal communication skills. Behavioral improvement is predictable by cortical processing as well as by network organization changes, with the highest classification accuracy of 81.57%. For cortical processing, behavioral prediction is better for the left than the right hemisphere and for speech than non-speech processing. For network organization, the best prediction is obtained for resting state, with greater contribution from inter-hemisphere connections between non-homologous regions than from within-hemisphere connections. Most interestingly, systematic connectivity-to-activity models reveal that speech processing of the left language network is developmentally supported largely by global network organization, particularly asymmetric inter-hemisphere communication, rather than functional segregation of local network. These findings collectively confirm the importance of asymmetric inter-hemisphere communication in formation of the lateralized language network and its functional development with early auditory experiences.

From Serendipity to Precision: Integrating AI, Multi-Omics, and Human-Specific Models for Personalized Neuropsychiatric Care

Background/Objectives: The dual forces of structured inquiry and serendipitous discovery have long shaped neuropsychiatric research, with groundbreaking treatments such as lithium and ketamine resulting from unexpected discoveries. However, relying on chance is becoming increasingly insufficient to address the rising prevalence of mental health disorders like depression and schizophrenia, which necessitate precise, innovative approaches. Emerging technologies like artificial intelligence, induced pluripotent stem cells, and multi-omics have the potential to transform this field by allowing for predictive, patient-specific interventions. Despite these advancements, traditional methodologies such as animal models and single-variable analyses continue to be used, frequently failing to capture the complexities of human neuropsychiatric conditions. Summary: This review critically evaluates the transition from serendipity to precision-based methodologies in neuropsychiatric research. It focuses on key innovations such as dynamic systems modeling and network-based approaches that use genetic, molecular, and environmental data to identify new therapeutic targets. Furthermore, it emphasizes the importance of interdisciplinary collaboration and human-specific models in overcoming the limitations of traditional approaches. Conclusions: We highlight precision psychiatry's transformative potential for revolutionizing mental health care. This paradigm shift, which combines cutting-edge technologies with systematic frameworks, promises increased diagnostic accuracy, reproducibility, and efficiency, paving the way for tailored treatments and better patient outcomes in neuropsychiatric care.

Brain-phenotype predictions of language and executive function can survive across diverse real-world data: Dataset shifts in developmental populations

Predictive modeling potentially increases the reproducibility and generalizability of neuroimaging brain-phenotype associations. Yet, the evaluation of a model in another dataset is underutilized. Among studies that undertake external validation, there is a notable lack of attention to generalization across dataset-specific idiosyncrasies (i.e., dataset shifts). Research settings, by design, remove the between-site variations that real-world and, eventually, clinical applications demand. Here, we rigorously test the ability of a range of predictive models to generalize across three diverse, unharmonized developmental samples: the Philadelphia Neurodevelopmental Cohort (n=1291), the Healthy Brain Network (n=1110), and the Human Connectome Project in Development (n=428). These datasets have high inter-dataset heterogeneity, encompassing substantial variations in age distribution, sex, racial and ethnic minority representation, recruitment geography, clinical symptom burdens, fMRI tasks, sequences, and behavioral measures. Through advanced methodological approaches, we demonstrate that reproducible and generalizable brain-behavior associations can be realized across diverse dataset features. Results indicate the potential of functional connectome-based predictive models to be robust despite substantial inter-dataset variability. Notably, for the HCPD and HBN datasets, the best predictions were not from training and testing in the same dataset (i.e., cross-validation) but across datasets. This result suggests that training on diverse data may improve prediction in specific cases. Overall, this work provides a critical foundation for future work evaluating the generalizability of brain-phenotype associations in real-world scenarios and clinical settings.

Brain-age prediction: Systematic evaluation of site effects, and sample age range and size

Structural neuroimaging data have been used to compute an estimate of the biological age of the brain (brain-age) which has been associated with other biologically and behaviorally meaningful measures of brain development and aging. The ongoing research interest in brain-age has highlighted the need for robust and publicly available brain-age models pre-trained on data from large samples of healthy individuals. To address this need we have previously released a developmental brain-age model. Here we expand this work to develop, empirically validate, and disseminate a pre-trained brain-age model to cover most of the human lifespan. To achieve this, we selected the best-performing model after systematically examining the impact of seven site harmonization strategies, age range, and sample size on brain-age prediction in a discovery sample of brain morphometric measures from 35,683 healthy individuals (age range: 5-90 years; 53.59% female). The pre-trained models were tested for cross-dataset generalizability in an independent sample comprising 2101 healthy individuals (age range: 8-80 years; 55.35% female) and for longitudinal consistency in a further sample comprising 377 healthy individuals (age range: 9-25 years; 49.87% female). This empirical examination yielded the following findings: (1) the accuracy of age prediction from morphometry data was higher when no site harmonization was applied; (2) dividing the discovery sample into two age-bins (5-40 and 40-90 years) provided a better balance between model accuracy and explained age variance than other alternatives; (3) model accuracy for brain-age prediction plateaued at a sample size exceeding 1600 participants. These findings have been incorporated into CentileBrain (https://centilebrain.org/#/brainAGE2), an open-science, web-based platform for individualized neuroimaging metrics.

TractGeoNet: A geometric deep learning framework for pointwise analysis of tract microstructure to predict language assessment performance

We propose a geometric deep-learning-based framework, TractGeoNet, for performing regression using diffusion magnetic resonance imaging (dMRI) tractography and associated pointwise tissue microstructure measurements. By employing a point cloud representation, TractGeoNet can directly utilize tissue microstructure and positional information from all points within a fiber tract without the need to average or bin data along the streamline as traditionally required by dMRI tractometry methods. To improve regression performance, we propose a novel loss function, the Paired-Siamese Regression loss, which encourages the model to focus on accurately predicting the relative differences between regression label scores rather than just their absolute values. In addition, to gain insight into the brain regions that contribute most strongly to the prediction results, we propose a Critical Region Localization algorithm. This algorithm identifies highly predictive anatomical regions within the white matter fiber tracts for the regression task. We evaluate the effectiveness of the proposed method by predicting individual performance on two neuropsychological assessments of language using a dataset of 20 association white matter fiber tracts from 806 subjects from the Human Connectome Project Young Adult dataset. The results demonstrate superior prediction performance of TractGeoNet compared to several popular regression models that have been applied to predict individual cognitive performance based on neuroimaging features. Of the twenty tracts studied, we find that the left arcuate fasciculus tract is the most highly predictive of the two studied language performance assessments. Within each tract, we localize critical regions whose microstructure and point information are highly and consistently predictive of language performance across different subjects and across multiple independently trained models. These critical regions are widespread and distributed across both hemispheres and all cerebral lobes, including areas of the brain considered important for language function such as superior and anterior temporal regions, pars opercularis, and precentral gyrus. Overall, TractGeoNet demonstrates the potential of geometric deep learning to enhance the study of the brain's white matter fiber tracts and to relate their structure to human traits such as language performance.

Cross-cohort replicable resting-state functional connectivity in predicting symptoms and cognition of schizophrenia

Schizophrenia (SZ) is a debilitating mental illness characterized by adolescence or early adulthood onset of psychosis, positive and negative symptoms, as well as cognitive impairments. Despite a plethora of studies leveraging functional connectivity (FC) from functional magnetic resonance imaging (fMRI) to predict symptoms and cognitive impairments of SZ, the findings have exhibited great heterogeneity. We aimed to identify congruous and replicable connectivity patterns capable of predicting positive and negative symptoms as well as cognitive impairments in SZ. Predictable functional connections (FCs) were identified by employing an individualized prediction model, whose replicability was further evaluated across three independent cohorts (BSNIP, SZ = 174; COBRE, SZ = 100; FBIRN, SZ = 161). Across cohorts, we observed that altered FCs in frontal-temporal-cingulate-thalamic network were replicable in prediction of positive symptoms, while sensorimotor network was predictive of negative symptoms. Temporal-parahippocampal network was consistently identified to be associated with reduced cognitive function. These replicable 23 FCs effectively distinguished SZ from healthy controls (HC) across three cohorts (82.7%, 90.2%, and 86.1%). Furthermore, models built using these replicable FCs showed comparable accuracies to those built using the whole-brain features in predicting symptoms/cognition of SZ across the three cohorts (r = .17-.33, p < .05). Overall, our findings provide new insights into the neural underpinnings of SZ symptoms/cognition and offer potential targets for further research and possible clinical interventions.

Triple Interactions Between the Environment, Brain, and Behavior in Children: An ABCD Study

BACKGROUND

Environmental exposures play a crucial role in shaping children's behavioral development. However, the mechanisms by which these exposures interact with brain functional connectivity and influence behavior remain unexplored.

METHODS

We investigated the comprehensive environment-brain-behavior triple interactions through rigorous association, prediction, and mediation analyses, while adjusting for multiple confounders. Particularly, we examined the predictive power of brain functional network connectivity (FNC) and 41 environmental exposures for 23 behaviors related to cognitive ability and mental health in 7655 children selected from the Adolescent Brain Cognitive Development (ABCD) Study at both baseline and follow-up.

RESULTS

FNC demonstrated more predictability for cognitive abilities than for mental health, with cross-validation from the UK Biobank study (N = 20,852), highlighting the importance of thalamus and hippocampus in longitudinal prediction, while FNC+environment demonstrated more predictive power than FNC in both cross-sectional and longitudinal prediction of all behaviors, especially for mental health (r = 0.32-0.63). We found that family and neighborhood exposures were common critical environmental influencers on cognitive ability and mental health, which can be mediated by FNC significantly. Healthy perinatal development was a unique protective factor for higher cognitive ability, whereas sleep problems, family conflicts, and adverse school environments specifically increased risk of poor mental health.

CONCLUSIONS

This work revealed comprehensive environment-brain-behavior triple interactions based on the ABCD Study, identified cognitive control and default mode networks as the most predictive functional networks for a wide repertoire of behaviors, and underscored the long-lasting impact of critical environmental exposures on childhood development, in which sleep problems were the most prominent factors affecting mental health.

Brain-phenotype predictions can survive across diverse real-world data

Recent work suggests that machine learning models predicting psychiatric treatment outcomes based on clinical data may fail when applied to unharmonized samples. Neuroimaging predictive models offer the opportunity to incorporate neurobiological information, which may be more robust to dataset shifts. Yet, among the minority of neuroimaging studies that undertake any form of external validation, there is a notable lack of attention to generalization across dataset-specific idiosyncrasies. Research settings, by design, remove the between-site variations that real-world and, eventually, clinical applications demand. Here, we rigorously test the ability of a range of predictive models to generalize across three diverse, unharmonized samples: the Philadelphia Neurodevelopmental Cohort (n=1291), the Healthy Brain Network (n=1110), and the Human Connectome Project in Development (n=428). These datasets have high inter-dataset heterogeneity, encompassing substantial variations in age distribution, sex, racial and ethnic minority representation, recruitment geography, clinical symptom burdens, fMRI tasks, sequences, and behavioral measures. We demonstrate that reproducible and generalizable brain-behavior associations can be realized across diverse dataset features with sample sizes in the hundreds. Results indicate the potential of functional connectivity-based predictive models to be robust despite substantial inter-dataset variability. Notably, for the HCPD and HBN datasets, the best predictions were not from training and testing in the same dataset (i.e., cross-validation) but across datasets. This result suggests that training on diverse data may improve prediction in specific cases. Overall, this work provides a critical foundation for future work evaluating the generalizability of neuroimaging predictive models in real-world scenarios and clinical settings.

Brain Status Transferring Generative Adversarial Network for Decoding Individualized Atrophy in Alzheimer's Disease

Deep learning has been widely investigated in brain image computational analysis for diagnosing brain diseases such as Alzheimer's disease (AD). Most of the existing methods built end-to-end models to learn discriminative features by group-wise analysis. However, these methods cannot detect pathological changes in each subject, which is essential for the individualized interpretation of disease variances and precision medicine. In this article, we propose a brain status transferring generative adversarial network (BrainStatTrans-GAN) to generate corresponding healthy images of patients, which are further used to decode individualized brain atrophy. The BrainStatTrans-GAN consists of generator, discriminator, and status discriminator. First, a normative GAN is built to generate healthy brain images from normal controls. However, it cannot generate healthy images from diseased ones due to the lack of paired healthy and diseased images. To address this problem, a status discriminator with adversarial learning is designed in the training process to produce healthy brain images for patients. Then, the residual between the generated and input images can be computed to quantify pathological brain changes. Finally, a residual-based multi-level fusion network (RMFN) is built for more accurate disease diagnosis. Compared to the existing methods, our method can model individualized brain atrophy for facilitating disease diagnosis and interpretation. Experimental results on T1-weighted magnetic resonance imaging (MRI) data of 1,739 subjects from three datasets demonstrate the effectiveness of our method.

The challenges and prospects of brain-based prediction of behaviour

Relating individual brain patterns to behaviour is fundamental in system neuroscience. Recently, the predictive modelling approach has become increasingly popular, largely due to the recent availability of large open datasets and access to computational resources. This means that we can use machine learning models and interindividual differences at the brain level represented by neuroimaging features to predict interindividual differences in behavioural measures. By doing so, we could identify biomarkers and neural correlates in a data-driven fashion. Nevertheless, this budding field of neuroimaging-based predictive modelling is facing issues that may limit its potential applications. Here we review these existing challenges, as well as those that we anticipate as the field develops. We focus on the impacts of these challenges on brain-based predictions. We suggest potential solutions to address the resolvable challenges, while keeping in mind that some general and conceptual limitations may also underlie the predictive modelling approach.

Differential responses in the mirror neuron system during imitation of individual emotional facial expressions and association with autistic traits

The mirror neuron system (MNS), including the inferior frontal gyrus (IFG), inferior parietal lobule (IPL) and superior temporal sulcus (STS) plays an important role in action representation and imitation and may be dysfunctional in autism spectrum disorder (ASD). However, it's not clear how these three regions respond and interact during the imitation of different basic facial expressions and whether the pattern of responses is influenced by autistic traits. Thus, we conducted a natural facial expression (happiness, angry, sadness and fear) imitation task in 100 healthy male subjects where expression intensity was measured using facial emotion recognition software (FaceReader) and MNS responses were recorded using functional near-infrared spectroscopy (fNIRS). Autistic traits were measured using the Autism Spectrum Quotient questionnaire. Results showed that imitation of happy expressions produced the highest expression intensity but a small deactivation in MNS responses, suggesting a lower processing requirement compared to other expressions. A cosine similarity analysis indicated a distinct pattern of MNS responses during imitation of each facial expression with functional intra-hemispheric connectivity between the left IPL and left STS being significantly higher during happy compared to other expressions, while inter-hemispheric connectivity between the left and right IPL differed between imitation of fearful and sad expressions. Furthermore, functional connectivity changes during imitation of each different expression could reliably predict autistic trait scores. Overall, the results provide evidence for distinct patterns of functional connectivity changes between MNS regions during imitation of different emotions which are also associated with autistic traits.

Machine Learning and Prediction in Fetal, Infant, and Toddler Neuroimaging: A Review and Primer

Predictive models in neuroimaging are increasingly designed with the intent to improve risk stratification and support interventional efforts in psychiatry. Many of these models have been developed in samples of children school-aged or older. Nevertheless, despite growing evidence that altered brain maturation during the fetal, infant, and toddler (FIT) period modulates risk for poor mental health outcomes in childhood, these models are rarely implemented in FIT samples. Applications of predictive modeling in children of these ages provide an opportunity to develop powerful tools for improved characterization of the neural mechanisms underlying development. To facilitate the broader use of predictive models in FIT neuroimaging, we present a brief primer and systematic review on the methods used in current predictive modeling FIT studies. Reflecting on current practices in more than 100 studies conducted over the past decade, we provide an overview of topics, modalities, and methods commonly used in the field and under-researched areas. We then outline ethical and future considerations for neuroimaging researchers interested in predicting health outcomes in early life, including researchers who may be relatively new to either advanced machine learning methods or using FIT data. Altogether, the last decade of FIT research in machine learning has provided a foundation for accelerating the prediction of early-life trajectories across the full spectrum of illness and health.

The Individualized Prediction of Neurocognitive Function in People Living With HIV Based on Clinical and Multimodal Connectome Data

Neurocognitive impairment continues to be common comorbidity for people living with HIV (PLWH). Given the chronic nature of HIV disease, identifying reliable biomarkers of these impairments is essential to advance our understanding of the underlying neural foundation and facilitate screening and diagnosis in clinical care. While neuroimaging provides immense potential for such biomarkers, to date, investigations in PLWH have been mostly limited to either univariate mass techniques or a single neuroimaging modality. In the present study, connectome-based predictive modeling (CPM) was proposed to predict individual differences of cognitive functioning in PLWH, using resting-state functional connectivity (FC), white matter structural connectivity (SC), and clinical relevant measures. We also adopted an efficient feature selection approach to identify the most predictive features, which achieved an optimal prediction accuracy of r = 0.61 in the discovery dataset (n = 102) and r = 0.45 in an independent validation HIV cohort (n = 88). Two brain templates and nine distinct prediction models were also tested for better modeling generalizability. Results show that combining multimodal FC and SC features enabled higher prediction accuracy of cognitive scores in PLWH, while adding clinical and demographic metrics may further improve the prediction by introducing complementary information, which may help better evaluate the individual-level cognitive performance in PLWH.