AI in Medicine News — 2026-10-20 (Tue) Evening News

Overview

Drug development and trial operations saw the most activity: AstraZeneca and Oracle announced a global AI collaboration, and Insilico Medicine and Servier reached a first milestone (AstraZeneca, WSJ, Insilico Medicine). In regulation, Japan's PMDA shared principles for evaluating generative AI-enabled medical devices, the AHA made recommendations to the FDA, and the FDA cleared TumorSight Risk (RAPS, American Hospital Association, OncLive). In research, a face-photo model for predicting cancer survival drew wide coverage (The Guardian, Business Standard).

Diagnostics and imaging

- Imagene AI and Mindpeak are collaborating to bring AI-powered NSCLC biomarker profiling into digital pathology workflows. No performance figures are given in the headline (Detroit Free Press). - Aignostics launched HETA Clinical for cancer tissue analysis. The headline gives no accuracy data (Trend Hunter). - An AI system reads ultrasound images to track gout tophi automatically. The headline gives no performance level or evidence stage (Bioengineer.org). - MedTech Dive carries a piece on closing the integration gap in enterprise imaging, titled "AI that understands, reasons and acts: Closing the integration gap in enterprise imaging" (MedTech Dive). - MDLinx asks whether doctors worried about AI taking radiology jobs should look at their own, in a piece titled "Doctors are focused on AI taking radiology jobs—but should they be looking at their own?" (MDLinx).

Clinical decision support

- Stanford HAI published "Measuring What Medical AI Can Actually Do". The headline does not describe the measures (Stanford HAI). - A piece on AI sepsis alerts is titled "AI sepsis alerts face the test of action at the bedside". It does not give results (Innovation News Network). - Pharmacy Times asks "Can AI Help Predict Drug Shortages?" The headline does not give an answer (Pharmacy Times). - An AI tool is reported to predict cancer patients' survival chances from a single face photo. See section 9 for the study coverage (Business Standard). - TRT World frames "Algorithmic gaslighting" as a hidden harm of AI in healthcare. The headline gives no detail on cases or data (TRT World). 【Generative AI and medical documentation】 - Notable launched Flow Compass, which gives healthcare operators plain-language control of their AI workforce (HIT Consultant). - Rotherham Doncaster and South Humber deployed an AI scribe. The headline gives no outcome data (Digital Health). 【Hospital operations and care delivery】 - Axios describes a cost-benefit dilemma for medical AI ("Medical AI's cost-benefit dilemma"). The headline gives no figures (Axios). - DOTmed warns that AI-generated codes could cost a practice more than they earn (DOTmed). - KT is teaming up with Acryl on AI for the medical sector (Telecompaper). Maeil Business Newspaper reports the work covers standardizing medical data, developing AI services and joint on-site validation, with the AI transformation (AX) effort tailored to each hospital's data environment (Maeil Business Newspaper). - A headline about 6-foot AI robots in grocery aisles is not about healthcare (geekspin). 【Drug discovery and clinical trials】 - AstraZeneca and Oracle launched a global AI collaboration to advance clinical trial optimisation. The headline does not name a trial phase (AstraZeneca, WSJ, European Biotechnology Magazine). - Insilico Medicine and Servier reached the first milestone in their AI-driven R&D collaboration, which triggered an undisclosed milestone payment. The headline does not state the development stage (Insilico Medicine). - Eli Lilly signed a delivery-technology deal with Chinese AI biotech Metis Techbio. The headline does not state the development stage (MedWatch). 【Medical device approval and regulation】 - Guidance: Japan's PMDA shared principles for evaluating generative AI-enabled medical devices (RAPS). - Recommendations (not a rule): the American Hospital Association detailed its recommendations to the FDA on regulatory frameworks for GenAI-enabled medical devices (American Hospital Association). - Clearance: the FDA cleared TumorSight Risk, an MRI-derived AI recurrence tool for HR+, HER2– breast cancer (OncLive). - GeneOnline describes a "Billion-Dollar Battle to Control Healthcare AI" between Big Tech and Big Pharma (GeneOnline).

Safety, errors and bias

- The US barred Chinese AI tool DeepSeek from federal civilian agency networks (TRT World). - MyBroadband discusses how AI could transform healthcare and patient safety in South Africa (MyBroadband). - Deep Health AI India Limited (BSE: 539559) is pivoting from diamonds to facial health scans (Kalkine India). - Times Now reports that AI is creating never-before-seen viruses and asks whether they could save lives or trigger a health crisis. The headline gives no evidence stage (Times Now). 【Research findings (major journals)】 - An AI model predicts cancer patients' survival from a single selfie. Coverage describes it as a facial health recognition tool that helps physicians assess age and survival risk to guide personalized treatment, and as flagging patients at higher risk of early death. The headlines call it a study and suggest results; none gives accuracy figures (EurekAlert!, The Guardian, BigGo Finance, Asia Economy). - Doximity and OpenEvidence settled lawsuits over the development of a healthcare AI chatbot. This is a legal development, not a clinical result (HealthExec).

Developments in Japan

- Japan's government is preparing rules for AI adoption in fields including healthcare and construction, covering clarification of responsibility and relief for harm (Nikkei). - A research institution participating in Rapidus, together with Hokkaido University and others, is working on an advanced-semiconductor medical AI system (Jiji Medical, 47NEWS). - Japan Cyber Security began offering SHIELD SCANNER AI, an AI-based vulnerability assessment service. The headline does not say it is specific to healthcare (d Menu News).

Statements by key figures

No notable developments.