MedAI Digest

[MedAI Digest @MedAIDigest]

TheODen enabled federated histopathology training across three German hospitals without VPNs or inbound firewall ports, reaching Dice scores of 0.764 and 0.754.
https://t.co/SI3Ap3FwxG
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

NDSNet estimated Huntington’s disease clinical scores from raw speech, tracked longitudinal change, and reproduced MRI associations with neurodegeneration.
https://t.co/cxLFsEQ2sX
#MedAIDigest #LancetDigitHealth

[MedAI Digest @MedAIDigest]

In young children, lower baseline epitope-specific IgE predicted peanut immunotherapy outcomes, while treatment reduced IgE breadth and increased IgG4.
https://t.co/hrL36XNpLm
#MedAIDigest #JAllergyClinImmunol

[MedAI Digest @MedAIDigest]

Early CT-detected vascular remodeling predicted benefit from anti-angiogenic therapy in advanced lung cancer, with an interpretable model validated at a second center.
https://t.co/M67ZsmJC6K
#MedAIDigest #SciAdv

[MedAI Digest @MedAIDigest]

In a multicenter randomized trial, AI assistance raised specialists’ top-5 genotype-category accuracy for inherited retinal diseases from 67.3% to 88.5%.
https://t.co/8hrWnfiKlB
#MedAIDigest #NatMed

[MedAI Digest @MedAIDigest]

An open-weight LLM detected clinical high risk for psychosis with 93.4% sensitivity, but limited specificity and contextual errors argue strongly for specialist oversight.
https://t.co/PQRqUQsWGV
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

In Ottoman Greece, ethnicity strongly predicted vineyard cultivation in the 1460s, but within a century economic and environmental factors largely displaced it.
https://t.co/Lj4cOjuxyv
#MedAIDigest #NatCommun

[MedAI Digest @MedAIDigest]

Only 19.5% of second-level appeals for denied anticancer drugs succeeded, while clerical problems accounted for more than half of unfavorable decisions.
https://t.co/24ZBwH17jX
#MedAIDigest #JAMANetwOpen

[MedAI Digest @MedAIDigest]

Advanced prompting raised mean health-checkup classification accuracy from 0.69 to 0.95, although blood pressure and demographic differences remained challenging.
https://t.co/yKLwHc81IJ
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

CLEAR matched strong chest X-ray performance with concept-level explanations, enabling clinicians to identify and reduce confounding without retraining the image encoder.
https://t.co/fNx5gYfV3l
#MedAIDigest #NatBiomedEng

[MedAI Digest @MedAIDigest]

Across 286 mental health apps, US developers disclosed more clinical activity, but app-store ratings did not reflect evidence—and satisfaction priorities differed by country.
https://t.co/k2ZRtVpwMJ
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

Across four phases of reported global mpox, model-associated correlates shifted from behavioral and mobility factors toward structural vulnerabilities.
https://t.co/otxYo6XQUM
#MedAIDigest #BMCMed

[MedAI Digest @MedAIDigest]

In 577,742 screening mammograms, an AI workflow with targeted expert review raised generalists’ cancer detection by 32.7% to specialist-level performance.
https://t.co/medoDuL8lz
#MedAIDigest #Radiology

[MedAI Digest @MedAIDigest]

Passive home sensors found that dementia is linked to sleep patterns that diverge from typical ageing, supporting scalable risk screening rather than diagnosis.
https://t.co/nnrmgZsmFf
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

A three-tier, guideline-grounded AI system produced safer diabetic kidney disease recommendations while handling nearly 88% of cases locally.
https://t.co/7nwMiIVuCa
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

A CT-based multi-task model predicted EGFR mutation status across external validation, with performance improving when clinical and radiological features were added.
https://t.co/XovB8dupYs
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

LLM judges rated clinical AI more consistently than humans, but even the best jury matched Rwandan clinicians on only 5 of 11 criteria and missed demographic bias.
https://t.co/XKOvAeCpYo
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

MEDRF combined multimodal classification with evidence-grounded LLM reasoning, improving cognitive impairment diagnosis when clinical data were incomplete.
https://t.co/yKpy08lTkN
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

A contrastive AI model identified two reproducible brain-based depression subtypes that differed in symptoms, illness duration, and response to medication and rTMS.
https://t.co/FBy1QfPH7V
#MedAIDigest #NPJDigitMed

[MedAI Digest @MedAIDigest]

A domain-aware agent significantly improved the quality and safety of atrial fibrillation guidance from four leading language models.
https://t.co/ubhlGEUCYm
#MedAIDigest #NPJDigitMed