Healthcare AI Daily Briefing

Diagnostics & Pathology

In histopathology, LLM2 showed lower misinterpretation and LLM3 higher hallucination rates than peers.

Preprint

Four general-purpose LLMs were evaluated on 153 multi-organ pathology cases without training or fine-tuning; LLM2 had significantly lower misinterpretation rates than the other three LLMs, while LLM3 had significantly higher fabricated-feature hallucination rates than LLM1 and LLM2. [1]

Medical Imaging

AI-ECG detects HCM phenotype and predicts HCM development in sarcomere variant carriers.

Preprint

Among genotype-positive, phenotype-negative individuals, higher AI-ECG scores predicted HCM development during follow-up. In 57,007 UK Biobank participants, high AI-ECG plus high polygenic risk yielded 60-fold higher HCM odds versus either alone. [2]

Policy & Governance

CMS Administrator Oz warns AI will initially inflate healthcare costs.

At Oracle's health summit, CMS Administrator Dr. Mehmet Oz said AI will 'turbocharge' medical billing and drive up costs short term because it enables existing billing systems to work more effectively, but he said the pain is worth it for long-term savings. [4]

Policy & Governance

Ohio health department’s AI policy focuses on accuracy and public trust.

The Ashland County, Ohio public health department’s newly approved AI governance policy explicitly addresses accuracy and public trust as core considerations for responsible AI use in health care. It requires staff to verify AI-generated claims, citations, and statistics before use and mandates human review for most non‑minor AI applications to ensure accuracy. [3]