Return to Transcripts main page


Detecting Parkinson's Disease by Voice

Aired December 2, 2012 - 14:00   ET



ANNOUNCER (voice-over): They're innovators, game changers, people pushing themselves, moving us all forward. They're the next scientists, musicians, poets, the next makers, dreamers, teachers, and geniuses. They are THE NEXT LIST.


DR. SANJAY GUPTA, CNN ANCHOR: Max Little has a bold idea. What if doctors could detect Parkinson's disease simply by the sound of your voice? Max Little is close to proving just that. He says one simple voice test can determine if someone has Parkinson's. All you need is a telephone.


UNIDENTIFIED MALE: We've got an ultra low cost way of detecting the disease.


GUPTA: He calls it the "Parkinson's Voice Initiative." He's testing voices from around the world to fine-tune his algorithm. He's collected over 17,000 voices and counting.


GUPTA: How confident can you be that person has Parkinson's?

UNIDENTIFIED MALE: Ninety nine percent.

GUPTA: That's pretty incredible.


GUPTA: Now if he succeeds, he could change the game for Parkinson's patients and for doctors. What makes this discovery even more incredible, Max isn't a doctor. He's a mathematician. I'm Dr. Sanjay Gupta and this is THE NEXT LIST.


MAX LITTLE, APPLIED MATHEMATICIAN, PROJECT DIRECTOR, "PARKINSON'S VOICE INITIATIVE": My name is Max Little and for a living I do applied mathematics. GUPTA: How do you go from your background in mathematics to essentially coming up with a way to diagnose Parkinson's?

LITTLE: It's an interesting story because in 2006, I was working on techniques for understanding, so say somebody would have had a full surgery and you want to know has the surgery improved their vocal function or not?

It didn't occur to me at the time that neurological disorders such as Parkinson's disease also affect the voice. I mean, I think I just thought about it as everyone else does at that point, which is you just see tremor in the limbs and that is pretty much how you detect it.

And until I had a chance e encounter with a researcher from Intel Corporation in a conference in 2006. He said, well, we've been doing this trial recording voices of patients so perhaps your techniques could be useful there.

GUPTA: Wow. It was a chance encounter?

LITTLE: Totally, yes. And the reason why they were doing this is because one of the co-founders of Intel Corporation, Andy Grove, was diagnosed with Parkinson's in 2000. And he realized, as I came to realize later, that there were no objective ways of measuring the progression of the disease.

The only way you could do it was by going to a neurologist and having a full exam. And so we met up and he said, so we've got all these recordings of Parkinson's disease voices.

Do you think the techniques you're using could be used here? I had no idea. You know, I've never looked to Parkinson's disease before. I thought, let's give it a go. Let's try it out.

GUPTA: So just in terms of the process, if someone picks up the phone and calls, what are they told to do?

LITTLE: So we have set up a study group, "Parkinson's Voice Initiative" and the idea of the study is to collect enough data that we can test this technology outside the lab to collect recordings from people in all sorts of circumstances that we can't control.

It will then test our ability to accurately detect whether they have Parkinson's disease or not. In lab based studies, we get 99.9 percent detection accuracy. But can we get that accuracy over the telephone when we don't have those control circumstances?

GUPTA: Ninety percent though.

LITTLE: Yes, that's the overall accuracy. So this here, these two formulas, describe typical algorithms used in clinical voice analysis. So these kinds of algorithms are the sorts of things we use to extract useful features from your voice recording.

So in this case what will happen is that people will call in. They simply say whether or not they've been diagnosed with Parkinson's and we ask them to say aah for as long as they can.

GUPTA: That's it?


GUPTA: It sounds so remarkably simple. As a neuroscientist myself, you know, we spend years studying this sort of thing. To think that literally it involves some guy saying aah into a phone or microphone. That could tell you so much information.

LITTLE: Yes. I mean, bear in mind voices are a very, very rich information source. Think of it this way -- there are 44,000 or so samples per second captured with this, these recordings, so there's a lot of data you collect over, say, 30 seconds. So now think about how a typical health record, how small that is by comparison.

GUPTA: Can we listen to what that sounds like?

LITTLE: Sure, yes.

GUPTA: So what did you hear there?

LITTLE: That's a classic example of vocal tremor. You can hear the tremor in the way the vocal folds, the muscles are twitching somewhat. And that affects the pitch.

GUPTA: So the same sort of tremor that you have in your hands, your arms, you're saying that's basically happening in the muscles around your vocal chords?

LITTLE: That's right, yes.

GUPTA: When you look at that, how do you analyze the data?

LITTLE: Typically, what we're looking at in this case would be the consistency of the interval between each vocal cycle. So in this case if you look at a healthy voice you can see this quite stable interval in between each vocal cycle whereas if you look at someone who has Parkinson's and has vocal tremor, an example where the interval becomes quite large.

GUPTA: Would that be something that this person would have noticed themselves?

LITTLE: This example of a tremor is probably fairly obvious to most people who would listen to it and all sorts of things can affect the voice such as smoking or you might have some kind of functional trauma to your local folds, say. We don't only look for vocal tremor. We look for lots of other symptoms. We look for about 130 or so different --

GUPTA: Is that right?

LITTLE: Yes, that's right.

GUPTA: A lot of doctors say, look, in the end, you have to lay your hands on the patient. You have to see them with your own eyes. You have to do all that. That's just how doctors diagnose patients. What do you say to those doctors?




LITTLE: An example future use of this technology could be that a neurologist has a number set up, a person could call in that number. They leave a voice recording. The algorithms would analyze the voice recording and the neurologist can get an indication whether they have Parkinson's and of course, they could get back to the patient and follow up.

GUPTA: A lot of doctors say, look, in the end, you have to lay your hands on the patient. You have to see them with your own eyes. You have to do all that. That's just how doctors diagnose patients. What do you say to those doctors?

LITTLE: Well, I say, of course, do that. It's very important to do that, but don't forget what's happening on the outside of the body is only one part of what happens in disease. We take a disease like Parkinson's, you can do something about it.

They can improve the quality of life and suffering. So it's very important to extend the reach of neurologists so that they can do this kind of detection and diagnosis remotely.

DR. NICTA MEJIA, NEUROLOGIST, MASSACHUSETTS GENERAL HOSPITAL: I'm a clinician. I see patients with Parkinson's disease and disorders, I also practice did --

GUPTA: How important is it to be diagnosed early with Parkinson's?

MEJIA: If you are having symptoms that are reflecting your day-to-day activities it's important to know what's causing them so that you can be appropriately treated.

SUSAN MILLAY, PARKINSON'S PATIENT: The first symptoms came on suddenly. People noticed a significant limp in my walk. And from time to time I would notice that I would have trouble writing.

GUPTA: What did you think it was?

MILLAY: I didn't know. There was a long process of being diagnosed before I actually got to Dr. Mejia.

GUPTA: When you hear about max little's work, somebody calling, actually giving their voice as part of the diagnosis, what do you think of that as a patient?

MILLAY: I think that's amazing. I think it's an amazing technology.

LITTLE: There are whole communities who don't have access to a neurologist. There are places in the first world people take it for granted, of course, that there is a high density to doctors to patients.

That's not true across the vast majority of the world. And so in those circumstances, lots of people will actually go about their lives not realizing that these -- they are developing quite severe symptoms to this disease and their families don't know it.

MEJIA: If I tell you 30, the number is expected to be around 9 million.

GUPTA: You know Parkinson's well. What does Max and his technology offer to you?

MEJIA: One of the mind blowing things when I said let's call Max is if it helps you screen for very low cost and, you know, very easily accessible technology, it would mean a lot, I think, to many patients.

GUPTA: So if you two give presentations together or talk to people here at the university about this?

MEJIA: No, not yet.

GUPTA: We're catching you guys at the very beginning.

MEJIA: We very recently met. You can think the promise was kind of technologists can hold, but then the science that has to happen. I think one of the steps ahead is really testing this further in a controlled manner.

GUPTA: How big did you think this could get, Max?

LITTLE: I think globally. The real value is we've got an ultra low cost way of detecting the disease that's accessible to 75 percent of the world's population.

GUPTA: If you could imagine what this all could become, this initiative, what do you foresee?

LITTLE: When I foresee for this kind of technology is the opportunity to radically reduce toxic health care that we're looking at, I'm hoping what we can do is we can start to radically reduce these costs, like getting accurate information about how to allocate those resources.

GUPTA: You're talking about everywhere in the world potentially?

LITTLE: Absolutely. I just had a very intense relationship with the music and sounds since being very young.



(BEGIN VIDEOTAPE) MARIA LITTLE, WIFE OF MAX LITTLE: My name is Maria and I'm married to Max. We've been married for 17 years. Max and I met a long time ago actually 20 years ago. So we have been together a long time.

And we met through some friends, and then he was a student and he asked if he could stay with me while looking for a place to live. He came to stay on my sofa and gradually moved into my bedroom and has never left.

MAX LITTLE: Yes, I think I probably did claim I was going to stay on the couch. If she didn't want to get with me -- well, I guess we just found that we had something in common. I'm just very lucky, I guess.

MARIA LITTLE: We live in oxford permanently, but we're here in Boston for two years because Max has a fellowship at MIT. And we have two children, Zach who is 3 and Aaron who is 2 months.

MAX LITTLE: Well, I was born in Scotland and then we moved down to the south of England and I grew up really there.

MARIA LITTLE: Max's mom, when she talks about him, she says he was quite an independent child. She says that his bedroom, he had a desk, which was a very large desk, and it was always covered in some pieces of electronics and wiring and stuff he was putting together.

MAX LITTLE: I think when I wrote my first code I was 7 or 8 years old. I wrote the codes by hand because I didn't have a computer. So I learned the program before I had a computer. I don't think I was particularly interested in mathematics at that time.

I only really became interested in mathematics in my early to mid-20s. I just had a very intense relationship with music and sound since being very young. I played in a band when I was in college. I wasn't very good. I used to write music for television.

I used to write music professionally for advertising purposes. I then realized I need to get a job. And what I discovered there was a very significant niche in computer games and video programs, which allowed me to compose music. It was a bit of an epiphany in a way.

I realized that what I was doing, I was creating sound effects with games. And the techniques that you could use in order to manipulate sound could be modelled. That's when I realized that I was really fascinated by the mathematics involved, that's what got me -- set me on that path.

PAUL WICKS, R&D DIRECTOR, PATIENTS LIKE ME: Max is smart and scary smart is an overused term but he's a very, very smart guy. He sees mathematics in a way that truly is a beautiful mind.

MAX LITTLE: I was studying in night class and so I did this college degree and got nearly the highest first class result that you could. And I think this was sufficient to impress Oxford that I was good enough to do their PhD course and to win a scholarship to do it.

MARIA LITTLE: Max is very English in the way he is self-deprecating and humble. So you would never hear him say anything about himself that makes him sound amazingly clever.

WICKS: He was at Oxford for his PhD and he got a welcome trust fellowship to study at MIT. In academics terms he's won a gold medal for his stage of things. I would love to have that on my CV.

LITTLE: I didn't think I would end up down this math, no. I couldn't have foreseen that at the time, I don't think.

GUPTA: When you imagine the future, what else can we learn from people's voice like this?



GUPTA: Welcome back to THE NEXT LIST. I'm Dr. Sanjay Gupta. Max Little believes his algorithms can someday help doctors diagnose Parkinson's disease, but he has even bigger plans than that.

He wants to put the technology in the hands of patients so they can manage and monitor their own symptoms accurately, and he's doing this with a very creative partnership.


WICKS: I first met Max when I was doing my post-on. I'm a neuropsychologist by training and part of my job was to go out and see Parkinson's disease patients in their own homes. I went and saw about 80 patients. It took me three years. And Max was coming up with a system that could do much of the same stuff immediately in real time using technology.

LITTLE: We're partnering up with a company called "Patients Like Me." What they want to do is they want to bring this technology to the general population.

What they've been doing over the years is setting up a system whereby people can look at their own systems. People, dedicated people, who have been entering their systems into the system for years.

WICKS: One of the tenets we have people want context. They want to know how sick they are. When you get dying those with Parkinson's, you are probably told there are meds for it, it's a slow progression.

There are things we can do to help. You don't know where you might be in five years or ten years. So what we've been trying to do is give people the tools to predict that. Say, well, here is how you are and here is how you compare to the rest of the population.

JAMES HEYWOOD, CO-FOUNDER, "PATIENTS LIKE ME": Max has been starting up this really amazing project that is increasingly accurate to measure or diagnose Parkinson's disease, and we're looking now to take this to the next level. We'll tell you something about the future of the patient.

This is an example of one of our patients with Parkinson's disease. I use the name Judy. She has had Parkinson's. You could tell immediately she has stage three because she joined in '08. She has some old data so she was diagnosed here in 2002.

How sick she's been over this intervening 10 years, so you can sort of make some predictions about where she might go with the disorder. And what you would see here in collaboration is see another line that was sort of the voice lab.

That we would record voice at the same time we started recording her Parkinson's and compare how sick she was in her voice and ultimately if Max's thing works out effectively it would be the driver or one of the main drivers of the score, how sick someone is with Parkinson's disease.

LITTLE: This sort of high-frequency data about how people's symptoms change over time has not readily been available for this. This is a new frontier in exploring how individuals respond to treatment to different drugs. And the value of this, this has enormous value in clinical practice.

For example, a neurologist might want to make a decision about whether or not to terminate a particular treatment or to change to a different drug at some point. Now usually what has to happen now is that someone has to remember how their symptoms changed at the time and of course it's actually very hard. It's hard for people to remember how their symptoms change because they're busy, first off.

WICKS: Think about a doctor sitting in an office and managing 1,000 people with Parkinson's disease. But he doesn't bring everybody in for an appointment every six months or a year. He gets everybody to do Max's system every 30 days or 60 days and the only people that he brings in to see are the people that need to come in.

That's going to allow us to prioritize those who need it not just an arbitrary thing. But when people are in need of an adjustment to their dosages or meds they're taking.

GUPTA: When you look at this and imagine the future, what else can we learn from people's voice like this?

LITTLE: Well, I'm really excited about the possibility that there are other disorders that display this kind of irregularity, anxiety, depression, posttraumatic stress disorder. These all have effects on the voice and we could measure something about somebody's state of mind using the voice.

There are problems to be solved. The problem is how do you get access to measuring somebody's symptoms over time? Well, this is one way of doing it. Any other way will require patients to have a neurologist on hand every day and I don't think that will happen.


GUPTA: Max Little is a daring innovator who is crossing disciplines and pushing boundaries that changed the way doctors diagnose and care for Parkinson's patients. And that's what earns him a spot on THE NEXT LIST.

Thanks for joining us. I'm Dr. Sanjay Gupta. Hope to see you back here next week.