Healthcare AI News 11/6/24

News

Adventist Health Rideout uses Viz.ai to reduce the time that is required to transfer suspected stroke patients to a hospital that offers specialty care from 202 minutes to 109. The system analyzes the CT scan and alerts the care team if it detects a potential stroke. The data is preliminary, as the hospital has used the technology on just 10 patients since it launched in May 2024.

OpenAI adds web searching capability to ChatGPT. It also collects up-to-date search results from news and data providers and displays them in visual categories such as weather, news, and maps. Results include source links. Unlike Google’s search results, it does not display promoted links, advertisements, or sources that are of questionable quality.

OpenNotes and Abridge partner to research the effectiveness of patient visit summaries that are generated from ambient listening by having those patients evaluate them.

Business

Aignostics, which applies AI to digital pathology to support precision medicine, raises $34 million in a Series B funding round. Mayo Clinic participated in the round and will work with the company to develop foundation models and biopharma product offerings.

Snoop Dog-backed cannabis biotech Oxford Cannabinoid Technologies collaborates with Oxford University to use AI drug discovery to develop non-addictive, cannabinoid-based pain medications. The company also announced that its shares will be de-listed because they have lost 97% of their value, leaving it with a market cap of $3 million.

France-based insurance startup Alan, which allows customers to ask questions to doctors via a chat interface and receive an answer back in 15 minutes, adds a virtual assistant to the feature. The virtual assistant, which is called Mo, rephrases the question and then asks if the person would rather interact with a doctor or Mo. Conversations with Mo are checked within 15 minutes by a doctor who can correct its recommendations. The company says it will enhance the tool to provide personalized guidance based on context and the user’s health history.

Research

The National Cancer Institute awards Pieces Technologies a $2 million grant to develop a conversational AI agent that will allow cancer patients the ability to ask questions about their care. The system, which will also collect social determinants of health information, will be co-developed by MetroHealth.

Penn State researchers find that patients are more satisfied with using a medical AI chatbot if the system remembers their social information and makes small talk about their job and hobbies. 

Other

Nvidia’s healthcare VP says that the next wave of AI will turn medical devices into robots. Kimberly Powell predicts that “This physical AI thing is coming where your whole hospital is going to turn into an AI.”

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 10/23/24

News

Amazon One Medical launches AI tools for its 1Life proprietary EHR – ambient documentation, a summarized medical history, draft responses to patient messages, and workflow routing.

Anthropic releases a beta developer version of Claude’s API that adds “computer use” capability, in which developers can program interaction with a user’s computer such as looking at a screen, moving the cursor, clicking buttons, and typing text.

Google Cloud announces GA of Vertex AI Search for Healthcare along with new features of its Healthcare Data Engine.

GE HealthCare announces CareIntellect for Oncology, which summarizes clinical reports, flags deviations from a patient’s treatment plan, and identifies relevant clinical trials.

The Coalition for Health AI (CHAI) publishes draft frameworks of how it will certify independent quality assurance labs and standardize their test results into what it compares to a nutrition label for AI product performance and safety.

Aidoc and Nvidia will co-develop a framework for integrating AI into clinical workflows.

Business

Ozarks Healthcare will integrate Avo’s AI clinical decision support with its Meditech EHR. The tools include scribing, chart summary and care guide, and the ability for clinicians to ask clinical questions about the patient’s information and relevant medical evidence.

Zoom will integrate Suki’s ambient documentation into Zoom Workplace for Clinicians, a recently announced paid offering.

Atropos Health announces GA of ChatRWD, an AI co-pilot that generates real-world evidence to answer clinical questions.

Tennr, which offers AI-powered healthcare document processing, raises $37 million in Series B funding.

AI-driven patient preference and informed consent system vendor HealthEx raises $14 million in seed and Series A funding.

Artera announces new AI co-pilots to its Harmony patient communications platform: Staff (translation, predictive text for patient inquiries, message shortening, and conversation summaries that can be saved to the EHR) and Insights (no-show reports).

HCA Healthcare selects Commure to develop and deploy ambient AI. Commure closed its $139 million acquisition of ambient documentation technology vendor Augmedix in July 2024.

Research

Researchers develop a system of AI-powered GoPro cameras that detects when the wrong drug syringe or vial is being used to prepare a patient’s doses, which the authors note could prevent serious medication errors in ORs, ICUs, and EDs.

Other

Harvard Medical School adds a required month-long AI course for students in its MD/PhD translational medicine program and engineering program that it offers with MIT.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 10/16/24

News

Patient engagement startup Parakeet Health launches its generative AI voice platform for health system contact centers and raises $3 million in seed funding.

Luma Health announces AI products for healthcare call centers: a fax processing tool and a patient-facing voice AI concierge.

Business

MIT News profiles alumni-founded Ambience Healthcare, which offers an ambient documentation solution that is being used in 40 large institutions.

Johnson & Johnson lists six ways it is using AI:

• Creating and analyzing videos of surgeries to create “highlight reels” for training, collaboration, and support.
• Create 3D maps for surgical procedures.
• Analyze de-identified EHR data to identify disease targets and drugs.
• Support clinical trial recruitment.
• Analyze genomic and clinical data from diagnostic tests to target treatments and identify candidates for clinical trials.
• Predict drug supply and demand to optimize distribution of products where they are needed most.

Research

A retrospective cohort study of Yale New Haven Health System medical-surgical inpatients finds that the accuracy of six AI-powered early warning tools for patient deterioration varies widely. ECARTv5 and the National Early Warning score were the top performers, while Epic’s Deterioration Index had the lowest positive predictive value. The authors recommend that health systems verify how the tools work and oversee their use.

Other

Angela Elena Olazaran Laureano, a 17-year-old student in rural Mexico, is chosen from 11,000 nominations to win a $100,000 student prize for her work developing an AI-powered virtual medical assistant that provides basic diagnosis for 21 conditions and warns users if they are likely to be contagious. She was previously part of robotics team that won the national Home Care Challenge. She will use the award to create a STEM classroom in her home town.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 10/9/24

News

WVU Medicine implements an AI-powered search system that allows employees to search for content that is contained in internal documents, such as clinical protocols, how-to guides, and tip sheets. It plans to eventually offer searching across all of its sites as well as across all Epic-using hospitals.

A newly signed California bill requires providers to include a specifically formatted disclaimer when AI was used in patient-facing clinical communication that has not been reviewed by a licensed provider. The communication must also contain instructions for the patient to contact an appropriate person.

HHS Assistant Technology for Technology Policy and AI leader Micky Tripathi, PhD, MPP says that HHS is looking at ways to create AI models using its vast data stores, but he cautions that its data is inherently biased since it excludes people who don’t have insurance or who don’t seek care from providers.

Wyoming’s Medicaid program approves coverage of Canvas DX, an online tool that assesses autism using family-provided information and videos. Families can use the tool instead of trying to find one of the few state clinicians that can administer the ADOS test.

The American College of Radiology will participate in the FDA’s program to boost innovation in breakthrough devices. FDA has expanded the program for its cardiology device origins to include neurology, ophthalmology, and radiology.

Business

The authors of a literature review on AI in healthcare suggest that its public health impact will be limited by the “politics of avoidance,” where the US health system’s treatment-focused model ignores broader social determinants of health. They express concern that the hype around AI may overshadow more effective, evidence-based interventions like using community health workers and implementing harm reduction programs, which address root causes rather than just offering suggestions for treating symptoms.

Research

A test of GPT-4’s ability to provide clinical recommendations for ED visits – including admission status, radiology requests, and antibiotic prescriptions – finds that it performs poorly compared to medical residents. Its recommendations were often overly cautious and called for medically unnecessary admissions and orders.

A review of AI-powered, FDA-approved medical devices finds that most did not include race or ethnicity, socioeconomic data, and the ages of study participants in their submitted approval documentation. The authors conclude that the lack of consistency and data transparency may exacerbate health disparities

Other

Drug maker Eli Lilly and Company names Thomas Fuchs, DrSc as its inaugural chief AI officer, where he will be tasked with leading the company’s AI initiatives in drug discovery, clinical trials, and manufacturing. He was previously chair of the AI program at Icahn School of Medicine at Mount Sinai.

Mayo Clinic will use a $25 million donation to provide funding and work time for its clinicians to purse AI-related projects.

Four high-profile cancer centers use $40 million in tech company funding to form the Cancer AI Alliance, which will study their collective patient data to find clinical insights.

US Army researchers develop SeptiBurnAlert, which uses AI to analyze the blood samples of burn patients to detect sepsis-related components. The system predicts risk within the first 24 hours of hospitalization with high accuracy. The team has applied for a patent and hopes to bring the product to the commercial market in three years, pending FDA approval. They are also working with companies that want to product a handheld device.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 10/2/24

News

Microsoft releases a major enhancement to Copilot that adds conversational capabilities, a virtual news presenter that will read headlines, webpage history recall, and the ability to answer questions about the text and images on a browser page.

FDA will hold the first meeting of its Digital Health Advisory Committee on November 20-21, which will address lifecycle considerations for AI-enabled medical devices. The session will be available via webcast with no registration required.

Business

Claimable launches an AI-powered appeals platform to provide documentation to help patients protest health insurer medication coverage denials for 60 major treatments. The service costs $39.95 per appeal and the company submits appeals via fax and first class mail. It claims that 80% of its appeals are accepted and most cases are resolved in less than 10 days.

WellSpan Health uses Hippocratic AI to develop an AI-powered conversational agent that contacts patients by telephone to close gaps in care.

GE HealthCare closes its $51 million cash acquisition of the ultrasound AI business of Intelligent Ultrasound Group.

Research

Clinicians compare the usefulness of OpenAI’s o1-preview, which features enhanced reasoning, to ChatGPT GPT-4 for medical AI:

• The new version can perform advanced, step by step reasoning. The effect on diagnostic accuracy has not yet been studied.
• o1-preview is slower.
• GPT-4 is good for patient communication and medical advice, while the new version is better at complex reasoning, genetic analysis, and research. The value of those changes for specific medical specialties has not yet been studied.
• GPT-4 is more prone to hallucinate.
• GPT-4 is better at human-like conversation.
• While transparency is limited with both versions, the new version provides chain of thought for double checking.
• GPT-4 is trained to provide information that is broad but not deep, while o1-preview applies PhD-level reasoning in physics, chemistry, and biology.

A ChatGPT review of 725 websites of clinics that promote “complementary” and “alternative” medicine finds that 97% contain false or misleading claims, including some that relate to cancer treatment.

Other

A woman is told by the ED doctor who sent her home that her face pain and drooping was not concerning gets a second (and correct) opinion from ChatGPT, which advised her to seek immediate medical attention for what was possible Bell’s palsy. She want back to the ED, where the doctor agreed with the diagnosis and started treatment immediately.

A study of LinkedIn and Glassdoor job postings that mention AI finds that Mount Sinai Health System is #6.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 9/25/24

News

The New York Times covers physician use of Epic’s In Basket Art to create draft responses to patient questions that have been submitted using MyChart. Some major health systems decided not to use the technology over concerns that doctors would approve responses without reviewing them, while others thought that patients would recognize the message as AI generated and devalue it. An Epic study found that doctors are sending unedited responses to one-third of messages. Duke Health Chief Health Information Officer Eric Poon, MD, MPH says that Epic’s product creates drafts that are still “moderate in quality,” which keeps doctors vigilant in looking for mistakes, but ponders whether they will let their guard down as the AI gets smarter. A Duke colleague adds that it tried to get Epic’s product to stop giving clinical advice, but “we couldn’t take out its instinct to try to be helpful.”

OpenAI will roll out Advanced Voice to paying subscribers of ChatGPT this week. The enhancement adds speedier conversational responses, the ability to pause when interrupted, and additional voices. However, it reportedly does not support web search, custom GPTs, images, or unlimited use.

Google Cloud releases upgraded versions of its Gemini 1.5 chat-based AI assistant that it says are twice as fast at half the cost, with the capability of handling 1,000-page PDFs and hour-long videos. The company says that its Cloud and DeepMind divisions are developing new AI products and models and that developers are starting to use Gemini to create their own chatbots and voice assistants. Google has added the standalone Gemini app to Google Workspace.

The Association for the Advancement of Medical Instrumentation (AAMI) will work with the Consumer Technology Association to create standards for post-market market surveillance of non-prescription health devices  and for evaluating and assessing AI/ML products.

Business

London-based Noki.ai announces an ambient AI medical companion that can transcribe visit conversions, automate scheduling, manage forms, verify insurance, display a patient dashboard, and exchange data in FHIR format. Monthly pricing ranges from free to $299 based on functionality and usage limits.

Ferrum Health, which offers a secure platform for health systems to deploy AI, raises $16 million in a Series A funding round. Co-founder and CEO Pelu Tran founded Augmedix in 2012, which at that time hoped to commercialize the use of Google Glass for medical documentation. He dropped out of Stanford’s medical school weeks before graduating to work on Augmedix full time, then left the company in 2018 and started Ferrum.

Healthcare AI call agent developer Hippocratic AI adds $17 million to its Series A funding round that was led by Nvidia’s venture arm, increasing its total raised to $137 million.

Research

A study finds that real-time screening of medical claims using AI can reduce healthcare fraud, waste, and abuse by allowing questionable charges to be reviewed before payment instead of after. Tests found that such screening reduced claim payments by 1.2%. 

Researchers from Brigham and Women’s Hospital use AI to improve the accuracy of results from the PREVENT cardiac risk calculator by calibrating it to local populations. The authors conclude that the black box nature of AI applications can be tailored while preserving their functionality.

Other

Singapore’s health minister says that general practitioners will use AI for assessing health risks, prescribing drugs, and recommending lifestyle changes, which will eventually be powered by mandatory use of its national EHR program. Ong Ye Kung told conference attendees that, “We have medical records, we have genome data, we have lifestyle data, we have socioeconomic data, and the technology is already available. We can train very sophisticated, high parameters, AI models to identify risk factors and to predictive preventive care.” He says that Singapore’s biggest health challenge is the “buffet syndrome,” where patients are overtreated because their insurance pays.

AAMC News profiles David Fajgenbaum, MD, MBA, who as a medical student in 2010 was dying of organ failure but was saved by experimental chemotherapy. The Penn Medicine professor created Every Cure, which uses AI to score 3,000 approved drugs for their possible use in treating other conditions, which gets them into the hands of patients quickly and inexpensively. That organization was awarded $48 million in federal funding earlier this year.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 9/18/24

News

An 18-month study by Toronto’s St. Michael’s Hospital finds that its use of the Chartwatch AI system that predicts patient deterioration was associated with a 26% drop in unexpected deaths. The system was developed with startup Signal 1.

Duke Health will partner with SAS to apply analytics, AI, and machine learning to healthcare operations.

A survey of 1,000 UK-based family doctors finds that one-fifth are already using AI in their clinical practice despite a lack of official guidance or work policies. About one-fourth of respondents report using AI to generate after-visit documentation, reviewing possible diagnoses, and suggesting treatments.

Business

Precision medicine technology vendor Tempus AI announces the beta launch of Olivia, an AI-enabled app that organizes the user’s personal health information.

Healthcare AI startup Evidium chooses Oracle Cloud Infrastructure to develop and train its AI models.

The director of professional services of virtual care solution provider OnCall Health by Qualifacts describes how a self-developed ChatGPT-powered form builder tool is saving the cost of at least one FTE. The tool, which took one afternoon to develop, allows customers to create their own forms that ChatGPT then turns into JSON code that the company’s platform can read.

Research

Oregon Health & Science University researchers determine that large language models outperform up to 75% of students on a test from an introductory course in biomedical and health informatics. The LLM outperformed students in answering questions quickly and with proper grammar and spelling. The authors express concern that such assessments could be gamed, especially for online courses whose exams are taken without in-person proctoring.

Other

Yale New Haven Health says that it has 50 AI projects underway and is already using it to predict patient outcomes, offer guidance for therapy selection, automate documentation, and prioritize radiology cases.

Oracle Chairman and CTO Larry Ellison calls for “omnipresent AI cameras” that maintain civil order by making people aware that they are being watched. Ellison says we’re already partly there with ever-present security cameras, police body cameras, and video technology running on doorbells and vehicle dashboards. He also says that high-speed police chases are unnecessary when autonomous drones could follow a car anywhere. He didn’t mention whether that concept could apply to the products of Oracle Health.

I ran across a weird but cool AI app called SocialAI. It’s like a private version of Twitter, except that you will instantly gain thousands of adoring followers that are actually AI bots of whatever type you choose – thinkers, trolls, jokesters, etc. – that will dutifully and realistically interact to whatever you post. It is compelling, entertaining, and perhaps useful for people who are stressed, lonely, creatively blocked, or reflective. My first post was “I’m bored – tell me something motivating” and the responses were realistic and generally useful, especially for folks whose primary social interaction is via keyboard. Imagine (positively and negatively) if the bots were programmed to support someone who has a significant medical condition or who is fretting over an impending medical decision. I posted that I was stressed at work and the responses were empathetic and actionable.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 9/11/24

News

Apple announces Apple Intelligence, which will add AI capabilities to the latest models of IPhones, MacBooks, and IPads with the launch of IOS 18 later this month. It will include:

• Writing tools that can rewrite text, proofread, summarize, and suggest content for an email reply.
• A new version of Siri that supports voice and typing.
• Photo tools for removing unwanted objects, searching by content, and creating user-defined memory movies.
• Phone call recording and transcription.
• Coming later will be Visual Intelligence (searching by taking photo), an emoji generator, custom image creation, and connections to ChatGPT and third-party apps.

Apple adds sleep apnea detection for the Apple Watch using a new Breathing Disturbances metric and AI analysis.

OpenAI will reportedly release its Strawberry reasoning-focused AI as part of ChatGPT within two weeks. The system “thinks” for up to 10-20 seconds before responding to provide better responses. 

Business

Japan’s drug and medical device agency approves IRhythm’s Zio AI-interpreted, 14-day cardiac monitoring patch that patients mail in to the company for analysis and reporting to their physician as an alternative to the Holter monitor.

Salesforce announces out-of-the-box AI services for 15 industries, including patient services and benefits verification for healthcare.

Philips will significantly expand its health technology work in India, especially that related to AI. The company’s innovation campus in Bengaluru houses 5,000 employees who focus on generative AI solutions for healthcare.

Drug maker Gilead Sciences contracts with Genesis Therapeutics to use its molecular AI platform to discover therapies. Gilead will pay Genesis $35 million upfront for three potential drugs, with additional payments for meeting development, regulatory, and commercial milestones. Genesis, which has raised $280 million in funding, is led by 32-year-old co-founder and CEO Evan Feinberg, PhD.

Research

Researchers from MIT, Massachusetts General Hospital, and Harvard Medical School develop ScribblePrompt, which segments biomedical images from human annotations.

Other

Hippocratic AI and healthcare education firm Adtalem Global Education will develop clinician AI training modules that will emphasize the learning needs of nurses.

A survey of UK doctors finds that nearly 90% are worried that patients who use generative AI for medical inquiries will receive misinformation. More than half are enthusiastic for using AI to provide diagnostic and treatment options to doctors themselves, while 83% think that government or medical associations should oversee AI’s use in healthcare.

Saudi Arabia’s King Faisal Specialist Hospital and Research Center, which established a dedicated AI center in 2019, has developed 20 AI-powered diagnostic and treatment applications. The applications automate tasks such as summarizing medical literature and monitoring patient progress, which it says has reduced bed wait times from 32 hours to six.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 9/4/24

News

A legal advisor recommends that doctors not use AI to respond to patient complaints. She says that AI’s responses may use country-specific laws (most likely from the US), generate misleading statements, and ask for confidential information that could breach privacy laws. She also notes that patients may find AI’s wording to be insincere or indicative that their complaint didn’t warrant a human response.

Business

Aidoc adds seven AI solutions for the European health market. Four of them address notification and triage of vessel occlusion, aortic dissection, vertebral compression fractures, and malpositioned endotracheal tube. The others involve quantitative assessments for midline shift, coronary artery calcification, and abdominal aortic measurement.

Imaging-based real world evidence vendor Segmed raises $10.4 million in a Series A funding round, with Advocate Health being among the investors. The company will use the proceeds to expand its use of AI.

Samsung acquires France-based fetal ultrasound AI software vendor Sonio, whose product is FDA 510(k) cleared in the US, for $92 million.

Research

A study finds that AI models can guess a patient’s self-reported race based on technical aspects of radiology images, which could perpetuate diagnosis bias. The authors note that setting a score threshold for the relevant factors can mitigate some of the bias.

Other

Digital transformation leader David Bray, PhD, MSPH says that it’s imperative that AI-generated outputs be labeled consistently to avoid “AI self-cannibalization,” which occurs when AI models are recursively trained on data that was created by previous AI work. He believes that HL-7 standards could be expanded to provide healthcare data provenance to create a traceable, transparent AI ecosystem and to support interoperability.

Actor Tom Hanks warns fans that online ads for “miracle cures and wonder drugs” are using AI-generate images of him, for which he didn’t consent.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 8/28/24

News

Medscape launches a free AI scribe for US physicians that can summarize a patient visit in SOAP, H&P, or POMR format.

Researchers find that about only about half of the 500 AI-powered medical devices that have been approved by FDA were validated on real patient data, raising concerns about their clinical effectiveness and the FDA’s review standards. Three-fourths of the authorizations involved radiology products.

Business

MUSC Health rolls out an Epic-integrated AI agent from SoundHound AI that can answer patient questions and manage appointments. SoundHound acquired Amelia AI, which sells conversational AI agents to multiple verticals outside of healthcare and had raised nearly $200 million, for $80 million three weeks ago. SoundHound went public via a SPAC merger in 2021 at a $2.1 billion valuation, now down to $1.7 billion.

Human Longevity, Inc., which uses AI to identify the early stages of age-related conditions for paying members of its 100+ longevity program, raises $40 million in a Series B funding round.

Other

Google’s API-accessible HeAR AI model (Health Acoustic Representations), which was trained on 100 million cough sounds, serves as the foundation of an India-based company’s tool that assesses lung health and detects tuberculosis.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 8/21/24

News

Epic R&D SVP Seth Hain, MS gives UGM attendees an idea of how the company is thinking about AI use cases. His example was an AI-initiated outreach call to a recovering wrist surgery patient that asks them to rate their pain, instructs them to move their hand in front of their phone’s camera, and then analyzes their range of motion to compare with similar patients in Cosmos to show them their relative progress. Epic also noted that 65 sites are live on Look Alikes, which taps into its Cosmos database to help diagnose rare conditions.

Singapore’s National University Health System is using an AI platform that it developed on Amazon Web Services to draft documents and summarize patient data, which it says has reduced the time required for administrative tasks by 40%. It is working on event-driven models that could perform tasks automatically, such as automatically drafting a discharge letter when a patient is going home.

Business

PayZen, which allows providers to offer their patients AI-powered installment payment plans, raises a $32 million Series B funding round that also includes a $200 million debt facility.

Research

Researchers find that AI analysis of facial expressions from images and video streams can predict health deterioration in patients who are unable to communicate, with an accuracy of 99.8%. Other potential uses include a “health monitoring mirror” for home use and a pre-visit analysis tool for telehealth encounters.

A machine learning model can predict autism spectrum disorder in children under 2 years old by using 28 data points from their medical histories and background. Current methods diagnose the condition at a median of 5 years old, providing an opportunity for earlier intervention. Date elements include sex, gestational age, problems with eating, birth complications or defects, growth and neurological conditions, household education and income, and ages at which the child first sat up, walked, and fed themselves.

Researchers apply AI to traditional Chinese medicine practices of identifying medical conditions by tongue color with 98% accuracy in a small study. Practitioners believe that the color, shape, and thickness of the tongue can reveal health conditions such as diabetes, cancer, stroke, anemia, COVID-19, asthma, and vascular and gastrointestinal issues.

Other

St. Catherine University analyzes LinkedIn job postings to determine which US cities have the highest demand for AI talent. The top five, with the number of AI jobs per 1,000 healthcare listings:

• Durham, NC (29)
• Colorado Springs, CO (28)
• Provo, UT (14)
• Ogden, UT (13)
• McAllen, TX (12)

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 8/14/24

News

Google launches the Pixel Watch 3, whose health-related features include workout biometrics, cardiac tracking, planning and guiding tools for runners, and loss-of-pulse detection that can contact emergency services. The watch costs $350 for the Bluetooth/WiFi version in 41 mm size.

Google also announces Gemini Live, which offers conversational AI for hands-free conversations and integration with apps. It will be available via the Gemini app on Android or as a tab on the Google app for IOS.

Microsoft announces enhancements to Nuance Dragon Ambient EXperience (DAX) Copilot that include the ability to generate referral letters, summaries of evidence, after-visit summaries, and encounter summaries. It also provides user coaching for areas where additional verbalizing would create more complete notes.  

Business

Medical device and equipment manufacturer Stryker will acquire Care.ai to offer customers smart hospital solutions that address nursing shortages, staff retention, and workplace safety. The technology will be integrated with Stryker’s Vocera system, which it acquired for $3 billion in January 2022.

Elise.ai, which sells AI assistants and customer relationship technology to property management companies, raises $75 million in a Series D funding round that values the company at $1 billion. Its Health AI business automates patient conversations and manages appointment scheduling and patient payment.

Amazon describes how its AI work is helping transform healthcare:

• Improving health visits via its HealthScribe ambient documentation service.
• A collaboration with EvolutionaryScale to enable researchers to design new proteins.
• More efficient prescription filling and better customer service in Amazon Pharmacy.
• Using AWS Textract intelligent document processing in Amazon Pharmacy to extract and structure information from digital and paper prescriptions, which allows order processing that is up to 90% faster.
• Partnering with health systems, insurers, and life sciences companies to uncover patient insights while ensuring privacy and security.

Research

Amazon describes the technical underpinnings of its Bedrock service that extracts unstructured data from standardized form entries, using healthcare as an example.

Other

Researchers postulate that clinicians, specifically radiologists, and AI do not make up a synergistic team, as humans rely on their knowledge and environment but AI learns from its own correlations and is not limited by context. They say that AI development is outpacing the understanding of its clinical value and the challenges that are involved in its integration.

Tests of open-source, locally hosted LLMs such as Meta Llama 3 matched the performance of ChatGPT and Claude in answering radiology board exam questions, raising the possibility of healthcare use without the expense and privacy concerns of hosted LLMs.

The minister of health of the Netherlands believes that AI can help solve staff shortages, as tightened immigration laws rule out bringing in workers from other countries.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 8/7/24

News

CHIME publishes a report of 10 principles, with detailed recommendations for each, that will ensure the ethical and effective implementation of AI in healthcare.

Amazon’s One Medical team is reportedly considering developing an AI agent called DoctorAI that would optimize clinician tasks, make product and service recommendations, and manage customer service requests.

In England, NHS trusts are using patient engagement prediction software from Deep Medical to identify patients who are likely to no-show or cancel their appointments with short notice. Volunteers reach out to the patients who are identified by the system three weeks ahead of time to see if they need assistance – such as with transport, translation, or parking fees – and to reschedule their appointments in the event of conflicts. NHS England says that patients don’t show up for 6.4% of its 125 million outpatient appoints each year at a cost of nearly $2 billion.

FDA announces the members of its newly formed Digital Health Advisory Committee, which will hold its first meeting in November to discuss total product life cycle considerations for AI-enabled devices.

Switzerland-based Sonova introduces the first hearing aid that uses real-time AI to improve speech clarify from background noise.

Business

CMS approves separate Medicare payment for Annalise.ai’s obstructive hydrocephalus portion of its critical care platform. The company says the solution, which triages non-contrast brain CT scans, is the only radiology triage device that can be billed for additional hospital payment.

Research

Researchers find that ChatGPT isn’t good at diagnosing patient conditions. It incorrectly analyzed half of a set of the 150 case challenges it was given, where it struggled with interpreting lab values and imaging results and failed to use complete information that was relevant to the diagnosis. They found that it works well as an educational tool, did a good job with differential diagnosis, and accurately suggested next diagnostic steps.

Other

In England, nine hospital trusts are using Ufonia’s Dora conversational clinical assistant to call patients three weeks after their cataract surgery and alert clinicians if they need further care. Two-thirds patients who undergo cataract surgery, which is performed 500,000 times per year, do not require an in-person appointment. The company is a spin-off of Oxford Unversity.

A pain patient’s op-ed piece expresses skepticism about AI in healthcare:

• Hospitals could have already solved the same patient problems if they really wanted to.
• The real motivation for health systems to use AI is to make more money rather than to lower patient bills.
• AI-drafted responses and emails to patients are easy to spot, and many of the initial patient emails were generated because of inefficient policies, such as requiring patients who have incurable conditions to contact their doctor every month to have their pain medication refill approved.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 7/31/24

News

OpenAI launches user testing of SearchGPT, which returns attributed web links from trusted publishers and can answer follow-up questions.

OpenAI begins rollout of Advanced Voice Mode for ChatGPT, which can conduct natural conversations, allows the user to interrupt, and tailors its responses based on speech cues about the user’s emotions.

Business

Bloomberg reviews Google’s healthcare projects – which include high-profile failures – and its work with HCA to use AI for nurse handoffs and ED physician documentation.

Cleveland Clinic hires Ben Shahshahani, PhD (SiriusXM) as its first chief AI officer.

Writer Inc. launches Palmyra-Med-70b, a large language model that the company’s benchmarks indicate is the most accurate available. Palmyra-Med named Vizient, CirrusMD, and Medisolv as Palmyra-Med users.

Research

Study participants trust medical advice less if AI was involved in its creation — even if they know that humans have reviewed the AI draft — and are less likely to follow any advice in which AI played a part.

Researchers find that AI can predict chronic pain in breast cancer patients, which could support early identification and personalized management.

Other

Doctors at University of Vermont Health Network are saving three hours per day in using Abridge for ambient documentation.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 7/24/24

News

Mayo Clinic will implement an ambient nursing documentation system from Abridge that integrates with Epic’s nursing workflows. Mayo, which is participating in the technology’s development, hopes to have nurses using the tool by the end of the year.

OpenAI creates compliance and administrative tools for enterprise users of ChatGPT who work in heavily regulated industries such as healthcare, legal services, and finance. The API sends time-stamped user records to EDiscovery and Data Loss Prevention service providers to support archiving, audit trails, data redaction and retention, and policy enforcement. A key use case is HIPAA compliance related to detecting and deleting protected health information. The company will also roll out a system this week that syncs employee directories with ChatGPT to provision and de-provision user accounts.

Sound Health announces Sonu, a $299, FDA-authorized, AI-powered wearable that analyzes the user’s sinuses and delivers acoustic waves that stimulate nasal vasoconstriction to treat rhinitis congestion.

Business

Drug design technology vendor Insilico Medicine announces DORA, an AI-powered system for researchers that summarizes relevant published literature and then uses the researcher’s hypothesis to create a draft article with citations.

Drug maker Takeda will use AI models from Nference to identify patients who could benefit from advanced interventions for inflammatory bowel disease.

Microsoft will collaborate with Mass General Brigham and UW Health to develop Azure-powered medical imaging models that will extend Nuance’s imaging copilot applications.

B. Well Connected Health adds configurable AI architecture to its FHIR-based platform to support network access, health data summaries and queries, and a consumer cost-saving recommendations engine.

Meta announces the open source Llama 3.1, which it says is being used for drug discovery, personalized medicine, clinical trials administration, and creating voice and vision ambient medical documentation.

Research

Researchers raise concerns about using AI in medical publishing:

• Real-world validity and clinical relevance needs to be improved and measured.
• The “publish or perish” model may tempt researchers into writing low-quality and misleading articles that are already common in some environments, countries, and settings.
• AI-created programming might increase errors when used by less-expert people.
• AI can generate plausible but incorrect medical articles.
• Future scientists may be hampered by homogenized articles, an increased number of published papers, and validity issues.

Other

HHS creates new positions for CTO, chief artificial intelligence officer, and chief data officer.

Hospitalist and AI-powered oxygen therapy platform company CEO Julio La Torre, MD, MBA says that output from always-on vital signs monitors is an untapped resource. He says that monitoring could make rounding and intermittent data capture obsolete and provide value to researchers. He notes challenges – data collection and standards, the need for robust security, developing the hospital infrastructure to integrate monitor data with existing systems, and developing a sustainable business model.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 7/17/24

News

Huma Therapeutics Limited announces Huma Cloud Platform, a no-code system for configuring regulated disease management tools that includes pre-built modules, device connectivity, cloud hosting, APIs, and a marketplace. The company built the system for its own products and will offer it as a software development kit. The London-based company, which has raised $250 million in funding, was founded by Dan Vahdat, who left his Johns Hopkins bioengineering PhD program in 2011 to start the company.

The VA will issue sole source bids to Abridge and Nuance to conduct ambient scribe pilot projects to transcribe clinical encounters and generate chart notes. The companies recently won a VA tech sprint for those functions. The VA will solicit feedback from other companies that believe they can meet its requirements.

The University of Alabama at Birmingham will launch a master’s in AI degree program next year that will teach students how to ethically integrate AI into healthcare software. It seeks applicants with a background in medicine, statistics, computer science, math, or biomedical engineering, also expressing a preference for healthcare professionals who have a programming background. Proficiency in the Python programming language is also required.

BBC reports that hospitals in England will use AI to improve patient flow, prepare radiology reports, and support rapid ED assessment. Patient records will be reviewed each morning to make sure that treatment is on track and that planned discharge dates are appropriate.

Family medicine physicians at University of Kansas Medical Center find that ChatGPT version 3.5 can produce summaries of peer-reviewed journal articles that are 70% shorter than the abstracts as posted, but with high quality, accuracy, and lack of bias. They documented ChatGPT hallucinations in four of 140 summaries. The tool fell short in being able to determine if a given article is relevant to primary care.

Business

Israel-based AI-powered disease modeling company CytoReason raises $80 million in a private funding round, with investors that include Nvidia and drug maker Pfizer. The company will open an office in Cambridge, MA later this year.

A law journal article says that the US patient system is not prepared to protect the AI-based technology innovation of companies. Companies struggle to disclose enough information about how their AI system works to earn an enforceable patent, especially if they cannot disclose the training data that was used. The authors also note that medical imaging analysis models are often built via trial and error, making those methods just as important as the training data.  

Research

Healthcare experts who were convened by Stanford University’s Institute for Human-Centered AI say that HIPAA as well as the FDA’s regulatory authority are too outdated to apply to rapid AI development. Specific recommendations:

• Streamline FDA market approval for AI-enhanced diagnostic capability, moving the emphasis to post-market surveillance; sharing test and performance data with providers to help them assess product safety; and creating finer-grained risk categories for medical devices.
• Participants were divided on the issue of requiring clinical AI tools to place a human in the loop, with some warning that such a requirement would create more busywork for doctors and make them feel less clinically empowered. Some said the model should be similar to that of laboratory testing, where devices are overseen by physicians, undergo regular quality checks, and send out-of-range values to humans for review. Participants were also mixed on requiring that patients be informed when AI is used in any stage of their treatment, although many felt that AI-created emails that are sent under a provider’s name should indicate that AI played a role.
• About half of participants said that chatbot-type AI tools should be regulated using medical professional licensure as a model, while a nearly equal number favored a medical device-like approach.

Pharmacy residents in the Netherlands test ChatGPT’s ability to respond to the clinical pharmacy questions of practitioners and patients. They conclude that it should not be used by hospital pharmacists due to poor reproducibility and a significant portion of answers that were incomplete or partly or completely wrong.

Other

A Phillips survey of 3,000 healthcare leaders in 14 countries includes questions about automation and AI. Snips:

• Nine out of 10 leaders see the potential of automation to support staff, reduce administrative tasks, and allow staff to work at their highest skill level, but two-thirds of them believe that healthcare professionals are skeptical and worry about inadequate AI and losing their skills due to overreliance on technology.
• Nearly all participants say that their organizations experience data integration challenges that hamper their ability to provide timely, high-quality care. They most often cited the time required to look up results, inefficiency, limited coordination among providers and departments, repeat tests, and risk of errors.
• More than half say that their organizations will invest in AI in the next three years, while 29% say they have already done so.
• Nine out of 10 leaders expressed concern about AI data bias.

People management software vendor Lattice cancels its week-old plans to create employee records for AI bots. The CEO had said that digital workers would, like their human counterparts, be assigned onboarding tasks, goals, performance metrics, IT system access, and a manager.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.

Healthcare AI News 7/10/24

News

OpenAI co-founder and CEO Sam Altman and Thrive Global founder and CEO Arianna Huffington announce that their companies will co-develop an AI-powered, hyper-personalized app for behavior change, which they say can be a “miracle drug” for preventing disease and optimizing health. Thrive AI Health will learn preferences and patterns related to sleep, food, movement, stress, and social connection and will be trained on the user’s biometric, lab, and other medical data. Some of the new company’s concepts are already being used by Thrive Global, which sells its product to employers.

Health wearables vendor Oura releases the AI-powered Oura Advisor, which it calls a personal wellness coach, for testing with its devices. Users can choose the chatbot’s communication style, notification preferences, and the training goals that it emphasizes.

A radiation oncologist creates prior authorization requests by telling ChatGPT the type of letter that he needs and the types of clinical studies that he could cite to support his request. He then tells ChatGPT to make the resulting letter four times longer because “if you’re going to put all kinds of barriers up for my patients, them when I fire back, I’m going to make it very time consuming.” A rehab medicine physician uses Doximity GPT, a HIPAA-compliant version of the chatbot, to analyze EHR data and coverage details to create a detailed request, which is says has boosted his coverage approval rate from 10% to 90%. The article notes that some Epic users are testing its new AI-powered PA functionality.

Business

South Korea-based AIRS Medical, whose FDA-cleared SwiftMR speeds up MRI scanning time by up to 50%, raises $20 million in a Series C funding round.

Amazon Web Services uses a large language model to design marketing content for disease awareness that it says balances accuracy and audience engagement. The company says that the AI assistant reduces creation time from weeks to hours and gives subject matter experts more control with its ability to create automatic revisions based on user instructions and comments. It also highlights any problems with required rules or regulations.

Research

Researchers survey oncologists about their possible use of AI, with these conclusions:

• They will use AI to make clinical decisions only if they as frontline practitioners can understand its logic.
• Most oncologists think that they should protect patients from biased AI tools, but few of them believe that they can recognize that bias.
• More than 90% of oncologists believe that AI developers should be held accountable if their products are involved in medical or legal problems, about double the number who believe that doctors and hospitals should share accountability.
• A medical oncologist predicts that AI tools will be used in three areas: making treatment decisions using a broader set of data, improving tumor characterization from images, and matching patients with clinical trials.

Boston researchers apply an AI model to demographic data, health history, exam results, neurological tests, and MRI scans to identify 10 distinct causes of dementia to support differential diagnosis.

Scientists at Dana-Farber Cancer Center and Weill Cornell Medicine develop digital pathology AI tools that can answer pathologist questions and compile results. They trained a private, secure Dana-Farber version of ChatGPT on 10,000 pages of recent digital pathology developments, which allows researchers to create detailed, summarized results quickly. They also developed a tool to allow pathologists who don’t program in the Python programming language to use the PathML computational pathology analysis tool.

Other

A philanthropy magazine profiles the $20 million donation to Mayo Clinic for AI work by Dwight Diercks, an engineering SVP and employee #22 at Nvidia, who grew up on a family farm in rural Minnesota. The focus of his donation is early cancer diagnosis.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Contact us.