Detecting suicide risk among U.S. servicemembers and veterans: a deep learning approach using social media data.

BACKGROUND: Military Servicemembers and Veterans are at elevated risk for suicide, but rarely self-identify to their leaders or clinicians regarding their experience of suicidal thoughts. We developed an algorithm to identify posts containing suicide-related content on a military-specific social media platform. METHODS: Publicly-shared social media posts (n = 8449) from a military-specific social media platform were reviewed and labeled by our team for the presence/absence of suicidal thoughts and behaviors and used to train several machine learning models to identify such posts. RESULTS: The best performing model was a deep learning (RoBERTa) model that incorporated post text and metadata and detected the presence of suicidal posts with relatively high sensitivity (0.85), specificity (0.96), precision (0.64), F1 score (0.73), and an area under the precision-recall curve of 0.84. Compared to non-suicidal posts, suicidal posts were more likely to contain explicit mentions of suicide, descriptions of risk factors (e.g. depression, PTSD) and help-seeking, and first-person singular pronouns. CONCLUSIONS: Our results demonstrate the feasibility and potential promise of using social media posts to identify at-risk Servicemembers and Veterans. Future work will use this approach to deliver targeted interventions to social media users at risk for suicide.

Year-Long Cannabis Use for Medical Symptoms and Brain Activation During Cognitive Processes.

Importance: Cannabis is increasingly being used to treat medical symptoms, but the effects on brain function in those using cannabis for these symptoms are not known. Objective: To test whether 1 year of cannabis use for medical symptoms after obtaining a medical cannabis card was associated with increased brain activation during working memory, reward, and inhibitory control tasks, areas of cognition affected by cannabis. Design, Setting, and Participants: This cohort study was conducted from July 2017 to July 2020 among participants from the greater Boston area who were recruited as part of a clinical trial of individuals seeking medical cannabis cards for anxiety, depression, pain, or insomnia symptoms. Participants were aged between 18 and 65 years. Exclusion criteria were daily cannabis use and cannabis use disorder at baseline. Data analysis was conducted from August 2021 to April 2024. Main Outcomes and Measures: Outcomes were whole brain functional activation during tasks involving working memory, reward, and inhibitory control at baseline and after 1 year of medical cannabis card ownership. Results: Imaging was collected from participants before and 1 year after obtaining medical cannabis cards, with 57 participants at baseline (38 female [66.7%]; 6 [10.5%] Black and 45 [78.9%] White participants; 1 [1.8%] Hispanic participant; median [IQR] age, 34.0 [24.0-51.0] years) and 54 participants at 1 year (37 female [68.5%]; 4 [7.4%] Black and 48 [88.9%] White participants; 1 [1.9%] Hispanic participant, median [IQR] age, 36.5 [25.0-51.0] years). Imaging was also collected in 32 healthy control participants at baseline (22 female [68.8%]; 2 [6.2%] Black and 27 [84.4%] White participants; 3 [9.4%] Hispanic participants; median [IQR] age, 33.0 [24.8-38.2] years). In all groups and at both time points, functional imaging revealed canonical activations of the probed cognitive processes. No statistically significant difference in brain activation between the 2 time points (baseline and 1 year) in those with medical cannabis cards and no associations between changes in cannabis use frequency and brain activation after 1 year were found. Conclusions and Relevance: In this cohort study of adults obtaining medical cannabis cards for medical symptoms, no significant association between brain activation in the areas of cognition of working memory, reward, and inhibitory control and 1 year of cannabis use was observed. The results warrant further studies that probe the association of cannabis at higher doses, with greater frequency, in younger age groups, and with larger, more diverse cohorts.

Brainchop: Providing an Edge Ecosystem for Deployment of Neuroimaging Artificial Intelligence Models.

Deep learning has proven highly effective in various medical imaging scenarios, yet the lack of an efficient distribution platform hinders developers from sharing models with end-users. Here, we describe brainchop, a fully functional web application that allows users to apply deep learning models developed with Python to local neuroimaging data from within their browser. While training artificial intelligence models is computationally expensive, applying existing models to neuroimaging data can be very fast; brainchop harnesses the end user's graphics card such that brain extraction, tissue segmentation, and regional parcellation require only seconds and avoids privacy issues that impact cloud-based solutions. The integrated visualization allows users to validate the inferences, and includes tools to annotate and edit the resulting segmentations. Our pure JavaScript implementation includes optimized helper functions for conforming volumes and filtering connected components with minimal dependencies. Brainchop provides a simple mechanism for distributing models for additional image processing tasks, including registration and identification of abnormal tissue, including tumors, lesions and hyperintensities. We discuss considerations for other AI model developers to leverage this open-source resource.

The past, present, and future of the brain imaging data structure (BIDS).

The Brain Imaging Data Structure (BIDS) is a community-driven standard for the organization of data and metadata from a growing range of neuroscience modalities. This paper is meant as a history of how the standard has developed and grown over time. We outline the principles behind the project, the mechanisms by which it has been extended, and some of the challenges being addressed as it evolves. We also discuss the lessons learned through the project, with the aim of enabling researchers in other domains to learn from the success of BIDS.

Resting-State Functional Connectivity of the Amygdala in Autism: A Preregistered Large-Scale Study.

OBJECTIVE: Three leading neurobiological hypotheses about autism spectrum disorder (ASD) propose underconnectivity between brain regions, atypical function of the amygdala, and generally higher variability between individuals with ASD than between neurotypical individuals. Past work has often failed to generalize, because of small sample sizes, unquantified data quality, and analytic flexibility. This study addressed these limitations while testing the above three hypotheses, applied to amygdala functional connectivity. METHODS: In a comprehensive preregistered study, the three hypotheses were tested in a subset (N=488 after exclusions; N=212 with ASD) of the Autism Brain Imaging Data Exchange data sets. The authors analyzed resting-state functional connectivity (FC) from functional MRI data from two anatomically defined amygdala subdivisions, in three hypotheses with respect to magnitude, pattern similarity, and variability, across different anatomical scales ranging from whole brain to specific regions and networks. RESULTS: A Bayesian approach to hypothesis evaluation produced inconsistent evidence in ASD for atypical amygdala FC magnitude, strong evidence that the multivariate pattern of FC was typical, and no consistent evidence of increased interindividual variability in FC. The results strongly depended on analytic choices, including preprocessing pipeline for the neuroimaging data, anatomical specificity, and subject exclusions. CONCLUSIONS: A preregistered set of analyses found no reliable evidence for atypical functional connectivity of the amygdala in autism, contrary to leading hypotheses. Future studies should test an expanded set of hypotheses across multiple processing pipelines, collect deeper data per individual, and include a greater diversity of participants to ensure robust generalizability of findings on amygdala FC in ASD.

Integrated platform for multiscale molecular imaging and phenotyping of the human brain.

Understanding cellular architectures and their connectivity is essential for interrogating system function and dysfunction. However, we lack technologies for mapping the multiscale details of individual cells and their connectivity in the human organ-scale system. We developed a platform that simultaneously extracts spatial, molecular, morphological, and connectivity information of individual cells from the same human brain. The platform includes three core elements: a vibrating microtome for ultraprecision slicing of large-scale tissues without losing cellular connectivity (MEGAtome), a polymer hydrogel-based tissue processing technology for multiplexed multiscale imaging of human organ-scale tissues (mELAST), and a computational pipeline for reconstructing three-dimensional connectivity across multiple brain slabs (UNSLICE). We applied this platform for analyzing human Alzheimer's disease pathology at multiple scales and demonstrating scalable neural connectivity mapping in the human brain.

Identifying bias in models that detect vocal fold paralysis from audio recordings using explainable machine learning and clinician ratings.

Detecting voice disorders from voice recordings could allow for frequent, remote, and low-cost screening before costly clinical visits and a more invasive laryngoscopy examination. Our goals were to detect unilateral vocal fold paralysis (UVFP) from voice recordings using machine learning, to identify which acoustic variables were important for prediction to increase trust, and to determine model performance relative to clinician performance. Patients with confirmed UVFP through endoscopic examination (N = 77) and controls with normal voices matched for age and sex (N = 77) were included. Voice samples were elicited by reading the Rainbow Passage and sustaining phonation of the vowel "a". Four machine learning models of differing complexity were used. SHapley Additive exPlanations (SHAP) was used to identify important features. The highest median bootstrapped ROC AUC score was 0.87 and beat clinician's performance (range: 0.74-0.81) based on the recordings. Recording durations were different between UVFP recordings and controls due to how that data was originally processed when storing, which we can show can classify both groups. And counterintuitively, many UVFP recordings had higher intensity than controls, when UVFP patients tend to have weaker voices, revealing a dataset-specific bias which we mitigate in an additional analysis. We demonstrate that recording biases in audio duration and intensity created dataset-specific differences between patients and controls, which models used to improve classification. Furthermore, clinician's ratings provide further evidence that patients were over-projecting their voices and being recorded at a higher amplitude signal than controls. Interestingly, after matching audio duration and removing variables associated with intensity in order to mitigate the biases, the models were able to achieve a similar high performance. We provide a set of recommendations to avoid bias when building and evaluating machine learning models for screening in laryngology.

Neurodesk: an accessible, flexible and portable data analysis environment for reproducible neuroimaging.

Neuroimaging research requires purpose-built analysis software, which is challenging to install and may produce different results across computing environments. The community-oriented, open-source Neurodesk platform ( https://www.neurodesk.org/ ) harnesses a comprehensive and growing suite of neuroimaging software containers. Neurodesk includes a browser-accessible virtual desktop, command-line interface and computational notebook compatibility, allowing for accessible, flexible, portable and fully reproducible neuroimaging analysis on personal workstations, high-performance computers and the cloud.

The Past, Present, and Future of the Brain Imaging Data Structure (BIDS).

The Brain Imaging Data Structure (BIDS) is a community-driven standard for the organization of data and metadata from a growing range of neuroscience modalities. This paper is meant as a history of how the standard has developed and grown over time. We outline the principles behind the project, the mechanisms by which it has been extended, and some of the challenges being addressed as it evolves. We also discuss the lessons learned through the project, with the aim of enabling researchers in other domains to learn from the success of BIDS.

Identifying bias in models that detect vocal fold paralysis from audio recordings using explainable machine learning and clinician ratings.

Introduction: Detecting voice disorders from voice recordings could allow for frequent, remote, and low-cost screening before costly clinical visits and a more invasive laryngoscopy examination. Our goals were to detect unilateral vocal fold paralysis (UVFP) from voice recordings using machine learning, to identify which acoustic variables were important for prediction to increase trust, and to determine model performance relative to clinician performance. Methods: Patients with confirmed UVFP through endoscopic examination (N=77) and controls with normal voices matched for age and sex (N=77) were included. Voice samples were elicited by reading the Rainbow Passage and sustaining phonation of the vowel "a". Four machine learning models of differing complexity were used. SHapley Additive explanations (SHAP) was used to identify important features. Results: The highest median bootstrapped ROC AUC score was 0.87 and beat clinician's performance (range: 0.74 - 0.81) based on the recordings. Recording durations were different between UVFP recordings and controls due to how that data was originally processed when storing, which we can show can classify both groups. And counterintuitively, many UVFP recordings had higher intensity than controls, when UVFP patients tend to have weaker voices, revealing a dataset-specific bias which we mitigate in an additional analysis. Conclusion: We demonstrate that recording biases in audio duration and intensity created dataset-specific differences between patients and controls, which models used to improve classification. Furthermore, clinician's ratings provide further evidence that patients were over-projecting their voices and being recorded at a higher amplitude signal than controls. Interestingly, after matching audio duration and removing variables associated with intensity in order to mitigate the biases, the models were able to achieve a similar high performance. We provide a set of recommendations to avoid bias when building and evaluating machine learning models for screening in laryngology.

NIDM-Terms: community-based terminology management for improved neuroimaging dataset descriptions and query.

The biomedical research community is motivated to share and reuse data from studies and projects by funding agencies and publishers. Effectively combining and reusing neuroimaging data from publicly available datasets, requires the capability to query across datasets in order to identify cohorts that match both neuroimaging and clinical/behavioral data criteria. Critical barriers to operationalizing such queries include, in part, the broad use of undefined study variables with limited or no annotations that make it difficult to understand the data available without significant interaction with the original authors. Using the Brain Imaging Data Structure (BIDS) to organize neuroimaging data has made querying across studies for specific image types possible at scale. However, in BIDS, beyond file naming and tightly controlled imaging directory structures, there are very few constraints on ancillary variable naming/meaning or experiment-specific metadata. In this work, we present NIDM-Terms, a set of user-friendly terminology management tools and associated software to better manage individual lab terminologies and help with annotating BIDS datasets. Using these tools to annotate BIDS data with a Neuroimaging Data Model (NIDM) semantic web representation, enables queries across datasets to identify cohorts with specific neuroimaging and clinical/behavioral measurements. This manuscript describes the overall informatics structures and demonstrates the use of tools to annotate BIDS datasets to perform integrated cross-cohort queries.

A reproducible and generalizable software workflow for analysis of large-scale neuroimaging data collections using BIDS Apps.

Neuroimaging research faces a crisis of reproducibility. With massive sample sizes and greater data complexity, this problem becomes more acute. Software that operates on imaging data defined using the Brain Imaging Data Structure (BIDS) - BIDS Apps - have provided a substantial advance. However, even using BIDS Apps, a full audit trail of data processing is a necessary prerequisite for fully reproducible research. Obtaining a faithful record of the audit trail is challenging - especially for large datasets. Recently, the FAIRly big framework was introduced as a way to facilitate reproducible processing of large-scale data by leveraging DataLad - a version control system for data management. However, the current implementation of this framework was more of a proof of concept, and could not be immediately reused by other investigators for different use cases. Here we introduce the BIDS App Bootstrap (BABS), a user-friendly and generalizable Python package for reproducible image processing at scale. BABS facilitates the reproducible application of BIDS Apps to large-scale datasets. Leveraging DataLad and the FAIRly big framework, BABS tracks the full audit trail of data processing in a scalable way by automatically preparing all scripts necessary for data processing and version tracking on high performance computing (HPC) systems. Currently, BABS supports jobs submissions and audits on Sun Grid Engine (SGE) and Slurm HPCs with a parsimonious set of programs. To demonstrate its scalability, we applied BABS to data from the Healthy Brain Network (HBN; n=2,565). Taken together, BABS allows reproducible and scalable image processing and is broadly extensible via an open-source development model.

OME-Zarr: a cloud-optimized bioimaging file format with international community support.

A growing community is constructing a next-generation file format (NGFF) for bioimaging to overcome problems of scalability and heterogeneity. Organized by the Open Microscopy Environment (OME), individuals and institutes across diverse modalities facing these problems have designed a format specification process (OME-NGFF) to address these needs. This paper brings together a wide range of those community members to describe the cloud-optimized format itself-OME-Zarr-along with tools and data resources available today to increase FAIR access and remove barriers in the scientific process. The current momentum offers an opportunity to unify a key component of the bioimaging domain-the file format that underlies so many personal, institutional, and global data management and analysis tasks.

A guide to the BRAIN Initiative Cell Census Network data ecosystem.

Characterizing cellular diversity at different levels of biological organization and across data modalities is a prerequisite to understanding the function of cell types in the brain. Classification of neurons is also essential to manipulate cell types in controlled ways and to understand their variation and vulnerability in brain disorders. The BRAIN Initiative Cell Census Network (BICCN) is an integrated network of data-generating centers, data archives, and data standards developers, with the goal of systematic multimodal brain cell type profiling and characterization. Emphasis of the BICCN is on the whole mouse brain with demonstration of prototype feasibility for human and nonhuman primate (NHP) brains. Here, we provide a guide to the cellular and spatial approaches employed by the BICCN, and to accessing and using these data and extensive resources, including the BRAIN Cell Data Center (BCDC), which serves to manage and integrate data across the ecosystem. We illustrate the power of the BICCN data ecosystem through vignettes highlighting several BICCN analysis and visualization tools. Finally, we present emerging standards that have been developed or adopted toward Findable, Accessible, Interoperable, and Reusable (FAIR) neuroscience. The combined BICCN ecosystem provides a comprehensive resource for the exploration and analysis of cell types in the brain.

Neurodesk: An accessible, flexible, and portable data analysis environment for reproducible neuroimaging.

Neuroimaging data analysis often requires purpose-built software, which can be challenging to install and may produce different results across computing environments. Beyond being a roadblock to neuroscientists, these issues of accessibility and portability can hamper the reproducibility of neuroimaging data analysis pipelines. Here, we introduce the Neurodesk platform, which harnesses software containers to support a comprehensive and growing suite of neuroimaging software (https://www.neurodesk.org/). Neurodesk includes a browser-accessible virtual desktop environment and a command line interface, mediating access to containerized neuroimaging software libraries on various computing platforms, including personal and high-performance computers, cloud computing and Jupyter Notebooks. This community-oriented, open-source platform enables a paradigm shift for neuroimaging data analysis, allowing for accessible, flexible, fully reproducible, and portable data analysis pipelines.

OME-Zarr: a cloud-optimized bioimaging file format with international community support.

A growing community is constructing a next-generation file format (NGFF) for bioimaging to overcome problems of scalability and heterogeneity. Organized by the Open Microscopy Environment (OME), individuals and institutes across diverse modalities facing these problems have designed a format specification process (OME-NGFF) to address these needs. This paper brings together a wide range of those community members to describe the cloud-optimized format itself -- OME-Zarr -- along with tools and data resources available today to increase FAIR access and remove barriers in the scientific process. The current momentum offers an opportunity to unify a key component of the bioimaging domain -- the file format that underlies so many personal, institutional, and global data management and analysis tasks.

TemplateFlow: FAIR-sharing of multi-scale, multi-species brain models.

Reference anatomies of the brain ('templates') and corresponding atlases are the foundation for reporting standardized neuroimaging results. Currently, there is no registry of templates and atlases; therefore, the redistribution of these resources occurs either bundled within existing software or in ad hoc ways such as downloads from institutional sites and general-purpose data repositories. We introduce TemplateFlow as a publicly available framework for human and non-human brain models. The framework combines an open database with software for access, management, and vetting, allowing scientists to share their resources under FAIR-findable, accessible, interoperable, and reusable-principles. TemplateFlow enables multifaceted insights into brains across species, and supports multiverse analyses testing whether results generalize across standard references, scales, and in the long term, species.

Can machine learning identify childhood characteristics that predict future development of bipolar disorder a decade later?

Early identification of bipolar disorder may provide appropriate support and treatment, however there is no current evidence for statistically predicting whether a child will develop bipolar disorder. Machine learning methods offer an opportunity for developing empirically-based predictors of bipolar disorder. This study examined whether bipolar disorder can be predicted using clinical data and machine learning algorithms. 492 children, ages 6-18 at baseline, were recruited from longitudinal case-control family studies. Participants were assessed at baseline, then followed-up after 10 years. In addition to sociodemographic data, children were assessed with psychometric scales, structured diagnostic interviews, and cognitive and social functioning assessments. Using the Balanced Random Forest algorithm, we examined whether the diagnostic outcome of full or subsyndromal bipolar disorder could be predicted from baseline data. 45 children (10%) developed bipolar disorder at follow-up. The model predicted subsequent bipolar disorder with 75% sensitivity, 76% specificity, and an Area Under the Receiver Operating Characteristics of 75%. Predictors best differentiating between children who did or did not develop bipolar disorder were the Child Behavioral Checklist Externalizing and Internalizing behaviors, the Child Behavioral Checklist Total t-score, problematic school functions indexed through the Child Behavioral Checklist School Competence scale, and the Child Behavioral Checklist Anxiety/Depression and Aggression scales. Our study provides the first quantitative model to predict bipolar disorder. Longitudinal prediction may help clinicians assess children with emergent psychopathology for future risk of bipolar disorder, an area of clinical and scientific importance. Machine learning algorithms could be implemented to alert clinicians to risk for bipolar disorder.

Neuroscout, a unified platform for generalizable and reproducible fMRI research.

Functional magnetic resonance imaging (fMRI) has revolutionized cognitive neuroscience, but methodological barriers limit the generalizability of findings from the lab to the real world. Here, we present Neuroscout, an end-to-end platform for analysis of naturalistic fMRI data designed to facilitate the adoption of robust and generalizable research practices. Neuroscout leverages state-of-the-art machine learning models to automatically annotate stimuli from dozens of fMRI studies using naturalistic stimuli-such as movies and narratives-allowing researchers to easily test neuroscientific hypotheses across multiple ecologically-valid datasets. In addition, Neuroscout builds on a robust ecosystem of open tools and standards to provide an easy-to-use analysis builder and a fully automated execution engine that reduce the burden of reproducible research. Through a series of meta-analytic case studies, we validate the automatic feature extraction approach and demonstrate its potential to support more robust fMRI research. Owing to its ease of use and a high degree of automation, Neuroscout makes it possible to overcome modeling challenges commonly arising in naturalistic analysis and to easily scale analyses within and across datasets, democratizing generalizable fMRI research.

Microscopy-BIDS: An Extension to the Brain Imaging Data Structure for Microscopy Data.

The Brain Imaging Data Structure (BIDS) is a specification for organizing, sharing, and archiving neuroimaging data and metadata in a reusable way. First developed for magnetic resonance imaging (MRI) datasets, the community-led specification evolved rapidly to include other modalities such as magnetoencephalography, positron emission tomography, and quantitative MRI (qMRI). In this work, we present an extension to BIDS for microscopy imaging data, along with example datasets. Microscopy-BIDS supports common imaging methods, including 2D/3D, ex/in vivo, micro-CT, and optical and electron microscopy. Microscopy-BIDS also includes comprehensible metadata definitions for hardware, image acquisition, and sample properties. This extension will facilitate future harmonization efforts in the context of multi-modal, multi-scale imaging such as the characterization of tissue microstructure with qMRI.