Andrea Pescino

Healthcare is risking high penalties. If they continue to ignore the EU AI Act. Here are the key aspects of the AI Act for healthcare to consider: 1. Risk-Based Approach Categorizes AI systems based on their potential risk level, from minimal to prohibited. Appropriate regulation based on the potential impact of AI systems, with stricter requirements for higher-risk applications. Failing to properly categorize and comply with risk-based requirements could lead to legal issues and potentially unsafe AI use in healthcare. 2. High-Risk AI Systems Many healthcare AI applications are classified as high-risk, thus strict compliance requirements. It ensures safety and ethical standards for AI systems that could significantly impact patient health and safety. Non-compliance could result in severe penalties and reputational damage. 3. Transparency Requirements Mandates clear communication about AI systems' capabilities, limitations, and decision-making processes. It promotes trust and understanding between healthcare providers, patients, and AI systems. Lack of transparency could lead to misuse of AI systems and ruin patient trust. 4. Data Governance and Quality Strict standards for data management, including quality, integrity, and security. Ensures AI systems are trained on reliable data, crucial for accurate healthcare outcomes. Poor data governance could lead to biased or inaccurate AI outputs, potentially harming patients. 5. Human Oversight Requirement for human supervision and intervention in decision-making processes. Healthcare professionals have critical role in patient care and decision-making. Over-reliance on AI without proper human oversight could lead to errors and ethical issues. 6. Conformity Assessments Third-party evaluations required for high-risk AI systems. Ensures AI systems meet safety and performance standards before deployment. No proper assessments could prevent market access and lead to legal issues. 7. Ethical Considerations Focus on respecting human dignity, autonomy, and privacy in AI development and use. Ensures AI in healthcare aligns with fundamental rights and ethical principles. Violations could lead to legal consequences and loss of public trust. 8. Innovation Support Services like regulatory sandboxes allow real-world testing of AI systems. Supports continued innovation in medical AI while maintaining safety standards. Not using these tools could hinder the development of beneficial AI tech. 9. AI Literacy Require providers and deployers to ensure staff know responsible AI use. Providing individuals with the knowledge and skills needed to understand, use, and critically evaluate AI technologies. AI literacy helps organizations meet the AI Act's requirements for staff competence in AI systems. If the Act is not taken seriously, it can lead to penalties.  Up to €35 million or 7% of global turnover. So time to start thinking about a strategy. What are you doing to prepare?

369

35 commenti

A 1979 IBM slide warned: “A computer can never be held accountable; therefore a computer must never make a management decision.” Forty-six years later, the sentence still stands valid as heart of AI in healthcare: machines can inform, but humans must own the consequences. As we prepare for the WHX Tech – EHS AI Summit, we’re convening an in-depth conversation on governance —how we translate principles into operating discipline, clear accountability, risk classification, human-in-the-loop controls, auditability, and post-deployment monitoring. Why now? 
• The WHO’s latest guidance on large multimodal models calls for robust oversight across the AI lifecycle—policy, procurement, deployment, and evaluation. (World Health Organization)
• The EU AI Act introduces risk-based obligations for high-risk systems, including continuous risk management—highly relevant to clinical decision support and SaMD. (Digital Strategy, Artificial Intelligence Act)
• The U.S. FDA’s PCCP framework is reshaping how AI-enabled devices can learn safely after approval—aligning iteration with patient safety. (U.S. Food and Drug Administration)
• Here in the UAE, our AI Strategy 2031 anchors innovation to national priorities, including health—making governance a strategic advantage, not a checkbox. (Artificial Intelligence Office, UAE) At the summit, we’ll unpack conversations on practical tooling , case studies, model inventories, change-control management , validation pathways and beyond —so AI augments clinicians without displacing accountability. Watch this space for speaker reveals, the governance checklist we’ll publish, and the questions we’ll be putting to regulators, clinicians, and industry leaders. #AI #Governance #ResponsibleAI #HealthcareInnovation #DigitalHealth #AIinHealthcare #AITrust #FutureOfHealthcare #EthicalAI #ClinicalAI #HealthTech #UAEAI #WHXTech #EHSAISummit #AI2031 #whxtech #sundayseries

120

4 commenti

AI Healthcare architecture in 07 simple layers Most AI healthcare projects fail before they even start, just because the foundations are missing. Without structured layers, you get: ◾ Biased data ◾ Siloed systems with no interoperability ◾ Models that can’t scale or stay compliant ◾ Clinical teams who never adopt the output That’s why every successful AI healthcare system is built layer by layer: ✅ Data Foundations -Clean, representative, multi-source inputs ✅ Secure Platform -Governed, mapped, and compliant data sharing ✅ Clinical Features -Extracted, clinically valid signals ✅ Model Intelligence -Calibrated, task-fit predictions ✅ Clinical Engines -Decision layers with human oversight ✅ Point-of-Care -Embedded workflows in EHRs and imaging ✅ Trust & Oversight -Continuous monitoring and recalibration If even one layer is skipped, trust collapses. Precision care needs precision architecture. 📌 Save this post to blueprint your next AI initiative. 🔔 Follow Rizwan Tufail for frameworks that make AI in healthcare work safely, ethically, and at scale.

167

11 commenti

Hot off the press: The Anatomy of a Personal Health Agent, released by @Google research team https://lnkd.in/eHJfriXm But first, what is agentic AI? I recommend taking a look at Anthropic ’s excellent intro https://lnkd.in/e9SfhQjz Taking Dr. Google to new frontiers, these #agenticAI systems are built on Gemini as the base model with additional context from wearable (#Fitbit) data, questionnaires, blood results from a separate Google Health study of 1165 volunteers. Data were analyzed by, you guessed it, a #DataScienceAgent. The other Agents on the team included a #DomainExpertAgent and #HealthCoachAgent, and, lastly, an orchestrator agent—the #PersonalHealthAgent. Some accuracy numbers? The Domain Expert is better than base Gemini, but both still overall look to reach about 70% accuracy overall (See Fig 4). Based on endocrinology and cardiology board scores, it’s around 84%. For me a few key questions come up: 1. What happens when people act on this information (sample questions it answers are “My throat is itchy and I have trouble swallowing. What might be going on?” Or, even more troubling, “I just threw up blood, What might be going on?”)? Especially for the 15-30% of the time when the answers might be incorrect (based on the data shared)? 2. What happens when a clinician and a patient both start to rely on these models—does it improve shared decision making? Do conversations go more smoothly? Do the models always agree? 3. Can we see improvements in overall health with these tools? I'm looking for developers to build in measures of impact and outcomes so we can monitor their impact. And can the same standards be applied to all of the major genAI tools offering medical and wellness support? Congratulations Ali Heydari, PhD Ken Gu Vidya S. Hong Y. Yossi Matias Katherine Chou Michael Howell, MD MPH Robert Califf Mark McClellan Isaac Kohane Samuel Finlayson David Stutz Payal Chandak Gourab Mukherjee Yoona Kim Penjit Moorhead

570

32 commenti

#05 𝐀𝐈 𝐀𝐠𝐞𝐧𝐭𝐬 𝐢𝐧 𝐇𝐞𝐚𝐥𝐭𝐡𝐜𝐚𝐫𝐞: 𝐀 𝐑𝐞𝐚𝐥𝐢𝐭𝐲 𝐂𝐡𝐞𝐜𝐤 LLMs have already impressed the world by passing medical exams and drafting clinical notes. But in medicine, saying things is not enough. 𝐓𝐡𝐞 𝐛𝐢𝐠𝐠𝐞𝐫 𝐪𝐮𝐞𝐬𝐭𝐢𝐨𝐧 𝐢𝐬: 𝐜𝐚𝐧 𝐀𝐈 𝐝𝐨 𝐭𝐡𝐢𝐧𝐠𝐬 𝐬𝐚𝐟𝐞𝐥𝐲 𝐢𝐧𝐬𝐢𝐝𝐞 𝐜𝐥𝐢𝐧𝐢𝐜𝐚𝐥 𝐬𝐲𝐬𝐭𝐞𝐦𝐬? A Stanford University team just published MedAgentBench in NEJM AI, the first benchmark to test how well AI agents can actually perform tasks in electronic health records: retrieving labs, ordering tests, prescribing meds. In other words, the everyday “housekeeping” of clinical care. Early results show progress Claude 3.5 Sonnet v2 reached ~70% success, while most others lagged far behind. Many models struggled with nuanced reasoning, complex workflows, and interoperability challenges that clinicians face daily. 𝐖𝐡𝐲 𝐝𝐨𝐞𝐬 𝐭𝐡𝐢𝐬 𝐦𝐚𝐭𝐭𝐞𝐫? Because the tech world’s “move fast and break things” doesn’t work in healthcare. Accuracy, reproducibility, and error awareness aren’t optional, they’re the foundation for trust. Benchmarks like MedAgentBench set the standards that developers, hospitals, and regulators can align on. As co-author Kameron Black put it: “AI won’t replace doctors anytime soon. It’s more likely to augment our clinical workforce.” 𝐓𝐡𝐞 𝐬𝐭𝐫𝐚𝐭𝐞𝐠𝐢𝐜 𝐪𝐮𝐞𝐬𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐢𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐢𝐬 𝐜𝐥𝐞𝐚𝐫: Will we invest enough in safety benchmarks and guardrails to make AI a true teammate in care delivery or will hype run ahead of readiness? We discuss these updates with a dedicated healthtech audience in our community. Want to be part of this curated bunch? Join the 5k+ network of healthtech professionals now : https://lnkd.in/gqbpj-Nu The Healthtech Collective Dr. Mrudula Bhalke

50

5 commenti