AI in Medicine - curated summaries making complex issues easy to understand

Mike Rawson

Details

AI in Medicine - Smart Summaries is a daily podcast for clinicians, health-system leaders and medtech, pharma and digital-health teams. Each episode unpacks one important public document on artificial intelligence in healthcare, such as FDA, EMA, MHRA, WHO or NHS guidance, OECD and World Economic Forum reports, or an open-access peer-reviewed review, in a clear 15 to 20 minute conversation. Topics include clinical AI and decision support, medical imaging and radiology AI, large language models (LLMs) and generative AI in clinical practice, AI-enabled medical devices and their regulation, the EU AI Act, drug discovery and clinical trials, digital health and wearables, health-system adoption, patient safety, bias and liability. The source of every episode is linked in the show notes, and full transcripts are at ai-in-medicine-podcast.vercel.app. The hosts' voices are AI-generated. For information only, not medical advice.

Recent Episodes

NOV 20, 2026
Who Gets Sued When an AI Scribe Hallucinates? The Hidden Liability of Automated Medical Notes
AI scribes are rapidly entering hospitals to save clinicians from documentation burnout, but who is legally responsible when the software hallucinates a diagnosis or drops a critical negation? This episode unpacks a PLOS Digital Health analysis of 74 official regulatory documents governing AI scribes. We explore the drafting fiction that shifts liability onto doctors, the biometric privacy risks of voice data, and why hospital leaders must rethink protected review time before deploying ambient tools. Key points Official guidance treats AI-generated notes as provisional drafts that require clinician authentication, a concept the study calls the drafting fiction. An analysis of 74 official documents reveals that while vendor contracts address broad product risks, the liability for individual note errors falls entirely on the verifying clinician. Medical device classification can shift after deployment depending on whether the AI performs simple transcription or interpretative summarisation. Privacy concerns extend far beyond secure storage, as spoken consultations contain biometric voiceprint data that can be virtually impossible to de-identify. Current guidelines fail to allocate protected time for clinicians to review AI notes, monitor for accent biases, or maintain manual documentation pathways for patients who refuse consent. Source: Regulating the drafting fiction: A qualitative content analysis of official guidance and regulator documents on AI scribe adoption and use in healthcare - PLOS Digital Health, 2026 This spot is available. Reach clinicians, health-system leaders and medtech and pharma teams following AI in medicine. Sponsor the show This episode is an AI-generated conversation summarising a public document; the hosts' voices are synthetic. It is for information only and is not medical advice. Always refer to the original source. Full transcript: https://ai-in-medicine-podcast.vercel.app/episodes/who-gets-sued-when-an-ai-scribe-hallucinates-the-hidden-liability-of-a-b2bc6f
19 MIN
NOV 20, 2026
Company Spotlight: Ever Fortune.AI ETT-XR-100 - AI Triage for Misplaced Endotracheal Tubes
This episode unpacks the FDA 510(k) clearance summary for Ever Fortune.AI's EFAI ETTXR system, an AI tool designed to flag malpositioned endotracheal tubes on adult chest X-rays. We break down the clinical testing data, the algorithm's specific measurement thresholds, and crucial limitations regarding esophageal intubation and double lumen tubes. Key points The EFAI ETTXR is a radiological triage software that analyzes adult chest X-rays to flag suspected vertically malpositioned endotracheal tubes in relation to the carina. In a retrospective, multisite clinical validation study of 940 studies, the device demonstrated a sensitivity of 0.890 and a specificity of 0.935. The software triggers a flag if the endotracheal tube tip is assessed as more than 7 cm above the carina, less than 3 cm above the carina, or below the carina. The device provides a passive notification via a text-based JSON file to prioritize worklists, but does not highlight or mark the original image. Crucial limitations include the inability to detect esophageal intubation, a lack of awareness of patient positioning, and the potential to trigger false alerts for properly positioned double lumen tubes. Source: FDA 510(k) summary K242821: EFAI Chestsuite XR Malpositioned ETT Assessment System (ETT-XR-100) (Ever Fortune.AI) - U.S. Food and Drug Administration, 2025 This episode is an AI-generated conversation summarising a public document; the hosts' voices are synthetic. It is for information only and is not medical advice. Always refer to the original source. Company: Ever Fortune.AI · Product: EFAI Chestsuite XR Malpositioned ETT Assessment System (ETT-XR-100) · FDA 510(k) K242821 Full transcript: https://ai-in-medicine-podcast.vercel.app/episodes/company-spotlight-ever-fortune-ai-ett-xr-100-ai-triage-for-misplaced-e-396e7d
13 MIN
NOV 19, 2026
Company Spotlight: Us2.ai Us2.ca - AI Detection for Cardiac Amyloidosis
We unpack the FDA 510(k) summary for Us2.ca, an AI software device by Us2.ai designed to detect cardiac amyloidosis from echocardiograms. We break down the clinical testing data, demographic performance differences, and key limitations hospital buyers need to know. Key points Us2.ca is cleared via the FDA 510(k) pathway as substantially equivalent to the predicate device, EchoGo Cardiac Amyloidosis 1.0. The software is indicated for adult patients with a specific left ventricular wall thickness of ≥ 12mm. In retrospective testing of 1,647 patients, the device produced a confident prediction for 1434 patients, yielding an 87.1% success rate. Performance on the confident predictions showed an overall sensitivity of 86.9% and specificity of 87.4%. The summary outlines strict limitations: it does not provide a diagnosis, does not highlight regions of interest on the image, and was only externally validated on GE and Philips ultrasound systems. Source: FDA 510(k) summary K250151: Us2.ca (Eko.Ai Pte) - U.S. Food and Drug Administration, 2025 This episode is an AI-generated conversation summarising a public document; the hosts' voices are synthetic. It is for information only and is not medical advice. Always refer to the original source. Company: Eko.Ai Pte · Product: Us2.ca · FDA 510(k) K250151 Full transcript: https://ai-in-medicine-podcast.vercel.app/episodes/company-spotlight-us2-ai-us2-ca-ai-detection-for-cardiac-amyloidosis-67a2d8
12 MIN
NOV 19, 2026
When Clinical AI Acts on Its Own: Building Defensible Agents in Healthcare
AI in healthcare is shifting from making predictions to taking autonomous actions. This document from IQVIA outlines how health systems and pharma companies must update their validation pipelines to ensure AI agents are defensible, auditable, and safe. Key points The software landscape has entered the agentic era, where AI no longer just drafts code or predicts outcomes, but actively takes actions like querying databases and updating records. Regulators have shifted their stance in just 24 months, moving from basic trustworthiness to requiring provable, continuous evidence of an AI system's safety and reliability. Risk tiering must be driven by the specific use case and downstream impact, with the autonomy of the agent acting as the primary amplifier of risk. Validating agents requires evaluating the entire dynamic workflow and how errors compound, not just testing an underlying model's accuracy on a static dataset. To govern autonomous systems at scale, organizations must use AI agents to validate other agents, while keeping humans accountable for the final judgment and acceptance criteria. Under Article 4 of the EU AI Act, AI literacy is no longer just an HR initiative, but a mandatory market-access requirement for everyone interacting with these systems. Source: Validating AI Agents and Software - IQVIA, 2026 This spot is available. Reach clinicians, health-system leaders and medtech and pharma teams following AI in medicine. Sponsor the show This episode is an AI-generated conversation summarising a public document; the hosts' voices are synthetic. It is for information only and is not medical advice. Always refer to the original source. Full transcript: https://ai-in-medicine-podcast.vercel.app/episodes/when-clinical-ai-acts-on-its-own-building-defensible-agents-in-healthc-d453a3
19 MIN
NOV 18, 2026
Company Spotlight: DigestAID Deep Capsule - Sifting Millions of Frames for GI Bleeds
DigestAID recently secured FDA 510(k) clearance for Deep Capsule, an AI tool designed to help gastroenterologists review capsule endoscopy footage for small bowel bleeding. Maya and Sam unpack the clearance summary to explore the algorithm's performance data, its substantial equivalence to a predicate device, and the real-world trade-offs between high sensitivity and low specificity when reviewing millions of video frames. Key points Deep Capsule is an AI software that acts as a computer-aided detection tool to help reviewers spot potential small bowel lesions in adults, outputting suspected frames into a thumbnail gallery. The FDA 510(k) clearance is based on substantial equivalence to the NaviCam ProScan, though Deep Capsule does not identify the anatomical location of the digestive tract like its predicate does. In a retrospective clinical validation study of 330 patients, AI-assisted reading achieved a diagnostic yield of 0.961, compared to 0.761 for standard-of-care reading. Patient-level sensitivity jumped from 0.762 with standard reading to 0.972 with AI assistance, but clinical specificity was notably low at 0.125 for the AI arm. The software was heavily tested on the PillCam SB3 system, which accounted for 307 of the 330 clinical validation exams, meaning compatibility should be checked before procurement. Source: FDA 510(k) summary K250655: Deep Capsule® (Deep Capsule US) (Digestaid - Artificial Intelligence Development) - U.S. Food and Drug Administration, 2026 This episode is an AI-generated conversation summarising a public document; the hosts' voices are synthetic. It is for information only and is not medical advice. Always refer to the original source. Company: Digestaid - Artificial Intelligence Development (digestaid.health) · Product: Deep Capsule® (Deep Capsule US) · FDA 510(k) K250655 Full transcript: https://ai-in-medicine-podcast.vercel.app/episodes/company-spotlight-digestaid-deep-capsule-sifting-millions-of-frames-fo-e5081f
13 MIN