This week’s Artificial Intelligence and Digital Health update highlights regulatory milestones, expanding CNS and cardiovascular AI pipelines, growing wearable-driven insights, and broader deployment across clinical workflows and population health.
In Today’s Newsletter
Dive deeper
🧪 MOLEXA reconstructs molecules from Coulomb explosions [1] [US • 23 Mar 2026]
https://www.chemeurope.com/en/news/1188335/ai-rebuilds-molecules-from-exploding-fragments.html
Context: SLAC National Accelerator Laboratory, European XFEL, and collaborators trained MOLEXA on simulated Coulomb explosion data, then tested it on water, tetrafluoromethane, and ethanol.
Key point: Source says the model reconstructed small-molecule geometries from ion momentum data, with prediction error reduced through two-step training and experimental reconstructions largely overlapping known structures.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧠 Lunai Bioworks adds BBB platform and Alzheimer’s assets [2] [28 Mar 2026]
Context: Lunai Bioworks described the deal as Series B Convertible Preferred equity with a fixed conversion price and a 19.9% beneficial ownership cap.
Key point: Company announced a $20 million strategic transaction to acquire a blood-brain barrier delivery platform plus CNS Alzheimer’s assets from the Clemann Group.
Implication: Signals pipeline investment and modality expansion.
🤝 Insilico Medicine and Tenacia expand CNS AI collaboration [3] [26 Mar 2026]
Context: The initial March 2025 program targeted small molecules with blood-brain barrier permeability for CNS disease.
Key point: Insilico Medicine and Tenacia Biotechnology expanded their partnership to develop an additional CNS candidate to preclinical candidate stage, with added deal value up to US$94.75 million.
Implication: Signals pipeline investment and modality expansion.
❤️ Noah Labs wins FDA Breakthrough Device Designation for Vox [4] [US • 27 Mar 2026]
Context: Noah Labs said Vox uses a 5-second voice clip and has been validated across multicentre clinical trials with academic partners.
Key point: FDA granted Breakthrough Device Designation to Noah Labs Vox, a voice-based algorithm intended to detect worsening heart failure before hospitalization.
Implication: May influence prescriber choice and payer reviews pending full data.
🫁 Mayo Clinic links wearable sleep data to COPD rehab engagement [5] [US • 26 Mar 2026]
Context: Researchers combined baseline sleep data from a wrist monitor with machine learning and clinical indicators in a 12-week home pulmonary rehabilitation setting.
Key point: Mayo Clinic reported that adding a Composite Sleep Health Score improved prediction of patient participation in remote COPD rehabilitation.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🫀 Viz.ai and Alnylam target earlier cardiac amyloidosis detection [6] [US • 24 Mar 2026]
Context: Viz.ai said the pathway will use its FDA-cleared echocardiography algorithm, Us2.ai, within a care coordination workflow.
Key point: Viz.ai and Alnylam Pharmaceuticals partnered to build an AI care pathway for identifying and evaluating cardiac amyloidosis earlier.
Implication: May expand screening, initiation, and follow-up at scale.
🧠 Neurophet joins ALZ-NET for Alzheimer’s imaging workflows [7] [US/South Korea • 27 Mar 2026]
Context: ALZ-NET is sponsored by the Alzheimer’s Association and operated by the American College of Radiology.
Key point: Neurophet signed an MoU with ALZ-NET to deploy FDA-cleared imaging software for monitoring amyloid-related imaging abnormalities and supporting treatment decisions.
Implication: May expand screening, initiation, and follow-up at scale.
🧲 Penn foundation model reads cardiac MRI at specialist level [8] [27 Mar 2026]
Context: University of Pennsylvania researchers trained the model on more than 300,000 MRI video clips from about 20,000 patients, according to the source.
Key point: Study summary said the model matched specialists on heart-function assessment, diagnosed 39 cardiac conditions, and flagged previously undiagnosed hypertrophic cardiomyopathy cases in real-world screening.
Implication: May influence prescriber choice and payer reviews pending full data.
🏥 Philips gets FDA clearance for DeviceGuide in mitral TEER [9] [US • 26 Mar 2026]
https://www.massdevice.com/fda-clears-philips-ai-heart-valve-repair-guidance/
Context: EchoNavigator R5.0 with DeviceGuide combines live echo and X-ray with AI visualization for minimally invasive mitral valve repair.
Key point: Philips announced FDA 510(k) clearance for AI-powered DeviceGuide to assist physicians during mitral transcatheter edge-to-edge repair procedures.
Implication: May influence prescriber choice and payer reviews pending full data.
🗣️ Noah Labs heart-failure voice RPM adds EU pathway details [10] [US • 26 Mar 2026]
Context: Source says Vox is also undergoing EU MDR certification and cites PRE-DETECT-HF as supporting evidence.
Key point: Noah Labs said Breakthrough Device Designation could accelerate US regulatory and commercial steps for its voice-based heart-failure monitoring software.
Implication: May influence prescriber choice and payer reviews pending full data.
🇰🇷 Dong-A ST broadens digital healthcare stack [11] [South Korea • 26 Mar 2026]
https://www.chosun.com/english/industry-en/2026/03/26/5DTGZGYVMZGORJLNS2MG2HDY64/
Context: The portfolio described includes HiCardi remote monitoring, Doctor Eye retinal AI tools, CareSens Air CGM, and workflow products such as Saelok, Miribom, and Yakmeokja.
Key point: Dong-A ST positioned digital healthcare as a growth engine spanning diagnostics, monitoring, and chronic disease management, with overseas expansion also cited.
Implication: May expand screening, initiation, and follow-up at scale.
💍 Oura highlights serious-illness signals from wearable trends [12] [US • 28 Mar 2026]
Context: Oura’s chief medical officer discussed Symptom Radar, cardiovascular age, and studies underway through Oura Labs.
Key point: Oura said repeated baseline-deviation alerts have helped prompt medical evaluation in multiple lymphoma cases and other conditions, though the ring does not diagnose disease.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
💊 eMed raises capital for employer-focused GLP-1 programs [13] [US • 28 Mar 2026]
https://technotrenz.com/news/emed-digital-healthcare-emed-just-raised-200m/
Context: Source says eMed will use funding for its AI platform and a capitated payment model tied to employer-sponsored obesity and metabolic programs.
Key point: eMed announced a $200 million Series A at a valuation above $2 billion, centered on AI-enabled population health and GLP-1/GIP care pathways.
Implication: DTC/telehealth: Could streamline initiation and adherence via remote prescribing and logistics.
💬 Study maps Chinese acceptance of LLM healthcare chatbots [14] [27 Mar 2026]
https://journals.sagepub.com/doi/10.1177/20552076261437614
Context: The survey study analysed 502 respondents using an extended UTAUT framework and tested moderation by prior telemedicine and LLM experience.
Key point: Performance expectancy, social influence, trust, and facilitating conditions were significant, while prior LLM experience, not telemedicine experience, showed meaningful moderating effects.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧬 Longitudinal digital health platform study links biomarkers, wearables, genetics [15] [24 Mar 2026]
https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001271
Context: Retrospective analysis covered more than 20,000 digital platform users and integrated blood biomarkers, wearables, and polygenic risk scores.
Key point: Users with suboptimal biomarkers showed sustained improvement over time in several measures, while step counts, REM sleep, and genetic risk correlated with the degree of improvement.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🩹 AI Healing Index beats PAR in wound-healing prediction [16] [25 Mar 2026]
https://bmjdigitalhealth.bmj.com/content/2/1/e000069
Context: Retrospective study evaluated 173,816 wounds across pressure injuries, venous ulcers, diabetic foot ulcers, and arterial ulcers.
Key point: The AI-powered Healing Index reached about 65% balanced accuracy by week 3, while percent area reduction reached similar performance at week 4.
Implication: May influence prescriber choice and payer reviews pending full data.
🦟 GAN imputation improves wearable malaria detection under data loss [17] [27 Mar 2026]
https://www.nature.com/articles/s41746-026-02518-4
Context: The study used rural Kenya wearable data with 50% data coverage and trained the imputation model on external COVID-19 data.
Key point: GAN-based imputation improved early infection alerting, including alerts that would otherwise have been missed because of missing heart-rate data.
Implication: May expand screening, initiation, and follow-up at scale.
📱 Mobile biometrics predict symptom shifts in complex chronic illness [18] [24 Mar 2026]
https://www.nature.com/articles/s41746-026-02543-3
Context: Analysis included 4,244 users and paired morning PPG biometrics with evening reports of crash, fatigue, and brain fog.
Key point: Within-person rises in heart rate and drops in heart-rate variability were associated with worse same-day symptoms, and adding biometrics improved predictive performance over prior symptoms alone.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧾 OpenEvidence adds coding suggestions inside clinical documentation [19] [US • 26 Mar 2026]
Context: The feature sits inside OpenEvidence Visits and suggests ICD-10, E/M, and CPT codes with rationale.
Key point: OpenEvidence expanded from clinical search and visit support into coding assistance, targeting documentation-to-claim workflows.
Implication: May streamline initiation and adherence via remote prescribing and logistics.
🚫 Irish mental health bodies call for ban on AI “therapy” [20] [Ireland • 27 Mar 2026]
https://www.imt.ie/news/mental-health-groups-call-for-ai-therapy-ban-27-03-2026/
Context: The College of Psychiatrists of Ireland, Psychological Society of Ireland, and Irish Association for Counselling and Psychotherapy jointly issued the position.
Key point: The groups urged legislation to ban AI therapy, arguing chatbots can mis-handle self-harm, suicide risk, delusions, and human relational needs.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🇳🇿 New Zealand’s diabetes roadmap includes AI and digital screening [21] [New Zealand • 27 Mar 2026]
https://www.healthcareitnews.com/news/anz/new-zealand-explores-ai-diabetes-roadmap-and-more-briefs
Context: Te Whatu Ora’s roadmap also references diabetic retinopathy screening, community retinal photography pilots, and HbA1c threshold alignment.
Key point: New Zealand placed AI and digital tools into a national diabetes roadmap focused on prevention, screening, and scaled management.
Implication: May expand screening, initiation, and follow-up at scale.
🧪 OpenAI Foundation outlines healthcare and drug-development push [22] [26 Mar 2026]
Context: The announcement described support for disease-pathway mapping, biomarker detection, drug repurposing, and open public-health data.
Key point: OpenAI Foundation said it will channel major funding into healthcare applications, including AI-driven life-sciences research and therapy development.
Implication: Signals pipeline investment and modality expansion.
🩺 MIT group proposes “humble” AI for clinical decision support [23] [US • 24 Mar 2026]
https://news.mit.edu/2026/creating-humble-ai-0324
Context: The BMJ Health and Care Informatics paper argues for AI systems that disclose uncertainty and prompt clinicians to gather more information when needed.
Key point: MIT-led researchers proposed a framework for clinical AI that is more collaborative and uncertainty-aware, rather than overly authoritative.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧠 Meta’s TRIBE v2 predicts brain responses to images, sound, and speech [24] [27 Mar 2026]
Context: Source says the model was trained on more than 1,000 hours of fMRI data from 720 subjects and predicts voxel-level responses.
Key point: Meta reported that TRIBE v2 often matched average brain-response patterns better than single-person scans and replicated known neuroscience findings in silico.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧠 Speech LLM adaptation study lifts dementia-screening performance [25] [26 Mar 2026]
https://ai.jmir.org/2026/1/e82608
Context: Researchers evaluated multiple text-only and multimodal models on DementiaBank speech datasets across in-context learning, reasoning, and fine-tuning strategies.
Key point: Token-level fine-tuning gave the strongest results overall, while prototype-style demonstrations and some reasoning strategies also improved smaller models.
Implication: May influence prescriber choice and payer reviews pending full data.
👁️ KongMing predicts anti-VEGF response in neovascular AMD [26] [26 Mar 2026]
https://www.thelancet.com/journals/landig/article/PIIS2589-7500(25)00153-0/fulltext
Context: Prospective multicentre study in China used OCT images and BCVA outcomes to build a transformer-based multitask model for anti-VEGF prognosis.
Key point: KongMing showed strong performance for predicting visual and anatomical outcomes after single injection, three loading doses, and 1-year 3+PRN treatment.
Implication: May influence prescriber choice and payer reviews pending full data.
Why it matters
- Cardiovascular AI remained a dense theme this week, spanning screening, monitoring, imaging guidance, and heart MRI interpretation [4][6][8][9][10].
- CNS and Alzheimer’s work leaned toward platform-building, with BBB delivery, AI-generated small molecules, speech models, and imaging-network deployment all advancing in parallel [2][3][7][25].
- Wearable-based evidence kept widening from engagement prediction and chronic-symptom monitoring to infectious-disease detection and consumer wellness signals [5][12][17][18].
- Several items showed AI moving beyond diagnosis into operational layers such as coding, medication workflows, diabetes pathways, and employer-sponsored care models [13][19][21].
- Governance is becoming more explicit, from “humble AI” design principles to professional calls to prohibit unsupervised AI therapy [20][23].
📚 Find your one-stop page for the full AI and Digital Health archive.
🎬 Watch on YouTube. Subscribe and never miss a video.
🚀 Accelerate your success. Contact us now
📂 Explore our case studies. See examples of our work.
💡 Read our insights. Learn from our latest reports and analysis
🎬 Watch on YouTube. Subscribe and never miss a video.
🧰 See our full range of services. Discover how we can help you.
FAQ
What did MOLEXA actually achieve in molecular imaging?
MOLEXA was reported to reconstruct likely pre-explosion molecular geometries from ion momentum data in Coulomb explosion imaging, and the team showed overlap with known structures for small molecules in experimental testing [1].
How material is the Insilico Medicine, Tenacia expansion?
The expansion adds a second CNS discovery program and carries potential deal value up to US$94.75 million, according to the announcement. It extends a March 2025 collaboration focused on BBB-penetrant small molecules [3].
What is the practical significance of Noah Labs Vox getting Breakthrough Device Designation?
BDD can accelerate interactions with FDA and support earlier commercial planning, but it is not a clearance or approval. Noah Labs also said an FDA trial is planned and EU MDR work is ongoing [4][10].
Did multimodal AI clearly beat text-only models in the dementia speech study?
Not in this report. The strongest results came from token-level fine-tuning of text-focused models, while multimodal performance was competitive but did not surpass the top text-only systems [25].
Why are Irish mental health groups pushing for an AI therapy ban?
The groups argue AI chatbots can mishandle crisis situations, reinforce delusions, and substitute simulated interaction for therapeutic human connection. Their position targets AI “therapy,” not human-delivered online therapy [20].
What stands out in the KongMing anti-VEGF study for AMD?
The model combined functional and anatomical prediction, including BCVA change, BCVA value, and post-treatment OCT image generation, across multiple treatment timepoints in a prospective multicentre setting [26].
Entities / Keywords
European XFEL, SLAC National Accelerator Laboratory, Lunai Bioworks, Clemann Group, Insilico Medicine, Tenacia Biotechnology, Noah Labs, Philips, Viz.ai, Alnylam Pharmaceuticals, Neurophet, ALZ-NET, Dong-A ST, Oura, eMed, OpenEvidence, OpenAI Foundation, MIT, Meta, University of Pennsylvania, Mayo Clinic, MOLEXA, Pharma.AI, Vox, EchoNavigator R5.0, DeviceGuide, Us2.ai, HiCardi, Doctor Eye CVD, CareSens Air, Symptom Radar, Healing Index, TRIBE v2, KongMing Model, blood-brain barrier (BBB), anti-VEGF, neovascular age-related macular degeneration (nAMD), cardiac amyloidosis, remote patient monitoring (RPM), continuous glucose monitoring (CGM), large language models (LLMs), multimodal AI, wearable biomarkers, Coulomb explosion imaging, COPD rehabilitation, dementia speech screening, ICD-10, CPT, E/M coding.
