Can photographs be used to measure Parkinson's?

We are all familiar with the Parkinson's masked face. It's difficult to pin-down exactly what it is but, to put it bluntly, we don't look quite right.

In my case, for as long as I can remember, certainly by the age of 11, people would ask why I looked so miserable, even though as far as I was concerned I was perfectly happy. (Also, while still a kid, people were saying how rounded my shoulders were.) I was diagnosed with Parkinson's at the age of 49.

If others are like me, something as simple as a photograph could form the basis of a very useful research and diagnosis tool that would go some way to answering questions such as:
- Did PD come on slowly or quickly?
- When did I get PD?
- Do non-PwP show any signs of PD?

Put simply, what's needed is a computer system to assess a photograph of someone's face and give a Parkinson's score.

This is basically a face recognition problem. Software exists to do that. But, whether it would be accurate enough to detect the masked face, I don't know. If not, the obvious approach is to use a type of computer program called an artificial neural network (ironically, in our circumstances, this is a type of computer program based on a model of how the brain works) which allows you, first to train a computer to do something and then to get on and do the job.

The collection and analysis of data in this way is similar to what Max Little et al. at the Parkinson's Voice Initiative are doing, but they are collecting voice data using a telephone. I think their system is likely to be more accurate. However, I suspect that few people have a large number of recordings of their voice over time. Whereas with photographs, I suspect that most people could provide at least one photograph for most years of their life.

I'd be grateful for comments?

John

Interesting concept, John. I haven't seen any research on this. I would think facial recognition software would be sophisticated enough to test out a theory. Should mention this to the PPMI team researching early biomarkers.

People thought i too looked miserable way before diagnosis, and I always looked awful in photos unless I arranged my face! It wasn't really till I had comtan added onto my sinemet that I got a more friendly face, for which I was grateful. On the inside I did not feel miserable at all.

You have the germ of an idea there, of course needs to be tested for. Would need a significant percentage of people to have this pre-diagnosis for it to work.

John, as I have previously posted on this board, I am currently actively involved in a research trial for a new drug being tested at Columbia University. While I was getting my drug transfusion today, I got to talking with the doctor running the trial about early biomarkers. One thing led to another and I mentioned your idea about photography and PD masked face. As it turns out, two other doctors at the University have just begun a study of that concept. They're using HD video rather than photos. He didn't know much about the study but told me the docs names who were doing it. The bigger coincidence, though, is that one of the two doctors is actually the husband of my regular MDS (who practices at a different hospital). When I go in for my next visit I will definitely ask her what she knows about the research.

Gary

I think the study has merit. I am always getting embarrassed because people say "I'm sorry" to me alot...I know it is the weird look my eyes have when l-dopa is low. At a Contra dance, one guy said "Don't look at me like that, I'm not going to step on you"...went to the restroom and sure enough..I needed tincture, but didn't realize I was getting low...having too much fun. Contra is like square dancing in a line, with people changing constantly as the couple moves down the line. Fast "Celtic -like" music, and as much energy out -put as aerobics. I hate to be in pictures...my face is so unpredictable.

It doesn't mention PD explicitly, but an article in the latest issue of New Scientist is of interest. Describing work done at Oxford, it says [1]:

"A computer program can now learn to identify rare conditions by analysing a face from an ordinary digital photograph. It should even be able to identify unknown genetic disorders if groups of photos in its database share specific facial features."

Does anyone know of a database containing photographs of people and their diagnoses, if any?

Reference

[1] http://www.newscientist.com/article/...l#.U6--lI1OUcA

John

There is a longitudinal Genetics/PD study being conducted at Beth Israel and Columbia U in NYC in which I participate. As part of the study, they take a series of videos which include facial images as well as walking, balance, coordination, handwriting, spiral drawing etc. This is just one part of an extensive study which is now going to be combined with PPMI. I'm not sure how they will be analyzing the images but I think one doctor at Columbia will be researching facial images and PD. There is also a non-PD control group that is included in the study. Although the PPMI data is open source to the research public, I don't think they would do that with the image for privacy concerns.

I was shocked when I was diagnosed with PD as were all my friends and family. The only symptom I went to have checked was a slowness rubbing 2 fingers together to clean my contact lenses. I initially went to an orthopedic surgeon whom was on my men's soccer team and after examining me he referred me to a MDS who quickly diagnosed me with pd, he also noted that the heel on my right shoe was worn but not on my left shoe. I got a 2nd opinion and that MDS said she could tell I had pd the instant she saw my face.

That said, I have to assume that with the current research being done on biomarkers and the cost of analyzing DNA coming down, it will be standard procedure to identify those who have a higher chance of getting pd while still in the womb and in subsequent years it will be a simple blood test for pd biomarkers and lots of other potential chronic diseases. Seems environment plays a bigger role than genetics so i assume just like with colonoscopies, neurological disease biomarker testing will be included in a standard checkup if your DNA predisposes you to one of those diseases?

Facial recognition is obviously cheaper and there's no doubt pd is underdiagnosed so doctors need every tool they can get. I occasionally played soccer with my GP and he didn't pick up on my total lack of right arm swing and that I fell a few times unexpectedly while planting on my right leg, it was odd enough to elicit laughs from other players. I think though that your're going to have to have a very high correlation for any procedure to be adopted, too many false positives would get people very very angry. So I assume facial diagnosis if found useful would just be part of the diagnostic process?

Thanks for the responses.

Tupelo3 is right in raising privacy concerns. Some of these are described in the thread:
"Ethical implications of the remote measurement of Parkinson's"
http://neurotalk.psychcentral.com/thread198299.html

soccertese raises an interesting anecdote:
"I got a 2nd opinion and that MDS said she could tell I had pd the instant she saw my face."
The opportunity is now lost, but it would be interesting to know whether her reaction would have been the same had you presented a year earlier, or two years earlier or whatever.

Diagnosis is just one of the uses of this approach. Other uses include:
- finding the time from the first signs to diagnosis;
- identifying geographical variations in the prevalence of Parkinson's;
- cluster analysis of PD sub-types;
- measuring the progression of PD;
- high frequency monitoring of symptoms.

I suspect that using automated measuring techniques such as this and other similar approaches (e.g. voice, movement etc.) will rapidly impact on the way that clinical trials are run.

John

this issue of the journal of parkinson's has some interesting abstracts on smell.
the bigger issue is why did england get booted out of the WORLD CUP so soon?

http://www.journalofparkinsonsdiseas...ent_Issue.html