Generative AI News Round Up – August 2024

Cranfield, UK, 3rd September 2024, Written by Vlad Kozynchenko –

Thank you for tuning in to this month’s edition of the latest news in generative AI in healthcare. Although it is summer and people should enjoy their holidays, this month has been overflowing with generative AI announcements. Below, I have curated some of my favourite ones.

Epic News

Epic has made a splash this month, as you have all heard about. In a nutshell, this happened.

At Epic’s recent UGM, AI was a major focus, with several noteworthy announcements:

• Speculative AI Bot: Epic showcased an early demo of an AI bot integrated with MyChart. This bot conducted a virtual follow-up after a fictional wrist surgery. It used voice and video inputs to assess recovery, compare it with similar cases from Epic’s Cosmos data, and manage follow-up appointments. This demo was notable for being a functional prototype rather than just conceptual slides, emphasising Epic’s commitment to exploring AI’s potential in patient interactions.
• Epic Payer Platform: The Payer Platform continues to evolve, incorporating AI to automate processes like prior authorisations, claims processing, and provider directories. This aims to streamline the interactions between payers and providers, potentially transforming administrative workflows through AI.
• Lookalikes: This AI feature searches Cosmos’ vast data set to find patients with similar conditions, helping clinicians connect with peers treating similar cases. It aims to enhance collaborative care by improving peer connections based on data-driven insights.
• Best Care Choices: This tool uses patient data to recommend treatments by comparing how other Epic users have treated similar cases. It leverages AI to support clinicians in choosing optimal treatment plans by analysing outcomes and practices from across the Epic network.
• Cosmo AI: Epic introduced Cosmo AI, which will utilise Cosmos data to develop high-quality healthcare AI solutions, though detailed information was limited.
• Ambient Clinical Voice Solutions: Epic is expanding its support for AI-driven voice solutions that capture and transcribe clinical conversations, streamlining documentation and reducing clinician workload.
• Conversational Search: Epic is developing a generative AI feature to provide summarised responses from EHR data, aiming to simplify information retrieval for clinicians.
• Aura and AI Integration: Epic’s Aura platform, which focuses on precision medicine, incorporates AI for improved diagnostic and genomic data management. This includes collaborations with medical device companies and expansion into new areas like continuous glucose monitoring.
• Patient-Facing AI: Epic is exploring AI-driven tools that interact directly with patients, including scheduling and managing appointments based on real-time assessments.
• Here are the programs the EHR vendor has coming (and their expected release dates).

Epic also announced that it would move its customers to TEFCA by the end of 2024, with the full transition expected by 2025. TEFCA aims to standardise health information exchange across the U.S. Epic’s subsidiary, Epic Nexus, and Carequality, an interoperability network including Epic, will align with TEFCA. Epic will maintain support through Carequality during the transition.

Earlier this month, Epic released its second Almanack volume, covering several key topics. Generative AI took centre stage. I have written about it earlier here.

Talkdesk Copilot is now integrated into Epic Cheers, enhancing the personalisation and efficiency of patient interactions through AI tools. This integration, part of Talkdesk’s Healthcare Experience Cloud, provides real-time transcripts, recommendations, and automated summaries to support agents in delivering better patient care and improving value-based outcomes. Talkdesk and Epic’s collaboration aims to make contact centres more effective and integrated within healthcare organisations.

Epic has awarded Solventum’s autonomous coding solution the Toolbox designation for its Fully Autonomous Coding category. This designation means Solventum’s technology, which automates and streamlines medical coding from patient encounter to billing, integrates seamlessly with Epic’s systems. Solventum’s solution addresses challenges like coding staffing shortages and budget constraints while ensuring accuracy and providing full customer control over automation.

Enhancing the Performance of Generative AI Models

Beyond Epic announcements, several new healthcare-specific foundation models have been announced. Considering the news, I wanted to dive deeper into how vendors can enhance foundation models. Prompt engineering and customisation efforts can enhance generative AI models. The graphic below summarises the approaches.

Prompt Engineering

InsideTracker has announced Ask InsideTracker, an AI-powered chat feature designed to provide real-time, science-backed health information. This new tool, built on OpenAI’s GPT-4, draws from InsideTracker’s extensive library of expert-written content to offer personalised answers on wellness topics. Ask InsideTracker likely leverages Retrieval-Augmented Generation (RAG) technology, which enhances its capabilities by combining information retrieval and generation. The RAG approach allows the system to search through an extensive database of up-to-date health articles and research, retrieving relevant content to inform its responses. The generative model then processes this information to create accurate, contextually relevant answers, ensuring users receive reliable health advice tailored to their queries. This integration of RAG technology helps Ask InsideTracker provide timely and precise insights while maintaining the quality and relevance of the information shared. My colleague Alan Stoddart published an Insight that dives into the topic in more detail.

Customisation

In August, Google, Cerebras, and Writer introduced significant advancements in foundation models, each bringing distinct capabilities and applications to the forefront of AI technology. These additions reflect diverse approaches to model development and utilisation, highlighting the evolution and specialisation within AI.

• Google’s Health Acoustic Representations (HeAR) is a bioacoustic model designed to enhance disease screening and diagnosis through sound analysis. Trained on a vast dataset of 300 million audio samples, including 100 million cough sounds, HeAR excels at identifying health-related patterns in acoustic data. Salcit Technologies in India is currently utilising its technology through its Swaasa® product, which assesses lung health by analysing cough sounds. HeAR aims to improve the early detection of tuberculosis (TB), a disease often plagued by diagnostic delays. This model is expected to make TB screening more accessible and affordable. Google Research is making the HeAR API available for researchers interested in leveraging this technology for their projects.
• Cerebras has introduced the DocChat suite, focusing on document-based conversational question answering. This suite includes two key models: Cerebras Llama3-DocChat and Cerebras Dragon-DocChat. Llama3-DocChat is an LLM based on the Llama 3 architecture, while Dragon-DocChat is a multi-turn retriever model. Llama3-DocChat was trained in a few hours, and Dragon-DocChat was fine-tuned in minutes. Cerebras has released the model weights, training recipes, and datasets to the AI community, which are available on HuggingFace. Llama3-DocChat was trained from Llama 3 8B base, and therefore is subject to the META LLAMA 3 COMMUNITY LICENSE AGREEMENT. Furthermore, it is trained on ChatQA’s synthetic conversational QA dataset, which was generated using GPT-4. As a result, this model can be used for non-commercial purposes only.
• Palmyra-Med, developed by Writer, is a specialised language model tailored for the healthcare sector. It claims to outperform models such as GPT-4 and Med-PaLM-2 in medical contexts. Built on the Llama architecture, Palmyra-Med incorporates high-quality biomedical data and uses Direct Preference Optimization (DPO) and a custom Medical Instruct dataset. It features a context window of 8192 tokens and is available under an open model license for non-commercial and research purposes. Palmyra-Med excels in analysing clinical notes, summarising electronic health records (EHR) data, and identifying medical concepts. Despite its fewer parameters than some competitors, it performs well across various biomedical datasets. The model includes watermarks to prevent misuse and is designed to support clinical decision-making and research, though it is unsuitable for direct patient care.

HeAR’s advancements in bioacoustic health diagnostics offer the potential for significant improvements in detecting diseases such as tuberculosis from cough sounds. However, integrating this technology into existing healthcare systems and addressing regulatory and technical challenges will be crucial for its widespread adoption. Similarly, Palmyra-Med’s watermarks and restrictions against direct patient care underscore the need for careful management to mitigate potential inaccuracies and biases. Ensuring the safe and effective use of these models in clinical settings will necessitate ongoing oversight and validation. Both Writer’s and Cerebras models face restrictions on commercial use, which may limit broader applications but help prevent misuse and foster ethical research practices.

While these new foundation models are currently designed for research purposes and are unlikely to see commercial deployment soon, simpler RAG applications are already being introduced to the market. Combining customised models with prompt engineering techniques could potentially enhance results, though the high cost of customisation raises questions about its practicality for low-risk use cases.

Despite these limitations, there is growing optimism about the future of generative AI in clinical settings. For instance, research from UCL has shown that a foundation model for retinal images consistently outperforms several comparison models in predicting complex systemic disorders such as heart failure and myocardial infarction. This suggests a promising future where generative AI can effectively assist doctors with diagnoses and serve as a valuable preliminary tool for identifying secondary findings.

Thank you for investing your time in this update, and I wish you a fantastic day ahead! 👋

Did you find this round-up useful?  The Signify Premium Insights require a subscription, but I’m happy to provide you with a free 30-day trial. Click below to sign up for a no-obligation trial.

If you would like to learn more about the generative AI healthcare landscape feel free to reach out, alternatively you can download the brochure to learn more about Signify’s market coverage in the area by clicking on the link below.

About the Author

Vlad joined Signify Research in 2023 as a Senior Market Analyst in the Digital Health team. He brings several years of experience in the consulting industry, having undertaken strategy, planning, and due diligence assignments for governments, operators, and service providers. Vlad holds an MSc degree with distinction in Business with Consulting from the University of Warwick.

About the AI in Healthcare Team

Signify Research’s AI in Healthcare team delivers in-depth market intelligence and insights across a breadth of healthcare technology sectors. Our areas of coverage include medical imaging analysis, clinical IT systems, pharmaceutical and life sciences applications, as well as electronic medical records and broader digital health solutions. Our reports provide a data-centric and global outlook of each market with granular country-level insights. Our research process blends primary data collected from in-depth interviews with healthcare professionals and technology vendors, to provide a balanced and objective view of the market.

About Signify Research

Signify Research provides healthtech market intelligence powered by data that you can trust. We blend insights collected from in-depth interviews with technology vendors and healthcare professionals with sales data reported to us by leading vendors to provide a complete and balanced view of the market trends. Our coverage areas are Medical Imaging, Clinical Care, Digital Health, Diagnostic and Lifesciences and Healthcare IT.

Clients worldwide rely on direct access to our expert Analysts for their opinions on the latest market trends and developments. Our market analysis reports and subscriptions provide data-driven insights that business leaders use to guide strategic decisions. We also offer custom research services for clients who need information that can’t be obtained from our off-the-shelf research products or who require market intelligence tailored to their specific needs.

More Information

To find out more:
E: enquiries@signifyresearch.net
T: +44 (0) 1234 986111
www.signifyresearch.net