Background: Pediatric heart disease (PHD), including congenital heart defects, is often incompletely captured in electronic health records, particularly when clinical significance must be inferred from unstructured echocardiogram reports. Automated methods capable of extracting clinically meaningful PHD from narrative reports could improve clinical decision support and research applications. Objective: The aim of the study is to evaluate the feasibility of using supervised fine-tuning of large language models (LLMs), with and without chain-of-thought (CoT) reasoning, to characterize patients with clinically significant or historical PHD from unstructured echocardiogram reports. Methods: We developed a PHD detection algorithm using fine-tuned open-source LLMs, including LLaMA (Meta) and Qwen (Alibaba), to analyze 9749 echocardiogram reports. A subset of 712 reports was adjudicated by 2 pediatric cardiac anesthesiologists, classifying 506 (71.1%) as clinically significant PHD and 206 (28.9%) as not significant. While DeepSeek R1 has shown improved performance with CoT reasoning, its application in medical contexts is underexplored. We incorporated R1-generated CoT into model prompts and fine-tuned backbone LLMs. Results: The fine-tuned Qwen-7B-10k-overthink-CoT achieved the highest accuracy (92.4%), outperforming Qwen-7B-without-CoT (90%), LLaMA-3B-without-CoT (87.9%), Qwen-3B-without-CoT (85.6%), Qwen-3B-10k-overthink-CoT (68.5%), and LLaMA-3B-10k-overthink-CoT (46.2%). In a second dataset, an external validation was performed (n=113; 64 positive, 49 negative), Qwen-7B-10k-overthink-CoT sustained a strong, balanced performance (82.7%), followed by Qwen-7B-without-CoT (88.4%), LLaMA-3B-without-CoT (86.8%), Qwen-3B-without-CoT (84.5%), Qwen-3B-10k-overthink-CoT (58.9%), and LLaMA-3B-10k-overthink-CoT (46.2%). The fine-tuned Qwen-7B model with overthinking CoT (10,000 tokens) achieved the highest internal accuracy (92.4%), with balanced sensitivity and specificity. Across repeated runs, CoT-enhanced models demonstrated improved classification consistency compared to non-CoT models (Qwen-7B-without-CoT: 90%, LLaMA-3B-without-CoT: 87.9%, Qwen-3B-without-CoT: 85.6%). In external validation (n=113), non-CoT variants achieved higher accuracy (up to 88.4%), whereas the Qwen-7B CoT model demonstrated more balanced class performance (accuracy=82.7%). Conclusions: Supervised fine-tuning of LLMs with CoT offers an effective approach for automated PHD detection within unstructured data in the electronic medical record. While CoT-enhanced models demonstrated improved internal performance and more balanced classification, they did not consistently achieve higher accuracy in external validation, highlighting trade-offs between accuracy and class balance. These findings highlight the promise of LLM-based approaches for clinical text phenotyping while underscoring the need for larger, multicenter validation and careful calibration for real-world deployment. Continued validation and integration into the electronic medical record are essential for real-world, artificial intelligence–driven clinical decision support.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/65aeb9ead7dafad2f061f602c5cfe3ed" />
Comparing the Accuracy of ChatGPT-4o, DeepSeek-V3, and Gemini 2.5 Flash in Answering Frequently Asked Questions About Systemic Lupus Erythematosus: Quantitative Study
Background: Systemic lupus erythematosus (SLE) is a complex, fluctuating disease, creating a continuous need for reliable patient information. A prior study concluded that patients with SLE often turn to the internet, including artificial intelligence (AI) chatbots, for information regarding SLE. The rise of AI chatbots as a primary information source presents a critical challenge regarding the accuracy of the information they provide. Objective: This study aimed to evaluate the performance of the latest generation of AI chatbots (ChatGPT-4o, DeepSeek-V3, and Gemini 2.5 Flash) in answering frequently asked questions about SLE. Methods: Twenty-two frequently asked questions about SLE in Bahasa Indonesia (the Indonesian language) were posed to each chatbot. Responses were independently and blindly evaluated for accuracy by 5 clinical immunologists using a 4-point Likert scale. Readability was assessed using the Flesch reading ease score formula. Statistical comparisons for accuracy and readability were performed using repeated-measures ANOVA or the Friedman test, followed by the Bonferroni test for pairwise comparisons. The Spearman ρ was used to evaluate correlations among accuracy, readability, and word count. Results: Gemini 2.5 Flash demonstrated the highest accuracy, with a mean score of 1.25 (SD 0.53), significantly outperforming ChatGPT-4o (mean 1.71, SD 0.61; <.001). Gemini 2.5 Flash significantly outperformed ChatGPT-4o in 2 evaluated domains. The interreliability analysis revealed a statistically significant level of agreement among the 5 evaluators across all responses (Kendall =0.389; <.001). Readability for all 3 chatbots was low (median Flesch reading ease score 42.22‐46.66). Gemini 2.5 Flash produced the longest responses (8509 total words), followed by DeepSeek-V3 (5410 words) and ChatGPT-4o (3632 words). A significant negative correlation was found between word count and lower accuracy (ρ=−0.401; =.001). Conclusions: Our study found that ChatGPT-4o, DeepSeek-V3, and Gemini 2.5 Flash provided overall satisfactory responses to SLE-related questions. The highest accuracy was demonstrated by Gemini 2.5 Flash; however, the absolute differences in scores among the 3 AI chatbots were relatively small. All 3 AI chatbots demonstrated low readability, which may limit accessibility for patient use. This finding highlights a critical “blind spot” in which clinical accuracy, as rated by experts, does not equate to patient accessibility. Thus, further research is required to develop more comprehensive evaluation frameworks incorporating safety, factuality, and calibration of AI chatbots across different medical fields and topics.
TARGA Imager Enables High-Resolution Imaging of Neurodevelopmental Models
Sponsored content brought to you by
Neurodevelopment in schizophrenia poses major challenges for experimental study due at least in part to the brain’s genetic complexity, cellular diversity, and limitations in accessing living human tissue. To overcome such barriers, researchers often use complementary human stem cell–derived models: adherent cortical organoids and Neurogenin-2 (NGN2) induced neurons. Adherent cortical organoids form three-dimensional cultures containing diverse cortical neuron types, enabling analysis of network development and long-term maturation over months.1 In contrast, NGN2 neurons generate rapid, two-dimensional, homogeneous populations of excitatory neurons that display robust activity within weeks, making them well suited for scalable, functional assays, and high-throughput screening.2
With the added insight that stem cell models offer into the neural development of the schizophrenic brain, the quantification of patient-derived neurons’ collective function is a priority.3 At Columbia University’s Mortimer B. Zuckerman Mind Brain and Behavior Institute, researchers use NGN2 neurons, yielding reproducible populations of excitatory cortical neurons that scale reliably across experiments.
Calcium imaging provides a powerful functional readout in these NGN2 neuron networks. When a neuron fires an action potential, voltage-gated calcium channels open and intracellular calcium rises sharply.4 Fluorescence calcium indicators convert transient, ionic changes into fluorescence emission that can be quantitatively detected via light microscopy across thousands of cells simultaneously. Coupling calcium-sensitive reporters with high-speed optical microscopy enables noninvasive, population-level measurement of neural activity, synchrony, and network dynamics. Interpreting these rich image sequences requires sophisticated theoretical and numerical approaches.5
Building on the ability to measure neural activity with calcium imaging at scale, Lumencor’s TARGA Imager represents a transformative step in the development of optical imaging hardware for the study of neurodevelopmental conditions such as schizophrenia. It is well suited to workflows where NGN2-neuron cultures are studied across multiple conditions in parallel (Figure 1). In these contexts, TARGA delivers calcium fluorescence images over millimeter-scale fields of view within standard 96-well plates, entire well areas, while maintaining high-speed, faster-than video rate imaging with precision resolution.
These capabilities allow researchers to observe chemical communications across large neuronal networks rather than isolated cells in real time. Images can be acquired at frequencies up to 100 Hz, enabling capture of fast calcium transients of collective neuronal dynamics. Rapid switching of multicolor excitation light supports multiplexed fluorescence dyes, linking functional activity with cellular structure and organization.
Overall, TARGA achieves high spatial, temporal, and spectral resolution simultaneously with precise, automated opto-mechanical architecture. These data are well matched to modern image analysis and AI algorithms, generating robust fluorescence traces from complex neuronal populations (Figure 2). In combination, these features make the TARGA Imager a revelatory neuroscience tool, uniquely enabling visualization of emergent collective behavior at millimeter scale with exceptional resolution. Such integrated performance accelerates discovery by bridging cellular mechanisms and systems-level phenotypes relevant to schizophrenia pathophysiology and therapeutic screening. By uniting scale, speed, and precision in a single optical platform, TARGA empowers researchers to probe experimental neuroscience and strengthens translational studies of complex psychiatric disease at population scale.
References
- Van der Kroeg et al. Human adherent cortical organoids in a multiwell format. eLife. 2024. 13:e98340.
- Shan et al. Fully defined NGN2 neuron protocol reveals diverse signatures of neuronal maturation. Cell Reports Methods. 2024. 4:100858.
- Rao et al. Aberrant pace of cortical neuron development in brain organoids from patients with 22q11.2 deletion syndrome‑associated schizophrenia. Nature Communications. 2025. 16:6986.
- Zhang et al. Fast and sensitive GCaMP calcium indicators for imaging neural populations. Nature. 2023. 615:884–891.
- Pasarkar et al. maskNMF: A denoise‑sparsen‑detect approach for extracting neural signals from dense imaging data. bioRxiv. 2023. 2023.09.14.557777.
To learn more, visit: lumencor.com or contact our team at: info@lumencor.com.
The post TARGA Imager Enables High-Resolution Imaging of Neurodevelopmental Models appeared first on Inside Precision Medicine.
360° Video-Based Virtual Reality for Preparing Medical Students for Body Donor Dissection: Randomized Controlled Trial
Background: Body donor dissection is fundamental to medical education but often induces anxiety and emotional distress in students, potentially impacting learning outcomes and well-being. Traditional preparation methods emphasize technical and procedural elements while inadequately addressing students’ emotional challenges. Recent advances in educational technology, particularly 360° video-based virtual reality (VR), may enhance students’ emotional readiness by providing immersive previews of dissection environments. However, the application of this technology specifically for emotional preparation for body donor dissection remains largely unexplored. Objective: This study aimed to develop and evaluate a 360° video-based VR application designed to enhance medical students’ emotional preparedness for their first body donor dissection experience. Methods: A randomized controlled longitudinal study was conducted with 43 first-year medical students (26/43, 60.5% female, mean age 20.9, SD 0.57 years) at Weill Cornell Medicine-Qatar in Fall 2025. Participants completed a baseline survey including the 40-item State-Trait Anxiety Inventory and were randomly assigned to intervention (n=22) or control (n=21) groups using computer-generated permuted block randomization. Before their first dissection session, the intervention group viewed a custom-designed 360° video-based VR experience that featured a virtual tour of the anatomy laboratory and a simulated first encounter with a body donor. The control group received no intervention. State-Trait Anxiety Inventory surveys were administered at baseline (survey 1, all participants), post-VR intervention (survey 2, intervention group only), and postfirst dissection (survey 3, all participants). A follow-up perception survey (survey 4) was administered to the intervention group 1 week into the dissection course. Data were analyzed using 2-tailed paired-samples and independent-samples tests, with qualitative responses analyzed using artificial intelligence–assisted thematic analysis. Results: The intervention group demonstrated a statistically significant reduction in trait anxiety (TA) immediately following the VR experience (mean difference 2.32, SD 4.95; =2.20; =.04), while the reduction in state anxiety (SA) was not significant (mean difference 2.41, SD 8.55; =1.32; =.20). No significant differences in SA or TA were found between intervention and control groups immediately before the first dissection session (SA: =0.03; =.98 and TA: =0.70; =.49) or in anxiety trajectories from baseline to postdissection (SA: =0.85; =.41 and TA: =0.46; =.65). Female students reported higher baseline TA compared to normative college populations (45.42 vs 40.40; mean difference 5.02, SD 7.72; =3.32; =.003). Qualitative analysis revealed positive perceptions, with 91% (10/11) reporting clear content and 82% (9/11) recommending it to future cohorts. Key perceived benefits included environmental familiarization, procedural understanding, and psychological preparation. Conclusions: The 360° video-based VR intervention significantly reduced TA and was perceived as valuable for emotional and procedural preparation. The intervention shows promise as a preparatory tool for enhancing emotional and procedural readiness; however, its impact on objective educational outcomes was not assessed and warrants further investigation. Trial Registration: ClinicalTrials.gov NCT07521033; https://clinicaltrials.gov/study/NCT07521033
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/2d931f56b90b14ca89868171f23345e7" />
Social Partner Effects on Type 2 Diabetes Prevention, Management, and Spillover Health Outcomes: Single-Arm Pre-Post Pilot Intervention
Background: South Asian Americans are at high risk of prediabetes and type 2 diabetes mellitus (T2DM). South Asian populations are typically close-knit communities, with support networks that could be leveraged in lifestyle interventions. Objective: This study was a single-arm, pre-post pilot study to evaluate the feasibility and efficacy of a culturally tailored telehealth intervention for South Asian adults with prediabetes or T2DM and their social partners (trusted household members) who agreed to complete preintervention and postintervention surveys. Methods: Participants attended 5-hour-long health education sessions delivered in English and Bengali. Participant outcomes included pre-post changes in hemoglobin A1c (HbA), BMI, blood pressure (BP), self-reported minutes of physical activity, and dietary choices at baseline and at the 6-month follow-up. For social partners, outcomes included pre-post survey changes in physical activity and dietary choices. We used Pearson chi-square tests and paired 2-tailed tests to compare baseline measures with postintervention outcomes. Results: This pilot study included 54 participants and 106 social partners in Atlanta, Georgia, between March 2021 and November 2023. All participants were Bangladeshi and spoke native Bengali. Social partners were most commonly participants’ children (39/106, 36.8%) or spouses (34/106, 32.1%). The participant baseline HbA level was 7.5% (SD 1.48%), which decreased by −0.83% (95% CI 0.42%-1.30%; <.001). Participants also improved systolic BP by −5.8 mm Hg (95% CI 0.196-11.37; =.04) with no change in diastolic BP (−0.451 mm Hg, 95% CI −1.49 to 2.39; =.60) or BMI (−0.642 kg/m, 95% CI −1.87 to 0.59; =.17). Compared with baseline, 39% more participants exercised at least 150 minutes weekly (<.001), but there was no difference in self-reported fruit and vegetable intake. However, the social partners increased fruit and vegetable intake (=.02), decreased soda intake (<.001), and increased daily moderate exercise (=.003). Conclusions: Including social partners in T2DM prevention and management is feasible and potentially beneficial, but comparative studies are needed to determine the incremental effects of social partners’ participation vs individual-focused lifestyle interventions. Trial Registration: ClinicalTrials.gov NCT05275231; https://clinicaltrials.gov/study/NCT05275231
<![CDATA[Lawsuits, research, and clinicians clash over youth screen addiction—learn more about how problematic social platform use evades diagnosis and how we can curb harm.]]>
STAT+: Pharmalittle: We’re reading about Lilly and Pfizer obesity drug data, Roche and J&J deals, and more
Good morning, everyone, and welcome to another working week. We hope the weekend respite was relaxing and invigorating because that oh-too-familiar routine of meetings, deadlines, and the like has returned with a vengeance. You knew this would happen, yes? To cope, we are relying, as always, on cups of stimulation. Our choice today is laced with chocolate raspberry. Feel free to join us. Remember, no prescription is required. Meanwhile, here are some tidbits to help you along. Best of luck accomplishing your goals today, and of course, do keep in touch. …
Revolution Medicines’ experimental pancreatic cancer drug has been the star of the oncology field in recent weeks, with new data showing the medicine produced unprecedented outcomes for patients, but its next act — this time as a co-lead — was just revealed, STAT points out. Tango Therapeutics said that in an early-stage clinical trial, a combination of its drug vopimetostat and Revolution’s daraxonrasib led to durable responses in the large majority of pancreatic cancer patients who received both medicines. Tango’s strategy of testing the two targeted drugs is notable because combination approaches in pancreatic cancer often include chemotherapy. The recent successful Revolution trial that has upended the specialty tested daraxonrasib versus chemotherapy as a second-line treatment.
Eli Lilly has already established that its next-generation obesity drug can lead to rapid weight loss, but researchers disclosed new data that provide more details on the safety and tolerability of the closely watched therapy, STAT writes. Lilly previously said that in one late-stage study, retatrutide helped people with diabetes lower blood sugar and lose a significant amount of weight, which is notable since those who have diabetes tend to lose less weight on treatment than those who do not. New data showed seven out of the 403 participants who received retatrutide experienced arrhythmias, and three treated participants experienced major cardiovascular complications, compared with none in the placebo group.
Gut microbiome dynamics in autism: a prospective nested case–control study demonstrates microbial-clinical associations following rehabilitation interventions
BackgroundChildren with autism spectrum disorder (ASD) commonly exhibit gut microbiota dysbiosis and metabolic abnormalities, yet the mechanisms linking these changes to clinical symptoms remain unclear.ObjectiveThis study employed a nested case–control design and multi-omics approaches to evaluate the effects of rehabilitation intervention on clinical symptoms and gut microbiota in children with ASD, identify distinct microbial-metabolic signatures, and explore their mechanistic links with sleep disorders and developmental abilities.MethodsWithin a prospectively established pediatric cohort (n = 45), we implemented a nested case–control design including 26 ASD children (18 males, 8 females; mean age 61.79 ± 11.15 months) and 19 age- and sex-matched healthy controls. All ASD participants received standardized rehabilitation therapy (2 h/day, 5 days/week for 6 months) comprising occupational therapy and cognitive-linguistic training. Primary outcomes included comprehensive clinical assessments [Griffiths Development Scales-Chinese (GDS-C), Children’s Sleep Habits Questionnaire (CSHQ), Autism Behavior Checklist (ABC), Childhood Autism Rating Scale (CARS)] and longitudinal multi-omics analysis (metagenomic sequencing and LC–MS-based metabolomics). Association analyses were performed with FDR correction (q < 0.05).ResultsFollowing the 6-month rehabilitation intervention, significant clinical improvements were observed in sleep quality (CSHQ total and subscores) and developmental performance (GDS-C). Multi-omics profiling revealed distinct biological signatures in ASD children compared to healthy controls, characterized by elevated Intestinibacter_bartlettii and reduced levels of ornithine and siderophore nonribosomal peptide biosynthesis. Crucially, correlation analysis demonstrated that, after FDR correction, ornithine levels were significantly positively correlated with multiple GDS-C developmental domains, while tyrosine was associated with parasomnias. These findings establish a potential mechanistic link where amino acid metabolism connects gut microbial shifts to clinical phenotypes.ConclusionThis study demonstrates that rehabilitation intervention synchronously ameliorates clinical symptoms and modulates the gut-metabolic profile in ASD. The identified associations between specific metabolites (ornithine and tyrosine) and clinical outcomes suggest a metabolic mechanism underlying the gut-brain axis, highlighting the potential of these metabolites as biomarkers for therapeutic monitoring. Further large-scale studies are needed to validate these findings.
Postnatal maternal mental health-related hospitalisation and its association with adverse child health and maltreatment outcomes: narrative review
BackgroundMaternal mental health conditions within the postnatal period can have long-lasting consequences on a child’s well-being. This study synthesises current evidence on the association between postnatal maternal mental health-related hospitalisation (MHrH) and adverse child health and maltreatment outcomes.MethodsWe conducted an inclusive search across multiple databases, including MEDLINE, PsycINFO, CINAHL, Scopus, Embase, Google Scholar, and the reference list of eligible papers. Studies that used standardised outcome measures were included. Study selection and data extraction were made using a standardised Joanna Briggs Institute (JBI) tool. The quality of the studies was assessed using the Newcastle-Ottawa Scale (NOS). Descriptive vote counting to map the direction of effects, alongside narrative thematic synthesis, was used to summarise the findings from the included studies.ResultsFour studies (n = 1,020,342) were included in the review. Three of these studies suggest a possible association between postnatal maternal MHrH and adverse child health outcomes (growth failure, respiratory and gastrointestinal infection) or maltreatment (sexual, physical, and emotional abuse). One study did not find a statistically significant association between postnatal maternal MHrH and infant growth.ConclusionThe available evidence suggests a possible association between postnatal maternal MHrH and adverse child health and maltreatment outcomes. However, given the small number of studies and methodological heterogeneity, these findings should be interpreted cautiously and highlight the need for further high-quality longitudinal research.Systematic Review Registrationhttps://www.crd.york.ac.uk/prospero/, identifier CRD42023446155.
Integrative mechanisms and intervention targets of the microbiota–gut–brain axis in depressive disorders: advances across immune, endocrine, and central nervous system pathways
Depressive disorders are highly heterogeneous syndromes characterized not only by depressed mood but also by cognitive impairment, sleep–circadian rhythm disturbances, altered appetite, somatic discomfort, and metabolic or gastrointestinal comorbidities. In recent years, the microbiota–gut–brain axis (MGBA) has been increasingly recognized as an integrative biological framework linking abnormalities in mood regulation, immune responses, endocrine function, metabolism, and neuroplasticity. This review provides a systematic synthesis of gut microbial ecology and host phenotypic features associated with depressive disorders, with particular emphasis on the depletion of short-chain fatty acid-producing commensals, the enrichment of potentially pro-inflammatory taxa, and the functional remodeling of key metabolic pathways, including the tryptophan–kynurenine pathway, short-chain fatty acids, bile acids, and trimethylamine N-oxide. We further discuss how bidirectional gut-to-brain and brain-to-gut communication may contribute to the onset and progression of depressive disorders through intestinal barrier disruption, low-grade systemic inflammation, hypothalamic–pituitary–adrenal axis activation, vagal signaling, and dysregulation of neurotransmitter and neurotrophic pathways. Current interventional evidence suggests that dietary and lifestyle modification, psychobiotics, and fecal microbiota transplantation may exert antidepressant potential in selected populations; however, the overall effect sizes remain limited and between-study heterogeneity is substantial. Patients with prominent gastrointestinal symptoms, metabolic abnormalities, or low-grade inflammatory states may represent priority candidates for MGBA-targeted interventions; nevertheless, a putative microbiota-responsive phenotype should not be simply equated with high stress exposure alone, and its definition requires prospective validation integrating stress burden, host responses, and microbial/metabolic readouts. Overall, MGBA research is gradually moving beyond descriptive profiling of microbial composition toward functional integration and clinical translation; however, causal inference, multi-omics standardization, and the identification of stratification biomarkers remain major challenges. Future studies should incorporate phenotype-based stratification, strengthened functional readouts, and precision intervention designs to determine which patients are most likely to benefit from microbiota-targeted therapies.