| | | | |  | | By Ruth Reader, Erin Schumaker and Daniel Payne | | | | | 
Google tested its own bot and found it was kind. | Getty | Google’s AI could have a better bedside manner than your doctor. An experimental chatbot from the tech giant called Articulate Medical Intelligence Explorer outperformed doctors in an evaluation conducted by Google researchers of the bot’s politeness, ability to explain a diagnosis and compassion. The interesting bit: Google researchers had difficulty finding conversations between doctors and patients to use to train the AI. So instead, they used medical records and transcriptions of medical conversations. They then programmed the AI to learn by directing it to converse with itself, playing the roles of both doctor and patient. Additionally, the researchers devised a method that would give them feedback on how to improve by asking the AI to play yet another role: a critic overseeing the conversation. The researchers tested the AI using actors posing as patients. They compared the results to that of primary care physicians in a randomized, double-blind crossover study of text-based consultations, also with actors playing patients. Even so: Doctors consulting with patients by text may not offer the empathy they would in person. And it’s important to note that Google’s self-evaluation is a preprint study that hasn’t yet undergone the peer-review process required before a study’s publication to ensure its validity. What’s next? Evaluating the chatbot to ensure it gives equivalent care to diverse groups. Then, researchers want to figure out how to ethically test the AI on human patients.
| | | | JOIN 1/31 FOR A TALK ON THE RACE TO SOLVE ALZHEIMER’S: Breakthrough drugs and treatments are giving new hope for slowing neurodegenerative diseases like Alzheimer’s disease and ALS. But if that progress slows, the societal and economic cost to the U.S. could be high. Join POLITICO, alongside lawmakers, official and experts, on Jan. 31 to discuss a path forward for better collaboration among health systems, industry and government. REGISTER HERE. | | | | | | | | | 
Boulder, Colo. | Shawn Zeller/POLITICO | This is where we explore the ideas and innovators shaping health care. A pig liver cleaned out toxins in a brain-dead man’s body at the University of Pennsylvania last month. The three-day experiment, in which the man was attached to an external pumping device containing the pig liver, could open doors to treating people suffering liver failure, MIT Technology Review reports. Share any thoughts, news, tips and feedback with Carmen Paun at cpaun@politico.com, Daniel Payne at dpayne@politico.com, Ruth Reader at rreader@politico.com or Erin Schumaker at eschumaker@politico.com. Send tips securely through SecureDrop, Signal, Telegram or WhatsApp. | | |  A new AI tool could prove useful to cancer surgeons. | ANNE-CHRISTINE POUJOULAT/AFP via Getty Images | An artificial intelligence tool developed by Penn Medicine researchers could one day help doctors catch cancer that might now go undetected. How so? iStar, short for Inferring Super-Resolution Tissue Architecture, interprets medical images broadly and at the individual cell level — making it helpful in determining whether cancer surgery was successful. To develop iStar, researchers used a hierarchical vision transformer, a machine-learning tool, which they trained on standard tissue images. “The power of iStar stems from its advanced techniques, which mirror, in reverse, how a pathologist would study a tissue sample,” Mingyao Li, a professor of biostatistics and digital pathology at Perelman School of Medicine at the University of Pennsylvania, said in a news release. “Just as a pathologist identifies broader regions and then zooms in on detailed cellular structures, iStar can capture the overarching tissue structures and also focus on the minutiae in a tissue image.” Li and her collaborators tested iStar on breast, prostate, kidney and colorectal cancer tissues, as well as healthy tissue. They found that iStar could detect difficult-to-see cancer and tumor cells. It was also much faster than competing products, completing an analysis of breast cancer tissue in just nine minutes. The National Institutes of Health-funded research was detailed in the journal Nature Biotechnology earlier this month. What’s next? In addition to helping doctors detect cancer, the researchers want to use iStar to analyze larger quantities of data, such as biobanks, which store thousands or millions of biological samples.
| | | | Enter the “room where it happens”, where global power players shape policy and politics, with Power Play. POLITICO’s brand-new podcast will host conversations with the leaders and power players shaping the biggest ideas and driving the global conversations, moderated by award-winning journalist Anne McElvoy. Sign up today to be notified of new episodes – click here. | | | | | | | | |  UCLA researchers helped put together an AI-friendly database of surgeries. | AFP via Getty Images | A database of more than 83,000 surgical outcomes could be a goldmine for researchers developing artificial intelligence tools. And now there is one: The Medical Informatics Operating Room Vitals and Events Repository, or MOVER, was assembled by University of California researchers and includes information about patients' surgical procedures, medicines taken, lines or drains used, and postoperative complications. "We expect it to help the research community to develop new algorithms, new predictive tools, to improve the care of surgical patients," Dr. Maxime Cannesson, professor and chair of anesthesiology and perioperative medicine at the David Geffen School of Medicine at UCLA said in a statement. Why it matters: There aren't currently many repositories of publicly accessible clinical data researchers can use to train and test AI algorithms, especially relating to surgery, the UCLA-UC Irvine team noted in a paper published in the journal JAMIA Open. MOVER, which is backed by the National Institutes of Health, is free to use as long as researchers sign a data use agreement. Privacy first: While the project is focused on information sharing, the researchers said they prioritized privacy. All data complies with patient privacy laws and is stripped of identifying information, like dates of surgery and ages for patients older than 90 years old who could be easier to identify than younger people. It's also designed to be transparent, with data that can be checked. "The goal is eventually to increase the trust that clinicians and patients have with what you are going to see in the near future — the development of more and more artificial intelligence-based models, especially for the surgical setting," Cannesson said. | | | | Follow us on Twitter | | | | Follow us | | | | |
