Google introduces new AI model for healthcare Here's how doctors are using it

Google on recently announced a series of new healthcare AI models called MedLM.

KEY POINTS

T.....The move marks Google's latest attempt to monetize its AI tools for the healthcare industry as competition for market share remains fierce among rivals such as Amazon and Microsoft.

 

Photo: Sundar Pichai, CEO of Google and Alphabet

Google said it plans to bring a health-specific version of Gemini, its latest and "most powerful" AI model, to MedLM in the future.

 

Google on Wednesday announced a new set of health-specific artificial intelligence models designed to help clinicians and researchers conduct complex studies and summarize doctor-patient interactions. A MedLM was announced.

 

The move marks Google's latest attempt to monetize its AI tools for the healthcare industry, as competition for market share remains fierce with rivals such as Amazon and Microsoft

 

It is something. CNBC spoke with companies that have tested Google's technology, including HCA Healthcare, and experts say that while they are taking cautious steps to implement it, the potential for impact is real.

 

The MedLM suite includes large and medium-sized AI models, both based on Med-PaLM 2. Med-PaLM 2 is a large-scale language model trained on medical data that Google first announced in his March. It will be generally available to eligible Google Cloud customers in the US starting Wednesday, and Google said the cost of the AI ​​suite will vary depending on how companies use the various models, but that the mid-range model is operationally cost-effective. Ta.

 

Google said it plans to bring a healthcare-specific version of its latest and "most powerful" AI model, Gemini, to MedLM in the future.

 

Aashima Gupta, global director of healthcare strategy and solutions at Google Cloud, said the company has discovered that various medically tuned AI models can perform certain tasks better than others. I did. This is why Google decided to introduce a variety of models rather than trying to create a "one size fits all" solution.

 

Google, for example, uses its large-scale MedLM model to perform complex tasks that require deep knowledge and large amounts of computational power, such as conducting research using data from a healthcare institution's entire patient population. It is said that it is suitable for But if a company needs a more agile model that can be optimized for specific or real-time functionality, such as summarizing doctor-patient interactions, a medium-sized model should work better. says Mr.Gupta.

 

Real-world examples

Photo Courtesy: Getty Images

When Google announced his Med-PaLM 2 in March, the company initially aimed to answer questions such as "What are the early warning signs of pneumonia?" He said it could be used for. ``Does it cure incontinence?'' ``Does it cure incontinence?'' But as the company tested the technology with customers, the use case changed, said Greg Corrado, head of health AI at Google.

 

Collard said Google doesn't see much demand from customers for these features because doctors don't often need help with "accessible" questions about the nature of a disease. . Rather, healthcare organizations want to use AI to solve other back-office and logistics problems, such as managing paperwork.

 

"They want something that solves a real pain point or slowdown in their workflow that only they know about," Collard told CNBC.

 

For example, HCA Healthcare, one of the largest health systems in the United States, has been testing Google's AI technology since the spring. In August, the company announced an official partnership with Google Cloud aimed at using generative AI to "improve workflows for time-consuming tasks."

 

Dr. Michael Schlosser, HCA's senior vice president of care transformation and innovation, said the company uses his MedLM to enable emergency physicians to automatically record patient interactions. For example, HCA uses an ambient audio recording system from a company called Augmedix to transcribe conversations between doctors and patients. Google's MedLM suite can split these transcripts into components of emergency room provider notes.

 

Schlosser said HCA uses MedLM in the emergency rooms of four hospitals, and the company plans to expand its use next year. Schlosser added that by January, he expects Google's technology to be able to successfully generate more than half of the notes without vendor assistance. Since a doctor can spend up to four hours per day on administrative tasks, the time and effort savings make a big difference, Schlosser says.

 

"This was a huge step forward for us," Schlosser told CNBC. "We think we'll get to the point where AI alone can create more than 60% of notes accurately, before a human reviews or edits them."

Schlosser said HCA is developing a handoff tool for nurses. He said he is also working on MedLM. The tool reads electronic medical records and identifies relevant information that nurses can pass on to the next shift.

 

Schlosser said automating the process would be "powerful" because handoffs are "boring" and a serious problem for nursing staff. Nurses at HCA hospitals perform approximately 400,000 handoffs each week, and two HCA hospitals tested the nurse handoff tool. Schlosser said nurses will compare traditional handoffs and AI-generated handoffs side-by-side and provide feedback.

 

Photo Courtesy: Getty Images

However, in both use cases, HCA determined that he did not find MedLM to be fool proof.

 

Schlosser said the fact that AI models can spit out false information is a major challenge, and HCA said he has worked with Google to develop best practices to minimize such inventions. said. He added that token restrictions that limit the amount of data that can be fed into models and managing AI over time posed additional challenges for HCA.

 

"What I can tell you right now is that the hype surrounding the current use of these AI models in healthcare exceeds the reality," Schlosser said. "Everyone is struggling with this problem, which is why no one is really implementing these models across the health care system."

Still, Schlosser said, providers' Initial response was positive and they said they recognized it still needed to be introduced. It will not work on the finished product. He said HCA is working hard to implement this technology in a responsible manner that does not put patients at risk.

 

"We're very cautious about how we approach these AI models," he said. “We do not use use cases where the results of the model could have any impact on human diagnosis or treatment.”

Google also plans to introduce health- care-specific performances of Gemini to MedLM in the future. Its shares popped 5 after Gemini’s launch before this month, but Google faced scrutiny over its demonstration videotape, which wasn't conducted in real time, the company verified to Bloomberg.

 

In a statement, Google told CNBC “ The videotape is an elucidative definition of the possibilities of interacting with Gemini, grounded on real multimodal prompts and labors from testing. We look forward to seeing what people produce when access to Gemini Pro opens on December 13.”

 

Corrado and Gupta of Google said Gemini is still in early stages, and it needs to be tested and estimated with guests in controlled health- care settings before the model rolls out through MedLM more astronomically.

 

“We ’ve been testing Med- PaLM 2 with our guests for months, and now we ’re comfortable taking that as part of MedLM,” Gupta said. “Gemini will follow the same thing."

 

Schlosser said HCA is “ veritably agitated ” about Gemini, and the company is formerly working out plans to test the technology, “ We suppose that may give us an fresh position of performance when we get that, ” he said.

 

Another company that has been using MedLM is BenchSci, which aims to use AI to break problems in medicine discovery. Google is an investor in BenchSci, and the company has been testing its MedLM technology for a many months.

 

Liran Belenzon, BenchSci’sco-founder and CEO, said the company has intermingled MedLM’s AI with BenchSci’s own technology to help scientists identify biomarkers, which are crucial to understanding how a complaint progresses and how it can be cured.

 

Belenzon said the company spent a lot of time testing and validating the model, including furnishing Google with feedback about necessary advancements. Now, Belenzon said BenchSci is in the process of bringing the technology to request more astronomically.

 

″(MedLM) does n’t work out of the box, but it helps accelerate your specific sweats,” he told CNBC in an interview.

 

Corrado said exploration around MedLM is ongoing, and he thinks Google Cloud’s health- care guests will be suitable to tune models for multiple different use cases within an association. He added that Google will continue to develop sphere-specific models that are “lower, cheaper, briskly, better.”

 

Like BenchSci, Deloitte tested MedLM “over and over” before planting the technology to health- care guests, saidDr. Kulleni Gebreyes, Deloitte’sU.S. life lores and health- care consulting leader.

 

Deloitte is using Google’s technology to help health systems and health plans answer members ’ questions about penetratingcare. However, for case, they can use MedLM to look for providers grounded on gender, If a case needs a colonoscopy.

 

Gebreyes said guests have set up that MedLM is accurate and effective, but, like other models, the AI isn't always great at decoding a stoner’s intent. It can be a challenge if cases do n’t know the right word or spelling for colonoscopy, or use other colloquial terms, she said.

 

“Eventually, this doesn't substitute a opinion from a trained professional,” Gebreyes told CNBC. “It brings moxie closer and makes it more accessible.”