J. David Giese

Psssst healthcare folks (and job seekers on the prowl)! AI in healthcare is a hot topic these days—so here’s a quick rundown of the major trends (credit to HLTH USA). ➡️ Imaging & Diagnostics: AI in medical imaging boosts diagnostic speed, accuracy, and efficiency, aiding clinicians in early disease detection and secure workflows. ➡️ Virtual Scribes & Assistance: AI-powered virtual scribes reduce admin tasks, letting providers focus on patients while improving workflow efficiency and documentation. ➡️ Patient Engagement & Care Management: AI enhances patient engagement through personalized care, streamlined communication, and coordinated care management. ➡️ Clinical Decision Support: AI supports data-driven decisions, improving patient care with real-time insights and evidence-based guidance. ➡️ Personalized Care & Workflow Automation: AI personalizes treatment with clinical/genetic insights, optimizing care for complex diseases and improving outcomes. ➡️ Predictive Analytics & RCM: AI-driven analytics enhance healthcare by predicting trends, improving outcomes, and streamlining billing for scalable, value-based care. ➡️ Healthcare Consulting & Administration: AI-powered consulting drives efficiency, innovation, and superior outcomes by transforming healthcare operations.

21

For those following AI governance, the FDA's guidance on pre-determined change control plans for AI in medical devices is a good example of a regulator with deep sector knowledge taking a scalpel approach to addressing unique challenges posed by AI, rather than trying to write an entire new rulebook.

12

This is a fascinating article by the talented Marty Stempniak of Radiology Business and goes to a debate about the purposes and value of AI today in radiology. On the one hand, the study authored by Cody Savage of University of Alabama at Birmingham argues that an AI application effectively didn't improve radiologist performance or turnaround time. Fascinating data! On the other, Jerome Avondo of Aidoc argued that the study missed the point of the AI application in practice, namely to be a safety net for patient safety rather than solely a productivity tool. This is a great debate and one that Nina Kottler, MD, MS, FSIIM and Woojin Kim have both eloquently weighed in on. Here is my take (admittedly very much a work in progress, so please weigh in!): Radiologists: Diagnostic AI has a funny impact on radiologists today. It is a bit intimidating. What happens when you disagree with the AI? It is a bit like a super version of QC that never stops running. As a result, we have seen that sometimes it creates an odd false positive effect in which radiologists worry about being overruled by AI and "missing" something. Certainly not all the time but it happens. Population Health: Diagnostic AI can run large scale screening and even catch edge cases (as Dr. Avondo points out) in a way that no human can. Jason Wiesner, MD, MBA at Sutter Health has been an early advocate for this. Still, even here, there are still challenges. As a general rule, we don't quite know how to operationalize and track the follow up for incidental findings at scale. Findings are often gray and the workup may not be trivial. Workflow: AI still shines in workflow. It is funny that the most mundane tasks can be the ones that bring a system to its knees. Hence the ability to streamline such tasks - automating them via AI - seems to produce dramatic benefits. As groups as diverse as Radiology Partners and Raleigh Radiology (full disclosure - both clients of Intelerad), implement efficiency solutions, we are seeing fascinating benefits. Satish Mathan Mustafa Khan Feedback and corrections most welcome! https://lnkd.in/en2tcEtb

44

19 Comments

The tool discussed in this article is a perfect example of the type of AI system that the #AIAct is needed for. Corti is a tool for AI-powered patient triaging and coding, determining the appropriate diagnosis and procedure codes, intended to be used in public health services, ambulatory clinics, etc. Safety is their business. Still, too often, builders of such AI companies don't plan to invest in safety and compliance. If they do, their incremental compliance costs on the AI Act should not reach six figures, as mentioned here. There's much to say about the AI Act, though. I am very aware that the compliance costs will be higher than the Commission expected or communicated in their cost analysis. One of the main causes, I think, is that everyone in AI in the EU must conduct an analysis of to what extent the AI Act applies to them and be able to maintain that confidence of the applicability across their AI inventory over time. While EC's estimate of 5-15% of AI systems falling into the high-risk category and the most strict requirements might hold, all of us are impacted. It's time to make smart choices on how to address it. #AIAct #EU #impactanalysis https://lnkd.in/gRf74Ksz

21

7 Comments

Is AI hype or reality in nuclear medicine? That question was addressed in point-counterpoint articles in AJR that highlight the progress -- and challenges -- facing AI adoption in the discipline. Read more at https://buff.ly/4aO7VIU #radiology #medicalimaging #nuclearmedicine #AI #ImagingAI

4

https://lnkd.in/gXxBxq5h Very insightful info on ML/AI medical technology by FDA experts Troy Tazbaz and John Nicol. ML/AI LCM may appear overwhelming, but by applying the most fundamental principles of safety engineering, one can connect the dots for focus and comprehensiveness. Here is the safety case structure for AI, with LCM being part of it. These principles align with AAMI TIR38, an extremely valuable yet underutilized standard. The safety of ML/AI technology is supported by: 1. There is reasonable assurance that ML/AI technology does what it is supposed to do: • Intended use, intended benefits, context of use, and system boundaries must be adequately documented. • Specifications (e.g., data, model) for the intended use must be defined, adequately verified, validated, and continuously monitored, controlled, and maintained. 2. There is reasonable assurance that ML/AI technology does no harm: • The risk management process (including cybersecurity) must ensure that potential risks, including risks of misuse, are identified. • Risk control measures (including security control measures) must be adequately established (identified, implemented, and verified). 3. There is reasonable assurance that ML/AI technology reliably does what it is supposed to do and does no harm: • ML/AI model reliability specifications (e.g., False Negative Rate, False Positive Rate) must be defined and adequately justified for the risks (and benefits) associated with the technology. • The reliability specifications must be adequately verified, monitored, and maintained. • There is reasonable assurance that an adequate life cycle process has been established to ensure the quality and reliability of the technology. Try to apply the same principles to other emerging technologies, and let me know if it makes sense.

7

1 Comment

#openemr along with right professional vendor is required combination for the effective usage of #openemr , especially on windows platforms

3

We've published our latest Yoti facial age estimation (FAE) white paper. Our Sep 24 model performance highlights are super good. * 99.3% of 13-17 year olds are correctly estimated as under the age of 21. * Over 95% of adults can use Yoti FAE to prove they're an adult without identifying themselves. * 99.5% of 6-12 year olds are correctly estimated as under the age of 13 * Over 98% of all teenagers and adults can use Yoti FAE to prove they're at least 13 without identifying themselves * The technology is being used by many companies globally to create age-appropriate experiences, across social media, gaming, vaping, ecommerce and online porn with over 650 million age checks completed worldwide Our Sep 24 FAE model will undergo independent testing by Age Check Certification Scheme from late September using a diverse sample of 18 year olds. Our current Mean Absolute Error (MAE) for 18 year olds is 1.2 and just 0.5% are estimated to be over 25 which are the two key performance measures ACCS test for (as they do not yet have test data for under 18s). Ofcom's has not published any credible evidence supporting their view that FAE 'does not perform very well for children'. Yoti's performance shows the opposite. Our MAEs for 15, 16 & 17 year olds are 1.0, 0.9 & 1.0. Our MAEs for 10 & 11 year olds are 1.0 & 1.1. Yoti FAE can be an effective age check to allow almost all teenagers and adults to prove they're at least 13 and to prevent the vast majority of under 13s from accessing 13+ content. The full white paper is available here: https://lnkd.in/emN4GCYm

26

1 Comment

Understanding the FHIR Patient Resource: A Review of Key Components and US Core Specifications for the Novice The FHIR (Fast Healthcare Interoperability Resources) designed to capture essential patient demographics and administrative information, it ensures interoperability across diverse healthcare systems. This resource conforms to the US Core Implementation Guide, which provides additional constraints to ensure compliance with US regulations. Profile Overview  The base Patient resource contains core fields such as name, gender, birth date, address, and contact details. In the US, implementations typically follow the US Core Patient Profile, which tightens constraints to ensure compliance with ONC’s 21st Century Cures Act. This profile requires specific fields (e.g., race, ethnicity, and birth sex) to meet US reporting standards. Extensions  FHIR’s flexibility allows for extensions that accommodate data not part of the core resource. Common US Core extensions in the Patient resource include:   • US Core Race – Captures race according to US standards (based on CDC race and ethnicity codes).   • US Core Ethnicity – Captures patient ethnicity using OMB standards.   • US Core Birth Sex – Required to capture the patient’s birth sex (male, female, unknown).  These extensions support interoperability while meeting US-specific healthcare regulatory requirements. Bindings The US Core version of the Patient resource ensures that key fields are bound to standard US-specific terminologies for consistency:   • Gender – Bound to the `AdministrativeGender` value set (male, female, other, unknown), in line with global FHIR standards.   • Race and Ethnicity – Bound to the US Core Race and Ethnicity value sets, which adhere to CDC and OMB standards.   • Language – US Core specifies binding to the `Language` value set, covering common spoken and written languages used in the US healthcare system.  By adhering to these bindings, data remains consistent and easily sharable across systems. Value Sets   The US Core implementation also uses specific value sets to standardize data fields for national interoperability. These include:   • Contact Relationship – A value set used for defining relationships such as parent, spouse, guardian, etc., constrained for US usage.   • Address Use – Standardized value set that defines whether an address is for home, work, or temporary use in a way that meets US requirements.   • US Core Race/Ethnicity – Value sets for capturing detailed race and ethnicity information (OMB’s five race categories and specific CDC ethnicity codes). The FHIR Patient resource, particularly in the US Core implementation, strikes the perfect balance between standardization and adaptability. By leveraging profiles, extensions, bindings, and value sets, healthcare organizations can ensure patient data is not only comprehensive but also compliant with US standards, enabling seamless and compliant health information exchange.

14

1 Comment

Here’s your recap of last week’s health IT news 🗞️ 👇 🥼AI Clinician Assistant • Suki : The AI scribe announced an SDK and APIs enabling developers to integrate ambient documentation into their applications. • Apricot: The gen AI pre-seed startup focused on reducing documentation overhead for home health nurses launched. • Tampa General Hospital + Palantir Technologies: The health system and the software platform are partnering to develop an AI care coordination and decision support platform. ☎️ AI Contact Center • Mediktor + Sensely: The AI symptom triage solution acquired the conversational member engagement platform. 📋  Patient Administration • Oracle (Cerner) + Loyal: The EHR now offers integration with the patient engagement solution through its Healthcare Marketplace. • HubSpot: The CRM and marketing platform announced it will now offer HIPAA support and sensitive data tools. 🏥 Provider and Practice Administration • LeanTaaS: The AI-enabled facility management solution released updates including: mobile-first live view of OR schedule, predictive staffing automation, and a gen AI solution allowing decision-makers to glean staffing and scheduling insights. ⚕️ Clinical Operations • Keragon: The HIPAA-compliant automation platform launched out of stealth with a $3M funding round. 🧑‍💻 Virtual and At-Home Care • AvaSure + CLEW: The virtual care platform is partnering with the AI-powered remote patient monitoring and analytics company to improve patient outcomes. 🍎 VBC • Evolent + Machinify, Inc.: The VBC org acquired AI Authorization assets from the AI operating system platform.    • ProgenyHealth, LLC + Wildflower Health: The maternity and NICU case management solution is partnering with the women’s health VBC enabler to “help health plans fill gaps, collapse silos and integrate into existing clinical and operational workflows.”    • Health Catalyst + Carevīve: The population health data and analytics company has completed acquisition of the cancer care management solution. 🧠 Food for Thought: • This article provided an interesting look at Nebraska Medicine’s efforts to design a next-generation hospital room, including the four technologies they’ve selected to anchor their work. https://lnkd.in/em-u-a5v    • McKinsey released a survey of 200 global health system executives on their investment priorities. Among the notable findings, 88% listed AI as an area of high potential impact, yet only 40% have implemented any solution thus far and over 20% don’t have plans to invest in the next two years. https://lnkd.in/eK3CeqQ2

38

4 Comments

Exciting news from HLTH USA – I’m thrilled to share that Aidoc, in collaboration with NVIDIA, is developing the BRIDGE Guideline, a new framework that I believe will help shape the future of AI in healthcare. BRIDGE is not just another guideline; it’s a strategic approach to overcoming the most critical barriers, scalability and integration, that have slowed AI adoption across our industry. At its core, BRIDGE is designed to support healthcare leaders and innovators as they navigate the complex journey of developing and integrating AI into clinical workflows. Partnering with NVIDIA on this has been instrumental in leveraging their deep expertise in computing platforms, which will help bring a higher level of innovation to healthcare systems. As we look ahead, I believe BRIDGE will play a pivotal role in helping healthcare systems realize AI’s transformative potential, empowering them to scale AI faster, with greater confidence, and deliver more efficient, higher-quality care to patients. Exciting times ahead! https://okt.to/D20fZU

311

12 Comments

Big healthcare AI news yesterday as Tempus AI received the first-ever CPT code for an algorithm-only analysis of previously-sequenced, laboratory developed test transcriptomic data. TLDR: AI molecular phenotyping has a new business model. The PurISTSM test uses RNA profiling to identify molecular subtypes in patients with advanced pancreatic ductal adenocarcinoma, which may inform first-line therapy selection and management. While I'm skeptical that payer balance sheets are going to solve the "who pays for healthcare AI" problem, the reimbursement pipes for clinical AI are maturing. I am surprised to see more progress for usage-based vs. performance-based models given the many challenges with FFS economics. It will be interesting to see how payers assess the "medical necessity" and therefore reimbursability of clinical AI as these models come to market. https://t.co/eLe18SmihV

420

9 Comments

Over 950 medical devices that are AI augmented have been approved to date by the FDA and there are thousands of startups in the health tech space. 1. How do we ensure best patient outcomes and not aggravate bias or inequities? 2. What are the optimal methods for workflow integration? 3. How can we best prepare users and stakeholders such as doctors and clinicians to ensure safety, privacy and best practices? I am excited to speak on these topics at the upcoming AI Med conference in Orlando, Florida in November. As AI and innovation evolves, it is essential that the healthcare team understands the fundamentals and is well-prepared to ensure optimal results. Join me! Harvey Castro, MD, MBA.Anthony Chang, MD, MBA, MPH, MS Gemma Lovegrove Sharief Taraman, MD, DABPN, DABPM, FAAP Emily Blum Mara Schenker, MD, MBA Amanda A. Jhonatan Bringas Dimitriades, MD Austin Lee Chiang, MD MPH Daniel Kraft, MD #ai #innovation #outcomes #healthequity #healthcare Madhu Mazumdar Nancy Spector Healio

53

2 Comments

AI Healthcare: “All the single images I gave it were sufficient context for a medical student or an early radiology trainee to make the diagnosis,” Heacock added. “Grok is not suitable for medical diagnosis in its current form. I agree that may change in the future, but that wasn't the intent of this test.” ----- RadiologyBusiness: "Elon Musk urges users to submit X-ray, PET and MR images to xAI chatbot Grok" (Marty Stempniak) (October 30, 2024) "Elon Musk is urging patients and providers to try submitting their X-ray, PET and MR images to his company’s artificial intelligence chatbot for analysis. Elon Musk shared the request on Twitter Tuesday, asking users to share where “Grok gets it right or needs work.” “This is still early stage, but it is already quite accurate and will become extremely good,” Musk, wrote Oct. 29. Radiologists and other physicians responded to the request Tuesday. French rad Dr. Thibaut Jacques input the last MR image in his queue, displaying adhesive capsulitis of the shoulder, which he deemed as “level: easy, typical pattern.” He said the results were mostly “not true,” with Grok’s output “too generic” and lacking a diagnosis. Laura Heacock, MD, a breast radiologist, deep learning researcher and associate professor at NYU, also responded. She previously ran a series of breast mammograms, ultrasounds and MRIs through GPT-4, sharing the results on Twitter. Heacock repeated the experiment with Grok, to disappointing results. The AI model thought a breast MRI was an image of the brain on the first try. On the second case, Grok correctly identified that it was looking at a breast MRI (which contained a “very obvious" left breast cancer). “However, Grok seemed unable to tell me anything more than a description of the ‘brightness’ evenness across the image,” Heacock added. “Correct identification, no points for diagnosis. Half point for probably realizing the left breast is abnormal.” After several more attempts with different medical images, Heacock shared her final assessment: --“Grok is OK at recognizing breast radiology image types, with some exceptions. --Grok is wrong at diagnosing breast imaging, although that capability may just be locked. --I prefer the explanations (even if wrong) to GPT-4v. --Real radiologists >>>>> Grok radiologist.” “All the single images I gave it were sufficient context for a medical student or an early radiology trainee to make the diagnosis,” Heacock added. “Grok is not suitable for medical diagnosis in its current form. I agree that may change in the future, but that wasn't the intent of this test.” “Nice work here,” responded Doug Lake, MD, a radiologist based in Ames, Iowa. “These are slam dunk, easy cases and it’s fascinating the tech sometimes struggles to recognize the modality and can’t nail down easy diagnoses that a medical student would get It will get better, but it has a long way to go.” RadiologyBusiness: https://lnkd.in/guc9ctZD #ai #fakeai #genai #degenai #aihealth

8

6 Comments

🔍 Beyond ChatGPT: Managing Autonomous Medical Agents ⚡ Your AI isn't just making suggestions anymore - it's taking action. The paper The 2024 ACM Conference on Fairness, Accountability, and Transparency outlines 3️⃣ critical mechanisms for AI agent visibility 👇 🏷️Agent Identifiers  Digital fingerprints mark every AI agent interaction across healthcare systems. These identifiers track autonomous decisions, from medication adjustments to treatment planning. Each identifier remains tamper-resistant and follows the agent across all system interactions. 👀Real-Time Monitoring Healthcare systems require continuous oversight of AI agent behavior. This enables: 🚨 Immediate detection of dangerous clinical decisions 🔒 Prevention of unauthorized data access 🔄 Tracking of multi-agent interactions across departments 🗒️ Activity Logging Comprehensive documentation creates an audit trail for every AI agent action. This system captures all inputs, outputs, and state changes across the healthcare ecosystem. 🏥Healthcare Takeaways The stakes in healthcare demand clear visibility into agent actions. Patient safety requires complete tracking of every autonomous decision. Clinical workflows must integrate seamless monitoring without compromising care delivery. Multi-agent interactions pose unique risks in healthcare settings. Consider an AI agent managing medication orders interacting with another handling lab results - both systems must maintain perfect coordination while preserving patient safety. The visibility framework enables: ⚡ Rapid intervention for potentially harmful decisions  ✅ Complete accountability for autonomous actions  🔐 Protection of sensitive patient information  🔄 Cross-system coordination tracking This represents a fundamental shift from passive clinical decision support to actions, requiring new governance frameworks unique to healthcare settings. #aiagents #genai #llm #healthcareai #digitalhealth

19

3 Comments