Background: Older adults facing social or structural marginalization for reasons such as lower literacy, digital exclusion, financial constraints, restricted living environments, or complex health histories, face persistent barriers to much-needed health screening. Digital health tools, particularly those using audio computer-assisted self-interview (ACASI) technology, offer potential to overcome these barriers (audio-delivered and self-administrable), but their application to marginalized populations remains underexplored. Moreover, guidance is limited for developing such tools which require collaboration within cross-disciplinary teams. This paper presents development insights and user testing findings from the ASCAPE (Audio App-Delivered Screening for Cognition and Age-Related Health in Prisoners) project, which aimed to develop equitable digital frailty and cognition screening for older people in Australian prisons. Objective: This study aims to describe the collaborative development of the “ASCAPE-HS” prototype, a tablet-based ACASI-delivered Frailty Index and aging screener, and to synthesize key lessons from the project that can inform equitable digital health tool development in hard-to-reach older adults. Also, to present findings on the usability and acceptability of ASCAPE-HS in a diverse community sample. Methods: The ASCAPE-HS prototype was developed through an iterative process involving researchers, clinicians, software developers, and end users under a digital health equity framework. The prototype included a self-administered, audio-delivered Frailty Index, alongside other items relevant to aging. The design process prioritized accessibility, sociocultural relevance, and technical feasibility, with regular multidisciplinary consultation and iterative refinement. Exploratory user testing with 20 older adults (aged 47‐93 years, including n=5 who had not finished secondary schooling, n=3 people with previous imprisonment history, and n=9 with mild or moderate cognitive impairment) provided feedback on usability and acceptability. Results: A 50-item Frailty Index was developed, alongside an additional selection of holistic questions that could meaningfully capture age-related health, and transferred to an iOS app (Apple, Inc), with ACASI features. Key elements included lay wording, consistent interface, simple “tapping” response options with repeatable audio feedback, a tutorial, and artificial intelligence–generated audio guidance. Key development considerations were synthesized into a checklist for teams undertaking similar projects. Successful strategies for the collaborative design process included diverse teams abreast of emerging literature and policy with varying expectations for engagement during development, and dedicating time to flexible, iterative development processes. Acceptability (median scores ≥4 out of 5 across 6 constructs) and usability (mean System Usability Scale score 79.0, SD 8.8) were high. Conclusions: A collaborative approach can produce ACASI-based health screening tools that are well-received by older adults. We highlight the feasibility of integrating frailty and aging assessment into a usable and acceptable digital tool, and offer actionable principles for collaborative, evidence-based development of equitable health screening tools in diverse, hard-to-reach populations.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/30699dc5ac2d0ad05f7755da3210033e" />
Machine Learning for Comparative Antidepressant Selection in Major Depressive Disorder: Systematic Review
Background: Major depressive disorder (MDD) affects approximately 1 in 6 adults during their lifetime, yet antidepressant selection relies predominantly on trial-and-error, with response rates of only 42% to 53%. While machine learning (ML) models have shown promise in predicting treatment outcomes, most focus on single treatments rather than comparative selection across therapeutic alternatives, limiting their clinical utility for the medication choice decisions that clinicians face in practice. Objective: This systematic review evaluates ML approaches that examine 2 or more pharmacological interventions for predicting treatment outcomes in MDD, with a focus on their capacity to facilitate comparative treatment selection between medications or medication classes for individual patients. Methods: Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, we searched PubMed, Scopus, and Web of Science for studies published from 2015 to 2025. We included studies involving adults with MDD that used ML models to predict treatment outcomes across 2 or more pharmacological treatments and reported medication-specific prediction outcomes. Risk of bias was assessed using PROBAST-AI (Prediction Model Risk of Bias Assessment Tool for Artificial Intelligence). We conducted a narrative synthesis organized by modeling strategies, data integration approaches, validation methodologies, and performance patterns. Results: From 5370 initial records, 19 studies met the inclusion criteria, with dataset sample sizes ranging from 49 to 77,226 participants. Studies employed 3 distinct modeling strategies: drug-specific supervised models trained independently for each medication, subtype- or trajectory-based approaches using clustering methods to identify differential response patterns, and a unified differential prediction framework generating calibrated cross-treatment predictions. Performance varied substantially, with area under the curve values ranging from 0.59 to 0.95 and classification accuracies between 62% and 95.4%, though high performance was concentrated in studies with small samples, high-dimensional neurobiological features, and internal-only validation. Only 7 studies conducted external validation, which generally yielded more conservative performance estimates. Feature informativeness was more consistently associated with performance variation than algorithm complexity. Most studies did not formally distinguish between prognostic features predicting general outcomes and predictive features identifying differential medication responses, and none applied formal explainability techniques. Conclusions: ML for comparative antidepressant selection remains in an early stage of development. Only 1 study implemented a unified framework directly supporting patient-level treatment ranking. Key barriers to clinical translation include insufficient distinction between prognostic and predictive markers, limited cross-trial validation, near-absent calibration reporting, and absent explainability. Future research should prioritize unified comparative frameworks with calibrated predictions, rigorous external validation on diverse cohorts, explicit modeling of heterogeneous treatment effects, and integration of explainability into model development.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/a59ed16526360bac492b29bed9672c02" />
iCARE Self-Guided Digital Intervention for Postpartum Depression in Danish Mothers: Formative Research Using User-Centered Design
<strong>Background:</strong> Postpartum depression (PPD) is a major public health concern. Despite advancements in treatment, many barriers to accessing care remain. There has been a growing interest in digital interventions for the prevention and treatment of PPD. However, for mothers with mild and moderate symptoms of depression, there is a limited offer of self-guided internet-based interventions developed with user input and with considerations on how to integrate the intervention into stepped care models for PPD. <strong>Objective:</strong> The aim of this study was (1) to describe the process of the design and development of iCARE, a self-guided digital psychological intervention for mothers with mild and moderate symptoms of PPD in Denmark, (2) present the program’s theory illustrated by a logic model, and (3) explore its initial usability and prospective acceptability. <strong>Methods:</strong> Applying user-centered design methods, the intervention development followed six steps: (1) a literature review to identify evidence‑based therapeutic components of self‑guided interventions for PPD, (2) interviews with women with lived experience of PPD and group discussions with mental health experts and home‑visiting providers to identify user needs, (3) iterative design and content development with stakeholder feedback in collaboration with the Department of Digital Psychiatry, (4) prototype testing using think‑aloud usability sessions and interviews with 5 mothers, (5) a group cognitive walkthrough with mental health experts, and (6) final refinement and implementation of the iCARE program with developers and designers. <strong>Results:</strong> Initial interviews with mothers and maternal health care providers emphasized the importance of a digital intervention offering timely psychoeducation, coping strategies, and pathways to in-person care while addressing the diversity of expressions of PPD symptoms. Stakeholders recommended a flexible program, multimodal content, and integration into maternal care systems with community health nurses supporting engagement and participation. The prototype was designed to be user-centered, engaging, and with multiple interactive features. It included components on psychoeducation, cognitive exercises grounded in cognitive behavioral therapy, acceptance and commitment principles, and mood-monitoring. The prototype was designed to be user-centered and engaging, with interactive features and components on psychoeducation, cognitive exercises grounded in cognitive behavioral and acceptance and commitment principles, and mood-monitoring. Prototype testing indicated high prospective acceptability and led to refinements across 6 themes: appropriateness of content; motivation and engagement; inclusivity and gender representation; clarity of instructions and data use; understanding of therapeutic method; and usability, layout, and navigation. <strong>Conclusions:</strong> iCARE is a self-guided internet-based psychological intervention for mothers with mild and moderate symptoms of PPD in Denmark. It was developed with user input by using qualitative methods, user-centered design, and psychological theory. Further research is needed to evaluate the feasibility and effectiveness of the program in a randomized controlled trial and its integration into maternal health care models such as universal PPD screening and home-visiting.
Early Detection of Alzheimer’s Disease and Related Dementias From Spontaneous Speech Using Foundation Speech and Language Models: Comparative Evaluation
<strong>Background:</strong> Alzheimer’s disease and related dementias (ADRD) are progressive neurodegenerative conditions where early detection is critical for timely intervention and care planning. However, current diagnostic methods are often inaccessible, costly, and delayed, especially for underserved populations. There is a growing need for scalable, noninvasive tools that can support timely diagnosis. Spontaneous speech contains rich acoustic and linguistic markers that can serve as noninvasive behavioral markers for cognitive decline. Foundation models, pretrained on large-scale audio or text data, generate high-dimensional embeddings that encode rich contextual and acoustic information. <strong>Objective:</strong> This study benchmarks open-source foundation language and speech models to evaluate their effectiveness in detecting ADRD from spontaneous speech as a potential solution for early, noninvasive, and scalable ADRD detection. <strong>Methods:</strong> In this study, we used the Pioneering Research for Early Prediction of Alzheimer’s and Related Dementias EUREKA (PREPARE) Challenge dataset, which consists of audio recordings from over 1600 participants with 3 distinct categories of cognitive decline: healthy control (HC), mild cognitive impairment (MCI), and Alzheimer’s disease (AD). We further excluded samples that are non-English, nonspontaneous speech, or of poor quality. Our final samples included 703 (59.13%) HC, 81 (6.81%) MCI, and 405 (34.06%) AD cases. We systematically benchmarked 18 open-source foundation speech and language models to classify cognitive status into 3 categories (HC, MCI, or AD). Post hoc interpretability analysis was performed for the best-performing model using Shapley additive explanations linking high-dimensional embeddings with explainable acoustic and linguistic markers. <strong>Results:</strong> Whisper-medium model achieved the highest performance among speech models at 0.731 accuracy and 0.802 area under the curve, while Bidirectional Encoder Representations from Transformers with pause annotation achieved the top accuracy of 0.662 and 0.744 area under the curve among language models. Overall, ADRD detection based on state-of-the-art automatic speech recognition model-generated audio-embeddings outperformed other models, and the inclusion of nonsemantic information, such as pause patterns, consistently improved the classification performance of text-embedding–based models. <strong>Conclusions:</strong> Our work presents a comprehensive comparative evaluation of state-of-the-art speech and language models for AD and MCI detection on a large, clinically relevant dataset. Embeddings derived from acoustic models, which capture both semantic and acoustic information, show promising performance and highlight the potential for developing a more scalable, noninvasive, and cost-effective early detection tool for ADRD.
The Download: making drugs in orbit and NASA’s nuclear-powered spacecraft
This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
A plan to make drugs in orbit is going commercial
A startup called Varda Space Industries is betting that the future of pharmaceuticals lies in orbit. The company has signed a deal with United Therapeutics to test whether drugs crystallize differently in microgravity, potentially creating improved versions with new properties.
The idea sounds futuristic, but falling launch costs and reusable rockets are making space-based manufacturing seem increasingly plausible. Varda says the partnership could mark an important step toward building products in orbit for use back on Earth.
Discover how space could become the next frontier for drug development.
—Antonio Regalado
MIT Technology Review Narrated: NASA is building the first nuclear reactor-powered interplanetary spacecraft. How will it work?
Just before Artemis II began its historic slingshot around the moon, NASA revealed an even grander space travel plan. By the end of 2028, the agency aims to fly a nuclear reactor-powered interplanetary spacecraft to Mars.
A successful mission would herald a new era in spaceflight—and might just give the US the edge in the race against China. But the project remains shrouded in mystery.
MIT Technology Review picked the brains of nuclear power and propulsion experts to find out how the nuclear-powered spacecraft might work.
—Robin George Andrews
This is our latest story to be turned into an MIT Technology Review Narrated podcast, which we publish each week on Spotify and Apple Podcasts. Just navigate to MIT Technology Review Narrated on either platform, and follow us to get all our new content as it’s released.
The must-reads
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
1 Sam Altman claims Elon Musk tried to seize control of OpenAI
Altman said Musk initially wanted 90% of the equity. (AFP)
+ And that control should go to his children when he dies. (BBC)
+ Altman also accused Musk of twice trying to end its non-profit status. (NPR)
+ Musk’s motivations for the suit are under scrutiny. (MIT Technology Review)
2 Google and SpaceX are in talks to launch data centers into orbit
SpaceX could join Suncatcher, Google’s orbital data center project. (WSJ $)
+ The project’s first launch is slated for early 2027. (Guardian)
+ Anthropic and SpaceX have also discussed orbital data centers. (Wired $)
+ But there are a few hurdles to overcome. (MIT Technology Review)
3 Jensen Huang has joined Donald Trump’s high-stakes mission to China
Nvidia is lobbying to sell its AI chips in the country. (Bloomberg $)
+ Elon Musk and Tim Cook are also on the trip. (CNBC)
+ But a tech rivalry and distrust have sapped hopes for big deals. (Reuters $)
4 ICE agents have a list of 20 million people on their iPhones, thanks to Palantir
An ICE official said Palantir is speeding up raids and arrests. (404 Media)
+ ICE has also used facial recognition and Paragon spyware. (TechCrunch)
5 Defense tech firm Anduril just doubled its valuation to over $60 billion
In a $5 billion funding round led by Thrive Capital and a16z. (FT $)
Anduril, which makes AI-backed weapons, may go public next year. (NYT $)
6 Meta employees are protesting computer-tracking at work
Flyers posted at offices are urging staff to oppose the program. (Reuters $)
+ Meta plans to track workers’ clicks and keystrokes to train AI. (CNBC)
7 OpenAI is facing another wrongful death lawsuit over ChatGPT medical advice
The chatbot’s tips allegedly led to a teenager’s overdose. (Ars Technica)
8 The Canvas learning platform has paid hackers to delete stolen student data
It caved to ransomware demands after the biggest-ever edtech breach. (BBC)
9 Scientific researchers are thinking twice about using AI
Due to price hikes, usage limitations, and unreliable outputs. (Nature)
10 The latest AI compute solution? Putting data centers in your home
Hardware hosts get subsidized electricity and internet. (Ars Technica)
Quote of the day
“Mr Musk did try to kill it.”
—Sam Altman claims that Elon Musk tried to destroy rather than protect OpenAI’s non-profit operations, the Guardian reports.
One More Thing
YOSHI SODEOKA
Why does AI hallucinate?
Chatbot fails are now a familiar meme. Meta’s short-lived scientific chatbot generated wiki articles about the history of bears in space. Lawyers have submitted court documents filled with legal citations fabricated by ChatGPT. Air Canada was ordered to honor a refund policy invented by its customer service chatbot.
This tendency to make things up—known as hallucination—is one of the biggest obstacles holding chatbots back from more widespread adoption. Here’s why they do it—and why we still can’t fix it.
—Will Douglas Heaven
This story is part of MIT Technology Review Explains, our series untangling the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here.
We can still have nice things
A place for comfort, fun, and distraction to brighten up your day. (Got any ideas? Drop me a line.)
+ A historian has unearthed the etymology of every single dinosaur name.
+ Humus on the moon is getting closer to reality after scientists grew chickpeas in lunar soil.
+ Witness the patience of a master paper artist in this gallery of intricate, handmade sculptures.
+ Want to tell the time alphabetically? Me neither, but this cursed clock is an intriguing reason to try.
Roles of NRXN1 in neuropsychiatric disorders: from genetic lesion to molecular mechanism
Numerous neuropsychiatric disorders frequently exhibit overlapping genetic risk factors, implying the molecular basis for their comorbidity. Nevertheless, the pathogenesis of these disorders remains elusive, particularly regarding how genetic variations impair the physiological function of risk genes and contribute to disease phenotypes. Neurexin 1 protein, encoded by NRXN1 gene, belongs to the neurexin family of presynaptic adhesion molecules. And neurexin 1 is involved in synaptogenesis and the maintenance of synaptic action. Genetic variations of NRXN1 have been demonstrated to be associated with a spectrum of neuropsychiatric disorders. Herein, this review focuses on the most recent and relevant literature concerning the genetic and molecular mechanisms through which NRXN1 variants contribute to the pathogenesis of neuropsychiatric disorders, particularly schizophrenia and autism spectrum disorder. Among them, we propose the isoform-dependent excitation-inhibition imbalance hypothesis of NRXN1 in autism spectrum disorder. And this hypothesis may account for both the elevated and decreased excitation-inhibition ratios observed in diverse individuals with autism spectrum disorder. Moreover, both schizophrenia and autism spectrum disorder involve deletions and alternative splicing of NRXN1, offering molecular evidence for their comorbidity. Then, we analyzed and summarized the current research status of NRXN1 in other neuropsychiatric disorders, including attention-deficit hyperactivity disorder, insomnia, epilepsy, suicide, and depression. Additionally, available limited researches on NRXN1-targeted therapeutic strategies and associated pharmacological studies are also incorporated. Finally, we discussed existing challenges in NRXN1 research within the context of neuropsychiatric disorders and proposed potential avenues to overcome these obstacles.
Direct modulation of human GABA-A α1β2γ2 receptors by the endocannabinoid 2-arachidonoylglycerol: implications for cannabinoid-related ligands and limitations for anxiolytic drug development
Anxiety disorders are associated with impaired inhibitory neurotransmission mediated by γ-aminobutyric acid type A (GABA-A) receptors. Although benzodiazepines remain effective anxiolytics, their clinical utility is limited by sedation, cognitive impairment, tolerance, and dependence, prompting the search for mechanistically distinct GABAergic modulators. Among cannabinoid-related molecules, the strongest evidence for direct GABA-A receptor modulation concerns the endocannabinoid 2-arachidonoylglycerol (2-AG), which potentiates recombinant human α1β2γ2 receptors through residues located in the M4 helix of the β2 subunit. Here, we review the structural architecture, biophysical properties, and pharmacological profile of the human GABA-A α1β2γ2 isoform as the relevant molecular framework for evaluating this mechanism, while discussing the broader relevance of cannabinoid-related ligands and selected phytocannabinoids without assuming mechanistic equivalence. We further assess the hypothesis that 2-AG reaches the β2-M4 site through a membrane-access route and identify five conceptual barriers that currently limit translation of this mechanism into anxiolytic drug development: supraphysiological effective concentrations, unresolved synaptic-versus-extrasynaptic actions, uncertain subtype selectivity, incomplete validation of lipid-environment effects, and lack of clinical evidence linking this mechanism to anxiolysis in humans. We conclude that direct modulation through β2-M4 defines a mechanistically intriguing allosteric pathway distinct from benzodiazepine action; however, its location on a shared β2 subunit and the micromolar concentrations required for modulation represent substantial obstacles to the rational design of anxioselective agents based on this mechanism.
Federated training of spiking neural networks on edge hardware for audio processing
Spiking Neural Networks have caught significant attention recently for their potential for energy-efficient computation on neuromorphic hardware and their event-driven processing. Spiking Neural networks employ spike-based learning paradigms, which require specialized training procedures such as Surrogate Gradient Descent. At the same time, Federated Learning allows collaborative model training on decentralized devices with preservation of data privacy protection. However, to date, few research has examined the suitability of Federated learning with ARM-based hardware. This work primarily investigates whether Federated Spiking Neural Networks training on ARM-based hardware is feasible with the Raspberry Pi 5 as a widely available and low-cost edge computing device for audio signal processing tasks. We perform a comparative analysis of federated Spiking Neural Network and federated convolutional neural networks on ARM processors and evaluate their performance on different data partitioning strategies using Dirichlet-based splits and various federated averaging algorithms. Using Federated learning, this work investigates the impact of data heterogeneity and aggregation strategies on model convergence, communication overhead, and latency in distributed training paradigms. The results provided showcases the important insights into the trade-offs of FL-SNN implementations on Von Neumann architectures and their applications in decentralized neuromorphic computing for audio processing.
Editorial: Research on the correlative mechanisms and clinical exploration of headache and cerebrovascular diseases
Assessing directional connections between symptoms, cognition, insight, and real-life functioning in schizophrenia: a partial ancestor graphs analysis
IntroductionSchizophrenia is a severe chronic mental disorder causing significant global disability. Understanding the intricate relationships between symptoms, cognitive functions, and real-life outcomes is essential for developing effective interventions. Prior research, while informative, could not often determine the direction of the association between these illness-related factors. This study aimed to investigate the possible causal connections within the interrelationships of these variables. MethodsThis cross-sectional study included 215 clinically stable patients diagnosed with schizophrenia. Comprehensive assessments covered psychopathology, neurocognition, social cognition, metacognition, clinical insight, and real-life functioning. Causal relationships were explored using Partial Ancestral Graphs, a causal discovery framework that accounts for mediators and confounders. The Greedy Fast Causal Inference algorithm was employed with 1,000 bootstrap replications to assign edge orientations.ResultsA central neurocognitive–metacognitive–functional system of directed connection emerged: visual learning was linked to attention/vigilance and working memory. Working memory showed a direct relationship with metacognition, which, in turn, was connected to real-life functioning. Two partly independent contributions to real-life functioning were also identified: conceptual disorganization and experiential negative symptoms, which were directly related to expressive deficits. Positive symptoms, depressive symptoms, and social cognition occupied peripheral positions, showing no significant connection with other variables. Unawareness and misattribution of symptoms showed an indeterminate association disconnected from the main network.DiscussionThe findings show a set of directed associations that start with neurocognitive abilities, pass through working memory and metacognition, and terminate in real-life functioning. Independently, conceptual disorganization and expressive negative symptoms also exert direct influences. These directed systems of connections provide operational guidance for clinical practice, highlighting critical targets for interventions such as cognitive remediation focused on working memory, metacognitive therapies, and strategies addressing disorganization and avolition, all aimed at improving real-life outcomes in schizophrenia.