AI in Medicine News — 2026-10-09 (Fri) Evening News

Overview

- The areas that are seeing notable movement are clinical decision support/medical records (Suki, Abridge/VA) and drug discovery financing (Iambic IPO, Rivercell, Phyraxis). - On the research front, AMIE-related research published in Lancet and real-world evaluation of AI for patients (BIDMC) were published, while there was also research (IIIT-H) warning against overconfidence in AI explanations (heat maps). - On the regulatory side, we can see the FDA's review of evaluation methods for AI-generating devices, the signing of three AI-related bills in California and the veto of two, and AHRQ's AI measures.

Diagnosis/Image

- Parasite diagnosis: Research on reading microscopic images using deep learning. The headline reads, ``Verification in a real environment is delayed.'' The numbers are unknown. - Neonatal pneumothorax: Discussion of the need for AI to capture clinically important pneumothorax. There are no performance values in the headline. - Research showing that AI chatbots can safely assist in diagnosis before visiting a GP (Rayo). The level of evidence and scale are unclear from the headline. - Public dataset for CT: The purpose is to let AI show the basis for its decisions (“show its work”) (Medical Xpress). - UK NHS: Deploying AI-equipped spirometry to address undiagnosed COPD (also from an operational perspective). - The market report (AI-enabled digital stethoscope, 2036 outlook) is market research and not clinical evidence.

Clinical decision support

- Suki announces "Ask". Wolters Kluwer UpToDate's expert AI-based complex clinical decision support. There is no performance evaluation in the headline. - South Korea: Researchers propose "1.5-tier" care for overburdened hospitals (Korea Biomedical Review). - Forbes outlines the use of AI for clinicians and oncology care providers (explanatory article). - Chatbot research (NYT: Clinical trial showing potential for emergency department support) in the Diagnosis/Image section is also related.

Generation AI and medical documents/medical records

- Abridge: Selected by VA corporate contract to provide ambient AI across Veterans Care nationwide. - Healthleap: $38M raised (Sequoia, First Round, Hummingbird). Expanding AI platform to detect undiagnosed conditions during hospitalization. There are no numerical values for detection performance in the headline. - Emergence: Announces acquisition of Clinect Healthcare. Patient experience/outcome intelligence domain.

Hospital management/medical provision system

- China: Medical AI penetrates hospitals (China Daily). - NHS: Deployment of AI spirometry (above). - AHRQ: Announces AI measures related to medical evidence reviews and data. - Mount Sinai: Wearable AI predicts prolonged sitting in women with chronic pelvic pain. It appears to be in the research stage, but details are unknown. - Q&A discussing whether the patient financial journey is the next area for AI.

Drug discovery/clinical trials

The development stage is not in the headline. Developed an AI virtual cell model for drug discovery (the stage is unknown as it is not specified as preclinical). - Phyraxis AI: Obtain initial funding. - Article about clinical data department shifting focus to AI validation and human oversight (Clinical Trial Vanguard).

Medical device approval/regulation

- FDA: Currently considering introducing evaluation methods that imitate doctors for AI-generated devices (not at the stage of approval or designation, but at the stage of consideration). - Explanation (Clinical Leader) that medical LLM selection requires a pre-introduction readiness assessment. - Approval/designation of individual items does not fall under the heading. - State law: California governor signs three healthcare AI-related bills and vetoes two (laws, not device regulations).

Safety/Error/Bias

- IIIT-Hyderabad: Warning against doctors relying entirely on AI for diagnosis. Findings that AI heatmaps can mislead radiologists (multiple media). - Experts say AI will help reduce medical errors (Pakistan). - Experts call for regulations to protect patients in Nigeria. It has been pointed out that there remain barriers to implementation in Canada.

Research results (major journals)

- Lancet: Suggests AI could improve patient-doctor relationships (Google). - Google AMIE: Lancet study shows promise (The Tech Buzz). Both studies appear to be the same study, but it is difficult to confirm from the headline. Design and result values are unknown. - BIDMC: The first study to examine the safety and quality of AI for patients in primary care in a real-world environment (Newswise). The results are not in the headline. - Public dataset for CT (above).

Japanese movements

- LLM developed by Sakana AI was adopted in "Evidence Finder" for doctors. - Nikkei Business: The tone that advanced AI will endanger critical infrastructure, including medical institutions. - The Phage (preventive medical AI) receives additional investment in Series A. - CoWorker: Provides free AI security vulnerability diagnosis (it is unclear whether it is specialized for medical care). - Vietnam.vn article discusses solving human and data bottlenecks in AI and robot applications in the medical field. This is not a Japanese move.

Remarks from VIPs

No relevant information (list of important people being tracked is empty).