Dr. David Jones, Director of the Mayo Clinic’s Neurology Artificial Intelligence Program, recalls a patient who was told there was no treatment for his dementia, predominantly because a previous neurologist could not find a diagnosis. “That’s not an acceptable answer,” he says. Instead, his team developed a program called StateViewer, powered in part by artificial intelligence, which searches brain scans of patients looking for patterns that can help rapidly — and accurately — diagnose 14 different types of dementia in clinical settings.
As Dr. Jones knows, not every patient can be seen or treated by the Mayo Clinic. However, a tool like StateViewer increases the potential for clinicians to diagnose — and treat — more patients.
“We want to democratize this stuff,” he says. “There are millions of people who need access to this.”
You can hear more from Dr. Jones and his team’s critical work during his talk in the DOC 2025 session, “Neurodegenerative Diseases: State of the Science and What You Can Do Now,” in our video, or read our lightly edited transcript below.
TRANSCRIPT:
Dr. David Jones
Thank you. It’s a pleasure to be here. The word dementia in, if you translate that into the original Latin is being without a mind. Now, I don’t want to lose my mind. You don’t want to lose your mind. It’s horrendous to watch our loved ones lose a piece of their minds. This is what we are missing in the current field of Alzheimer’s disease. We don’t. The mechanisms of the mind are what we’re missing. And how do they relate to the disease? That’s what we need to sort out. And that means, I believe that if we solve dementia, we will solve intelligence. And if we solve intelligence, it will help us solve dementia. These are synergistic problems. And when we overlap them, we can.
Where you’re stuck in one field, you can go to the other. And you can synergistically creatively go back and forth. And we can finally start to make progress both on intelligence and on dementia. So one, every three seconds someone in the world develops dementia. There’s obviously a horrendous financial cost, but, there’s this is devastating for human beings.
And by 20, 50, 130 million human beings, and we don’t want, the individuals lost in this, every single patient dementia happens to them in a unique way. For them, an average is really don’t count. There’s many different quote unquote causes. So we need to get to the bottom of this at an individual level.
When a patient comes to see me, I think about them, not the 139,000,000 in 2050. The problem is we can’t measure the mind. We’re terrible at it. We don’t understand it. How do we even begin to approach this? Cardiology is able to make advances with a lot more numbers that relate directly to the physiology that we care about, which is the heart needs to pump.
The brain needs to think, how do we measure this? We just have metaphors. We talk about proteins. Proteins don’t think so how do they relate to us? Thinking and having intelligence. So the breakthrough that we are working on is we look at brain energy. Those mitochondria need fuel. They need glucose. There’s a scan that’s been around for about almost 50 years. FDG, Pet, widely available, just isn’t used. But we’re understanding it’s telling us it’s a good biomarker of mental function in all neurodegenerative diseases. And so we are able to simplify the information. The FDG, Pet scan color, code it into a small number of dimensions. Now, each one of these dots you see here is a patient and it’s color coded by, which areas of their brain are not working.
If they’re close together, they’re similar to one another. And so you see these different dementia syndromes clustering like Lewy body disease post your cortical atrophy your vision is gone. Typical Alzheimer’s disease, semantic dementia. You no longer have semantic information limbic predominant or late disease taking away memories in late life. Behavior problems. Behavioral frontotemporal dementia. So all of the mind is laid out there in that heterogeneity.
We put this into a diagnostic tool, that, really understands this functional organization of the brain. And so when a patient comes in, this dot in the middle represents the patient. This is the patient’s brain. Blue is where it’s not working. We find matches, matches I’ve seen at Mayo Clinic in the past. And I know what happened to those patients.
I kind of call that brain like mine. It’s like, yes, I’ve seen about 20 patients have a brain that looks just like yours. What happened to them? What was their diagnosis? How did we treat them? Those areas in blue are, the problem. We like to explore these things in detail, really looking closely and make sure, these things are lining up with their clinical symptoms, the way they’re losing their intelligence.
Here’s some matches, as I get further away, these are other patients we’ve seen that’s informing the advice I’m going to be giving you. And so here is a different plot, same sort of thing. Each dot is a patient, and how close together they are, how similar their brains are. And you can see if you could describe the whole field of neuro behavioral neurology in one slide.
This kind of does it because you get visual symptoms. Maybe you get executive symptoms. Maybe you get language symptoms. Some people can’t speak behavior symptoms, semantic knowledge memory. Some people fall. They lose the ability to manipulate their mouth. Lewy body disease. You have hallucinations. They’re all here on one slide. And then the proteins associated with those conditions are there too but this is what we’re use to, to, well, this I didn’t project right.
But, what we did was I looked at 15,000 scans at Mayo Clinic and ran the tool. And what we found was, you know, it’s not all Alzheimer’s disease. There’s different forms of Alzheimer’s disease. We found 11% had this condition called normal pressure hydrocephalus. What is that? That causes dementia symptoms, and it’s 100% treatable. So why would you want that and not know it?
We recently had to do a randomized, placebo controlled trial published in New England Journal to prove doctors should be doing this. Just published last month. This condition has been around for 75 years. Why do we have to do this? Trial people are nihilistic about getting the diagnosis right. These patients shouldn’t be missed. They need to be treated.
Here’s just a clear kind of patient example. My new patient has given us, permission to share his story. But it took him several years going to see neurologist after neurologist across the country to try and get a diagnosis. Every doctor he saw. I know you have dementia. It’s one of these. I don’t know what it is, but I can tell you there’s no treatment.
That’s not an acceptable answer. If somebody tells you there’s no treatment, you want to know what’s the diagnosis? That’s how you know there’s no treatment. You don’t even know the diagnosis. How do you know there’s no treatment? He came to see us and we could say, well, here’s all these different possible ways you could get dementia. Which one did it to you? He landed right there. That’s the non degenerative cause and that’s surgically reversible. So we did surgery that week. And now he says I can enjoy time with my family again. I can go out with my friends. I can even go down do my own taxes. These are moments that I thought I’d last forever. So how many of cases like this are acceptable to miss?
How many have to exist before we say we should catch all of them? Let’s say one. We need to catch all of them. And so this tool we’re using now, it’s been, rolled out enterprise wide at Mayo Clinic for about four months on this investigational use. And it’s very quick, wide adoption. And people have logged on to this application about 2000 times, and there’s about 70 plus users and growing, it increases diagnostic accuracy.
Reading these brain scans by 3 to 5 times reduces reading time by 50%. Has a lot of integration with our EMR, and it’s built on Google Cloud architecture. And, there’s these static reports and there’s these dynamic reports. Hey, this is new technology. It’s helping us right now, today in clinical practice. And we’re just getting started. And so just as an example, this is a study, where we, we gave the radiologist instead of 14 different types of dementia.
We said three picked three different categories. It’s either post your cortical atrophy, Lewy body disease or something else. This is all autopsy confirmed cases. The tool itself is about 87% accurate in this data set. The for readers about 70%. This is a trainee who’s almost at random chance. But then you give them all the tool and they all become better than expert level.
That’s what we want. We want to democratize this stuff. We want to digitize my knowledge and understanding, put it in a tool, and you don’t fly to the Mayo Clinic anymore. You get this done where you are because there’s millions of people who need access to this. And so here’s, Nathan Young. He’s a community neurologist in a community practice, and he says FDG pet with state here is a game changer in practice. And very practical for cases in the community practice, sort of knowledge sharing and expertise helps us set us apart. It’s a great for patient care. The state we are no longer need to bug David to help you read scans. It’s like having an expert work with you. In each case, I put a little error there because he does not bug me.
I love this, this is my expertise. But they don’t need me as much, which is great. It’s why we made the tool. I’m not going to go into details about this, but I think design, designing your technology within the environment that’s going to be used to help you sharpen it, make sure it’s right and iterate quickly and really make friends with your information technologies to do this right. Then you can bottle it up, maybe in like our cloud platform, give people access to it after, you know, it works. So key takeaways I think there is a new era for brain health coming. I think we should be seeking precision and dementia diagnosis. I think AI transforms neurologic expertise into scalable precision, and with the right partnerships, we can make expert brain care universal all. So thank you very much for your attention.