How AI is changing the way doctors treat their patients | 7.30

0:00 From cancer breakthroughs, >> this is the first time that we've been able to see the complexity of cancer >> to helping GPS save time and see more patients. >> It just streamlines the work. It's good for the patient because I'm concentrating on them. >> Artificial intelligence is giving researchers and doctors the answers to problems they faced for years. But the same technology is posing complications. The greatest concerns are medical misinformation and protecting patient privacy. >> I think it's a huge challenge and I think it's a growing challenge every single day. >> This is an ordinary doctor's consulting room, but what you can't see is the AI

0:47 scribe listening to every word. >> So that records our consultation and types up all my notes for me. >> Dr. Grant Blashki has been a GP for more than 20 years in Melbourne. He's showing me how he uses AI in a mock appointment. >> You can tell me what's brought you in today. >> I've been having a headache uh on and off pretty much every morning for the last month. >> Oh, that's no good at all. Tell me a bit more about it. >> I wake up sort of on both sides of my head. It's worse in the morning and sort of you know peters off gets a bit better during the day. >> Have you had this problem before? Do you get migraines? Anything like that? >> Not really. >> Then in what the AI scribe calls making

1:33 magic, it delivers its notes. The AI scribe has offered differential diagnosis to the symptoms I described. Attention type headache and a cervoggenic headache which means pain originating in the neck. Dr. Blashki has been using this Heidi health AI for a year. >> The doctor really needs to turn their mind to it and look at them more as suggestions than the answer. And I guess with the new generation of doctors coming up who will be living with AI, they need to understand its benefits and its limits. >> Clinicians, say goodbye to painful note takingaking and admin. >> And hello to Heidi. 7:30 requested an interview with Heidi Health. It declined, but in a statement, CEO and

2:21 co-founder Dr. Thomas Kelly said, "We summarize the clinical encounter reflecting their lines of questioning and using appropriate clinical terminology to describe them. Heidi does not provide a differential diagnosis, absent the clinician, and it is still up to the clinician to review their documentation for accuracy. Then there's the issue of patient privacy. Kai Vanlashout is the CEO of Lieird Health, another AICribe company. He says the software automatically deletes notes after 7 days. >> We've had doctors that have needed something that we've had or wanted it or maybe you get, you know, in the thousands of people, they don't realize that it's deleted after 7 days and there's nothing we can do.

3:08 He says all the patient data stored on his software is encrypted. >> It's basically the same level as what would be used kind of with highly sensitive bank data, right? And if someone was to see or access some of that encrypted information, it's completely iniccernible. >> In a statement, Heidi Health told 7:30 its software is offered in Australia, New Zealand, Canada, US, and the UK. In each region, data is stored in compliance with the healthcare regulations and privacy policies of the region. >> There's always some element of risk when you're dealing with digital data, and you know, you can't really see exactly where it goes. >> John Lawler researches different models of AI from bots to online tools. He says their effectiveness relies on storing a

3:54 lot of information. >> A lot of those models, they're very data driven, so the more data they have, usually the better they get. So uh on the one hand if it has a lot more data from patients they can typically improve the models. >> He says AI models are now so advanced that on social media it's hard for anyone to distinguish content made by bots from humans. >> The bot can reply to posts, make new posts under very strict conditions when it sees a certain post that has a keyword or a post by a certain individual. Um and then it becomes much more automatic and automated. AI bots posting about vaccines and spreading misinformation on social media sites is on the rise. It's currently the focus of researchers at the Murdoch Children's

4:40 Research Institute who uncovered numerous examples, including this thread on Reddit. Why the world needs fewer vaccines. One comment says, "The US has some of the highest vaccine rates in the developed world, but also the highest vaccine injury rates." Another says, "I got vaccinated, went to the doctors, and was in pain. All I could do was cry on the couch. I was so scared." And there's a question. What's the deal with multi-dosese vaccines? Are they just a marketing gimmick? Every single comment in this conversation was created by AI. Reddit declined an interview with 7:30, but in a statement said the thread is an openly labeled bot project and that

5:27 labeled bots have long been allowed on Reddit and in some cases serve as a unique part of the Reddit experience or utility. Inauthentic content that is presented to mislead, including deceptive bots, is against our policies. Vaccine hesitency's been uh on the rise to a certain degree uh and and vaccine uptake has dropped by a couple of percentage points. >> Good morning everyone. >> Brett Sutton is the former chief health officer of Victoria. He says vaccine misinformation online is having a realworld impact. We've seen what's happened in the US and it's dropped precipitously and um and that's playing out in uh in measles outbreaks and uh you know the the potential harms that

6:13 might arise more broadly around the world um uh I think are significant. >> He now works at the CSRO and says social media companies should be actively identifying and deleting misleading medical information. >> It's not just a threat to science. It's a threat to democracy, to national security. It's a threat to trust in national institutions and to government. [Music] Moving offline and into this Sydney lab, where AI is being used to tackle one of the world's worst diseases. >> Hey Shan, how you going? >> Hey Christine, are you good? An international team of researchers co-led by associate professor Christine

7:00 Schaffer are using an AI tool known as a net. It enables us to understand the structure of a tumor. And by structure, I really mean how uh similar is one cancer cell to the next. We're ready to send you the data for you guys to start analyzing. Christine collects tumor samples in Sydney then sends them to associate professor Smitta Krishna Swami at Yale University in the United States who processes the samples through A8. >> Hey Christine um we have some preliminary analyses and it looks really exciting. Would would you like to see it? >> Yes. Cool. What we do with that information then is to work out ways to eradicate some of the key components of each groups of cells.

7:47 This color-coded image of a tumor sample created by AAET shows the different cells within a tumor. >> The orange parts of those tumors are the cells that we know sit with a specific grouping. So those cells we know are proliferating uh quite quickly and are really contributing to the bulk of the tumor. But on the inside of that tumor where the blue cells are, this is where we've found an hypoxic niche. And so this is where cancer cells are hiding. Thanks to this AI tool, researchers hope it may be possible to eliminate all cancer cells the first time a patient is diagnosed. We want to stop those cells from returning and coming back. So, we're really focused on coming up with ways to

8:33 eradicate every cell within a primary tumor so that there aren't any left behind. >> It's discoveries like this that make Dr. Grant Blashki excited about the future of AI. I think the health system is going to have a better exchange of information, much better use of resources and ultimately better outcomes for the patient.