Tatyana Kanzaveli

University of California, San Francisco Chancellor, Sam Hawgood, in an interview with Matthew Lungren MD MPH at #ConvergeSummit has outlined the use of #AI scribes to create clinical notes as well as other #AI-driven technologies. Key to their success: a group of practicing physicians with knowledge of AI ensuring that the technology fits into the physicians’ existing workflow, which improves physician adoption of the technology. Another key: measure success through reduction in physician burnout and improvement in patient satisfaction, not increases in physician productivity. Final key: have a real-time, frequent review of implemented technologies to ensure that they still support and are relevant to changing clinical care standards - and shut the technologies down if they are not supporting their purpose.

10

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

Highlights from the Chandra Ramanathan facilitated panel on #AI innovation with Petrina Kamya, Ph.D., Chris Regan, Anna Marie Wagner, and Ramesh Durvasula during MassBio #AlignSummit24 included: “50% of current pre-clinical assets are AI driven” “Why do we continue to think about sequential drug development when we can use AI to design molecules that address downstream issues?” “We need positive and negative data as well as feedback loops to continue to improve an AI model.” “The model needs to inform us on what additional data should be collected to drive improvement.” “We need more highly diverse, multi-modal data.” “Knowledge is now a commodity. Differentiation now comes from speed.”

34

What are important benchmarks you use with LLMs with #AIinhealthcare? Can #LLM excel in medical encoding? Yes! But what is your measurement of success? That you found a billable code or is it the RIGHT billable code? Does it help you ALSO find the RIGHT non-clinical code because of Regulatory pressures like #USCORE and #USCDI? Or for Research, Adverse Events, Observational outcomes, etc...? Questions you should be asking yourself in measurements of success: - Did you benchmark manual review vs LLM review? Was the encoded encounter the same code that left the organization on the bill? Was there a human in the loop to confirm its accuracy? - Was the encoded encounter changed from point of care to #CDI? Did CDI recommend a change after LLM? - Is the LLM recommending other code sets for the encounter because encoding is only a portion of the clinician's need? Are they the right codes or purely a crosswalk? - Did the LLM provide you an international version of your encoding requirement or did they give you a US version (AKA training Hallucination)? Here at IMO Health, we leverage terminology services as an expert system to couple IMO Clinical AI to train LLMs. This greatly improves the accuracy and support of medical encoding and supporting the overall F-Score of what truly occurred during the encounter. If you’re curious about how technology can positively impact medical coding accuracy, this read is for you. https://bit.ly/3VBZFqn  #DataQuality #NLP #ClinicalAI #ClinicalTerminology

10

2 Comments

As the discourse shifts from models to compound AI systems / agents, we need better AI benchmarks to evaluate multi-modal and multi-step task performance, especially in healthcare and life sciences. When we wrote the first paper demonstrating ChatGPT's performance on the USMLE, we chose the US Medical Licensing Exam as a benchmark for accessibility, speed, and ease. This benchmark was never intended to represent AI model performance on real-world clinical tasks. Today, I still see so many research teams and startups using benchmarks (like the USMLE) that are ill-fitted for assessing the true clinical or scientific performance and utility of the models they are developing for real-world contexts. Benchmark development may be seen as a "less sexy" area of research, but it is of paramount importance. Years after the rise of the transformer, we still lack adequate benchmarks for so many single-step tasks in biomedicine. With compound AI systems (i.e., architectures that integrate multiple AI models to perform complex tasks) emerging, we need new benchmarks for agentic behaviors. I'd argue that developing an agent with novel capabilities without at least proposing a companion benchmark (if an industry standard does not yet exist) may hinder the adoption of said agent, especially for high-stakes workflows. Designing more benchmarks that capture/simulate real-world clinical and scientific workflows will help us mitigate the major discrepancies observed between in silico and in vivo performance and better support safe + effective deployment of AI in biomedicine. There are already brilliant people focused here, and we need more. DMs are open if you're researching or working in this area of multi-step/multi-modal benchmarking in healthcare and life sciences! #healthcare #ai #artificialintelligence #generativeai

100

18 Comments

As a hot topic, you can imagine there are strong opinions on both sides of the spectrum. Supporters of the veto believe what was proposed went too far and would completely hamstring AI progress and innovation while grossly-restricting access to the technology. Opponents of the veto believe this bill barely scratched the surface in limiting tech companies’ excessive power and influence with the veto resulting in inevitable harm to humanity from non-governed AI development. So, who’s right? Neither. As with everything, both sides have legitimate points and concerns that should be taken into consideration. Both sides are attempting to validate their position by comparing their steel-man arguments to their opponents straw-man arguments. Both sides are also approaching the situation as a zero-sum game, resulting in an unwillingness to compromise. For the sake of everyone, we can do better. Perhaps, instead of choosing a polar end of the spectrum and doubling down on a no-holds-bar competition for victory, both parties can work, with reasonable comprompromises, toward a mutually beneficial solution that puts reasonable governance and oversight on what we already know to be the most powerful technology in human history while continuing innovation and democratizing it’s capability and access for the benefit of humankind. #ai #governance #technology

1

Regulating #AI is a sticky wicket. As is often the case, California is ahead of the curve. SB 1047 is the first legislation to impose safety mandates for large language models. I'm as eager as anyone to see some guardrails put in place, but the risks of getting it wrong are not trivial. Solid reporting here via @jacoberobbins of PitchBook #artificialintelligence #regulation #machinelearning

10

1 Comment

Curious about AI and Healthcare? Learn more right here!

6

There appears to be a "Superstar Effect" in healthcare and biomedical AI research right now. A small percentage of publications are truly novel methods contributions that introduce new capabilities or behaviors; whereas a majority of contemporary research publications are focused on implementation science to assess the real-world utility and performance of new methods. This dynamic seems different from prior eras, when the translational infrastructure for implementing novel AI methods was largely unavailable. While methods research often captures more attention and accolades, both types of research are critical – novel methods without implementation science lacks translatability; implementation science without novel methods lacks novelty. If our goal is to invent and deploy transformative AI methods in real-world healthcare and life sciences contexts, we ought to continue encouraging and rewarding stellar implementation science work alongside breakthrough methods development.

90

3 Comments

🚨 Researchers at UCLA Health Jonsson Comprehensive Cancer Center have unveiled a dual role of the protein interferon regulatory factor (IRF1) in cancer treatment, shedding light on its complex behavior in tumor progression and immune response. This protein can either hinder or help the body's fight against cancer depending on its cellular environment. Their findings, published in Cell Reports, suggest that manipulating IRF1 levels could significantly enhance the effectiveness of immunotherapy, providing new hope for cancer patients. Dive into this groundbreaking study to see how IRF1 could revolutionize cancer treatment strategies! 💉🔬✨ #CancerResearch #Immunotherapy #MedicalInnovation #HealthcareTransformation Read more: https://lnkd.in/gDCe-Sie

6

🌟 BREAKTHROUGH DISCOVERY IN CANCER GENETICS 🌟 UCLA researchers have published a groundbreaking study in Nature Communications that revolutionizes our understanding of how noncoding DNA influences cancer progression. Their discovery of mutations in these "silent" regions significantly impacting mRNA levels, leading to abnormal protein production which can accelerate or hinder cell proliferation, challenges previous assumptions about noncoding DNA's role. The study also introduces a new prognostic tool, the untranslated tumor mutation burden (uTMB), which could predict patient survival in certain cancers. This pioneering research paves the way for developing targeted treatments and personalized medicine strategies, offering new hope in the fight against cancer. #CancerResearch #Genetics #InnovativeScience

16

This California Management Review article explores the ROI of AI ethics and governance investments through two lenses: loss aversion and value generation. While many companies focus on minimizing risks like legal penalties and reputational harm, the article argues that a value generation approach—using ethics to build trust, foster innovation, and create competitive advantages—offers greater long-term benefits. By integrating ethics into strategic planning, organizations can both mitigate risks and enhance overall growth.

49

3 Comments

One of the most fascinating discussions you can watch on what #AI can do for science. It really does make you dream https://lnkd.in/eGBf9qGF

3

Will Generative AI Transform Healthcare? Uncover the Opportunities and Challenges. Generative AI is poised to transform healthcare, offering unprecedented capabilities in data management and patient interaction. As investments from both tech giants and innovative startups flood this emerging sector, the potential for groundbreaking developments is significant. However, this enthusiasm is tempered by considerable concerns. The readiness of AI systems, coupled with issues of bias and privacy, has prompted calls for a cautious approach. Healthcare professionals and industry advocates stress the importance of establishing stringent regulatory frameworks and conducting thorough testing to ensure that AI technologies are both safe and effective. For investors, the integration of AI into healthcare presents a unique dual opportunity: to lead the charge in technological innovation while addressing the ethical and practical challenges it brings. This careful balance of advancement and accountability is crucial in navigating the future landscape of healthcare technology. https://hubs.la/Q02t4NwV0 What are your views on AI’s role in modern healthcare? Join the conversation below and share your insights.

19

2 Comments

NVIDIA's latest move brings generative AI to the forefront of healthcare innovation. This fusion of cutting-edge technology and medicine promises to revolutionize patient care and empower biomedical professionals. The future of healthcare is being rewritten - are we ready for the transformation? #GenerativeAI #HealthcareInnovation #PrecisionMedicine #NVIDIA #AIinHealthcare What potential breakthroughs do you foresee from this convergence of AI and healthcare?

8

2 Comments

The FDA’s latest Executive Summary highlights a significant evolution in regulatory thinking for generative AI (#GenAI) devices. While previous guidance addressed AI broadly, this new approach focuses on the unique complexities of GenAI: + Foundation Models: Acknowledges challenges in managing and regulating third-party models that power GenAI, emphasizing the need for transparency and adaptability. + Non-Deterministic Outputs: Spotlights risks like "hallucinations," requiring stricter controls and tailored performance metrics. + Postmarket Monitoring: Proposes new strategies to address GenAI’s continuous learning and real-world variability. This marks a shift toward a more dynamic, lifecycle-based approach to ensure safety. What are the implications for entrepreneurs? Transparency is crucial—entrepreneurs must document their model's design, training data, and performance metrics, even when leveraging third-party foundation models. Narrowly defined use cases are equally important to ensure products are safely and effectively evaluated, mindful of the competing pressure to create applications with broad utility (and market potential). Startups should prioritize robust monitoring systems to track real-world performance, mitigate risks like bias or hallucinations, and maintain consistent outputs across diverse deployments. As always, human-in-the-loop safeguards and transparent, user-centric designs can enhance trust and safety. FDA’s lifecycle approach means compliance doesn’t stop at approval—continuous evaluation, updates, and monitoring will be increasingly expected to keep products aligned with regulatory and patient safety standards. Explore the full document for insights: https://lnkd.in/dXefRaaS #wearenina #healthcaretechnologies Nina Capital

77

5 Comments

The "Lab in a Loop" framework presented by Aviv Regev from Genentech is essential to driving accelerated innovation in drug discovery --> Experiments --> Data ---> AI/ML ---> Experiments --> Data ---> AI/ML --->Experiments --> Data ---> AI/ML It's an essential framework in the broader ecosystem of leveraging advancements in AI, but there are other fundamental considerations that we must keep in mind in this pivotal and transformative period.

39

1 Comment