Artificial Intelligence and Digital Health News – September 30th 2025

Timos Papagatsias

2 months ago

AI in Healthcare and Digital Health Update: AI Imaging, RCTs, Brain-Computer Interfaces, Digital Therapeutics, Biomarkers, Payer Adoption, Strategic Capital, and More

From payer momentum as Cigna backs AI imaging and large randomized RCTs like PRISM in mammography to Neuralink’s brain-computer interface for thought-to-text—this week’s digital health signals are powerful. We span clinical practice, payer adoption, neuroscience, and digital therapeutics, with early-stage biomarkers and strategic capital highlights across areas, underscoring clinical integration and frontier innovation. These are just a few of the updates—our full newsletter below explores the landscape in depth.

🏥 LLM-led visits at Akido Labs (ScopeAI) [1] [US • 22 Sep 2025]

https://www.technologyreview.com/2025/09/22/1123873/medical-diagnosis-llm/

Context: Clinics in SoCal use ScopeAI (fine-tuned Llama + Claude) to run history, differentials, plan; doctors approve later.
Key point: Akido claims 4–5× physician productivity; critics cite automation bias and unclear FDA/licensure fit (endpoint not specified).
Implication: Could streamline initiation and adherence via remote prescribing and logistics.

👶 NHS to trial AI at 20-week scans [2] [UK • 24 Sep 2025]

https://practicebusiness.co.uk/news-nhs-to-trial-ai-tool-for-detecting-congenital-conditions

Context: 12-month, phased pilot across 4 trusts (n>9,500) vs standard ultrasound; NIHR-funded; start winter 2025.
Key point: Study measures detection, workflow, cost-effectiveness; results expected early 2027.
Implication: May influence prescriber choice and payer reviews pending full data.

📱 Spren smartphone body composition validation [3] [28 Sep 2025]

https://www.trendhunter.com/amp/trends/ai-body-scan

Context: Pennington Biomedical compared app vs three DXA scanners.
Key point: Article reports ~2.6% error vs multiple DXAs (study specifics not fully detailed in source).
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.

🩺 AmplifyMD raises $20M for AI-driven virtual specialty care [4] [US • 26 Sep 2025]

https://hitconsultant.net/2025/09/26/amplifymd-secures-20m-to-scale-ai-driven-virtual-specialty-care/

Context: Series B led by Forerunner Ventures; integrated network across 15+ specialties; EHR-integrated platform.
Key point: Reported outcomes include 52% lower TeleStroke ops cost and 2× TeleHospitalist cross-cover consults per shift.
Implication: Signals pipeline investment and modality expansion.

❤️ Cigna to cover Cleerly AI plaque analysis [5] [US • 24 Sep 2025]

https://hitconsultant.net/2025/09/24/cigna-to-cover-cleerlys-ai-based-cardiovascular-imaging/

Context: Coverage for AI-QCT/AI-CPA starting Oct 1, 2025; joins UHC/EviCore; CPT I code planned Jan 2026 (per source).
Key point: Criteria include chest pain, eligible CCTA, intermediate risk/CAD-RADS 1–3, ACS ruled out.
Implication: Introduces competition that may affect pricing and formulary access.

🧠 AI chatbots for teen mental health in India [6] [India • 28 Sep 2025]

https://www.eurasiareview.com/28092025-ai-chatbots-for-teen-mental-health-augmenting-indias-counselling-services-analysis/

Context: ORF analysis on integrating chatbots with Tele-MANAS/e-Sanjeevani amid workforce shortages.
Key point: Recommends WHO-aligned safety testing, data-protection compliance, multilingual design via Bhashini.
Implication: May expand screening, initiation, and follow-up at scale.

😴 Dreem Health raises $29M for end-to-end sleep care [7] [26 Sep 2025]

https://www.fiercehealthcare.com/health-tech/digital-sleep-clinic-dreem-health-raises-29m-end-end-sleep-care

Context: Sunrise Group expands digital sleep clinic; FDA-cleared at-home diagnostic; AI plans for co-pilot, in-house build.
Key point: Aims to scale specialists and nationwide offerings; links poor sleep to comorbidities (metrics not specified).
Implication: Signals pipeline investment and modality expansion.

⚠️ Study: Gender-skewed chatbot outputs in diabetes eye disease [8] [29 Sep 2025]

https://www.devdiscourse.com/article/technology/3640406-chatbots-may-misguide-patients-with-gender-skewed-diabetes-responses

Context: Journal of Patient Experience study on LLM responses for diabetic retinopathy.
Key point: Readability above guidance; DeepSeek showed largest gender differences in urgency/referral language.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.

🧬 AI links aging biology to IPF pathways [9] [27 Sep 2025]

https://www.ajmc.com/view/ai-driven-research-toolkit-links-aging-biology-and-fibrotic-disease

Context: Aging study introduces proteomic aging clock + ipf-P3GPT transformer; UK Biobank data and COVID-19 testing.
Key point: Limited gene overlap suggests IPF dysregulation distinct from normal aging; validation limits noted.
Implication: May influence prescriber choice and payer reviews pending full data.

👶 AI flags severe pediatric asthma risk by age 3 (Mayo) [10] [US • 23 Sep 2025]

https://respiratory-therapy.com/disorders-diseases/chronic-pulmonary-disorders/asthma/ai-predicts-severe-asthma-risks-in-young-children/

Context: EHR ML/NLP on 22k children; applies PARC/API checklists.
Key point: Identifies high-risk subgroup with more infections and exacerbations; next-step external testing planned.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.

📈 Philips–Optum deal for ambulatory cardiac monitoring [11] [US • 25 Sep 2025]

https://www.massdevice.com/philips-strikes-digital-health-deal-with-major-u-s-health-insurer/

Context: Optum to offer MCOT & ePatch to 3.4M members across 22 states.
Key point: Aims to improve detection, referrals, adherence via connected monitoring; Epic integration cited.
Implication: May expand screening, initiation, and follow-up at scale.

💊 Cedars-Sinai HF-AI licensed to RhythmScience [12] [US • 24 Sep 2025]

https://fox4kc.com/business/press-releases/ein-presswire/851820008/rhythmscience-licenses-groundbreaking-ai-algorithm-from-cedars-sinai-to-optimize-heart-failure-treatment/

Context: HF-AI supports GDMT titration; TURING-HF study referenced.
Key point: Cardiologists agreed with AI recommendations ~91% (per source); reported doubling of med use, fewer hospitalizations.
Implication: Signals pipeline investment and modality expansion.

🗣️ A.Cure voice AI screens heart failure severity [13] [South Korea • 25 Sep 2025]

https://www.donga.com/en/article/all/20250925/5869543/1

Context: “Heart to Voice” analyzes vocal changes + 9 clinical indicators; clinical trials ongoing (n=124 stated).
Key point: Reports 90.4% current accuracy; targets MFDS approval and wellness integrations (details preliminary).
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.

🩺 UCLA-led PRISM RCT of AI in mammography [14] [US • 23 Sep 2025]

https://www.uclahealth.org/news/release/ucla-lead-16-million-national-study-artificial-intelligence

Context: $16M PCORI-funded, multi-state pragmatic RCT; Transpara via Aidoc aiOS; hundreds of thousands of mammograms.
Key point: Randomizes AI-assisted vs radiologist-only reads; evaluates detection, recalls, patient/clinician trust.
Implication: May influence prescriber choice and payer reviews pending full data.

🩸 Leukemia prediction model review [15] [27 Sep 2025]

https://bioengineer.org/evaluating-predictive-models-for-leukemia-types-review/

Context: Systematic review of ML prognostic models across leukemia types.
Key point: Notes validation gaps and population bias; calls for standardization and inclusive datasets.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.

❤️‍🩹 mHealth for cardio-oncology survivorship (systematic review) [16] [25 Sep 2025]

https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001027

Context: 14 studies; apps/wearables tracked PA, HR; EMR integration rare.
Key point: Underuse of HRV/sleep; calls for standardized metrics and workflow integration.
Implication: May expand screening, initiation, and follow-up at scale.

🧪 iCARE: individualized next-test recommendation framework [17] [25 Sep 2025]

https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001022

Context: Locally weighted models + SHAP to pick patient-specific added assessments.
Key point: Improved accuracy/AUC (6–12%) in select datasets; no benefit when features not discriminative.
Implication: Signals pipeline investment and modality expansion.

📝 Clinical AI scribes: accuracy vs. safety trade-offs [18] [29 Sep 2025]

https://bmjdigitalhealth.bmj.com/content/1/1/e000092

Context: 7 CAIS products, 8 standardized audio consults.
Key point: No error-free notes; omissions = 84% of errors; hallucinations less frequent but often severe.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.

🧑‍⚕️ SHAP vs. clinician-friendly explanations in CDSS [19] [26 Sep 2025]

https://www.nature.com/articles/s41746-025-01958-8

Context: Vignettes with surgeons/physicians evaluating transfusion CDSS.
Key point: Clinical narrative explanations increased acceptance/trust vs SHAP plots or results-only.
Implication: May influence prescriber choice and payer reviews pending full data.

🧾 TRIALSCOPE: simulate trials from EMRs [20] [22 Sep 2025]

https://ai.nejm.org/doi/full/10.1056/AIoa2400859

Context: LLM structuring + causal methods on >1M oncology patients.
Key point: Reproduced RCT-like lung cancer effects; simulated pancreatic cancer trials; ablations included.
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.

🧬 AI-designed phage genomes function in lab (preprint coverage) [21] [26 Sep 2025]

https://finance.yahoo.com/news/ai-designed-living-genomes-worked-180103219.html

Context: Report of 16 functional AI-generated phages out of 302; outcompeted ΦX174 in some tests (details summarized).
Key point: Highlights potential for phage therapy design and associated biosecurity concerns.
Implication: Signals pipeline investment and modality expansion.

🧠 Neuron–silicon biocomputer CL1 demos Pong [22] [Germany • 26 Sep 2025]

https://interestingengineering.com/science/lab-grown-brain-powered

Context: Cortical Labs’ CL1 uses ~800k human neurons on MEA; showcased at Fraunhofer IPA Summit.
Key point: Early platform for disease modeling/drug discovery; ~6-month neuron lifespan (per source).
Implication: Signals pipeline investment and modality expansion.

🧠 Neuralink US thought-to-text clinical trial timing [23] [US • 22 Sep 2025]

https://www.teslarati.com/neuralink-plans-launch-us-trial-translating-thoughts-into-text/

Context: FDA IDE-approved trial reported to start October; aims LLM communication “at speed of thought.”
Key point: Focus on severe speech impairment initially; expansion ambitions noted (timelines per source).
Implication: May expand screening, initiation, and follow-up at scale.

🧑‍⚕️ Policy watch: US federal AI posture & healthcare [24] [US • 25 Sep 2025]

https://ai.nejm.org/doi/full/10.1056/AIpc2500892

Context: Commentary on July 23, 2025 EOs/action plan affecting AI regulation/use.
Key point: Discourages regulation, flags DEI limits in “woke AI” EO; healthcare singled out as over-regulated.
Implication: Regulatory/generics: Introduces competition that may affect pricing and formulary access.

🧠 MS real-world insights from unstructured EHRs [25] [EU/US • 26 Sep 2025]

https://www.neurologylive.com/view/leveraging-ai-extract-real-world-insights-ms-rebekah-foster-john-foley

Context: ECTRIMS 2025: CHARM model abstracts MS notes/MRIs to derive EDSS, relapses, treatment patterns.
Key point: Estimated EDSS at 97.9% of visits; 79.5% had ≥1 relapse (database details in source).
Implication: Could inform practice and payer discussions; interpretation depends on study design and confounding control.

🦿 Cionic Neural Sleeve 2 for spasticity + gait [26] [US • 23 Sep 2025]

https://go.cionic.com/the-neural-sleeve-2-choreographing-muscles-to-work-together

Context: FDA-cleared “bionic clothing” adds MultiStim (activation + anti-spasm sensory stimulation).
Key point: Home study respondents reported ~30% average improvement across spasticity aspects (duration/controls not specified).
Implication: May influence prescriber choice and payer reviews pending full data.

💶 Sanofi Ventures adds $625M for biotech/digital health [27] [EU • 24 Sep 2025]

https://www.sanofi.com/en/media-room/press-releases/2025/2025-09-24-05-05-00-3155175

Context: AUM now >$1.4B; focus on immunology, rare, neuro, vaccines, AI/digital health.
Key point: Evergreen CVC cites recent exits; aims to support early-stage innovation.
Implication: Signals pipeline investment and modality expansion.

Why it matters

Clinical reliability is under active test: Large, pragmatic RCTs (e.g., PRISM [14]) and health-system deployment studies (e.g., causal ML at Kaiser [20]) will shape payer and regulator confidence.
Coverage shifts accelerate adoption: Cigna’s policy for Cleerly’s AI plaque analysis [5] and Optum’s Philips deal [11] expand access pathways.
Equity & safety need guardrails: Evidence of readability and gender bias in chatbots [8,25] and scribe omissions [18] underscores the need for oversight.
Novel modalities are diversifying: Voice biomarkers [13], smartphone body scans [3], biocomputers [22], and AI-designed genomes [21] widen the innovation frontier.
Capital is flowing to platforms: Funding/news from AmplifyMD [4], Dreem [7], Sanofi Ventures [27] signal continued appetite for AI-enabled care models.

📢 Stay Ahead in Artificial Intelligence and Digital Health!

✅ Make sure to like, share, and subscribe to our YouTube channel to catch all future highlights on AI and digital health.

✅ Contact LucidQuest at info@lqventures.com for strategic guidance on leveraging AI and digital health innovations across healthcare.

FAQ

What exactly is ScopeAI doing at Akido’s clinics? [1]

ScopeAI runs structured interviews and drafts differentials and plans from visit audio; a physician reviews/approves asynchronously. Experts warn of automation bias; regulatory status relies on human-in-the-loop.

When will the NHS fetal anomaly AI results be known? [2]

The 12-month trial (phased from winter 2025) targets early 2027 for outcomes on accuracy, workflow, and cost-effectiveness.

What conditions qualify for Cigna’s coverage of Cleerly’s AI plaque analysis? [5]

Patients with chest pain eligible for CCTA, no known CAD, intermediate risk/CAD-RADS 1–3, and negative/inconclusive ACS evaluation starting 01 Oct 2025.

Are clinical AI scribes ready for unreviewed use? [18]

No. In testing, none were error-free; omissions were common and can be clinically significant. Clinician review remains essential.

How is PRISM different from prior mammography AI studies? [14]

It’s a large randomized US trial comparing AI-assisted vs radiologist-only reads, also measuring patient/clinician perceptions—not just AUC.

Does the gender-bias chatbot study recommend a specific model? [8]

It assesses multiple LLMs and notes differences; emphasizes physician oversight and readability limits rather than endorsing a single tool.

Entities / Keywords

Akido Labs / ScopeAI; NHS fetal anomaly AI; Spren body scan; AmplifyMD; Cleerly LABS AI-QCT/AI-CPA; Cigna; Dreem Health / Sunrise Group; DeepSeek, Gemini, ChatGPT, Claude; Idiopathic Pulmonary Fibrosis; Mayo Clinic pediatric asthma AI; Philips MCOT/ePatch; RhythmScience HF-AI / Cedars-Sinai; A.Cure Heart to Voice; PRISM / Transpara / Aidoc; Leukemia ML models; mHealth cardio-oncology; iCARE individualized assessments; Clinical AI scribes (CAIS); SHAP explanations; TRIALSCOPE (NEJM AI); AI-designed phages; Cortical Labs CL1; Neuralink BCI; US AI policy (NEJM AI commentary); CHARM MS EHR abstraction; Cionic Neural Sleeve 2; Sanofi Ventures.

References