Hacker News new | past | comments | ask | show | jobs | submit login
Say 'Ahhh': A Simpler Way To Detect Parkinson's (npr.org)
91 points by gruseom on July 22, 2012 | hide | past | favorite | 28 comments



The paper from the same group last year on the topic is here: http://rsif.royalsocietypublishing.org/content/8/59/842.full...

They are using random forest out-of-sample error as a metric but doing feature selection before this step (see table 6).

As far as I can make out from a quick reading they are essentially making the error described here: http://www-stat.stanford.edu/%7Etibs/sta306b/cvwrong.pdf

and elegantly described in this recent blog post: http://blog.kaggle.com/2012/07/06/the-dangers-of-overfitting...

On a sample size of only 42 people, overfitting seems very likely.


One big issue with gathering this kind of data over the phone is the frequency cutoff on voice-only lines—above a certain frequency (I want to say 4kHz? maybe I'm misremembering), the information is lost. It's basically as if you took a Fourier transform, zeroed out everything above the threshold, and then transformed back.[0] For humans (and even computers) trying to interpret the sound as language, that's not a huge problem, although you might lose some of the higher formants. But for an acoustic analysis that's trying to do voiceprinting—in this case to detect Parkinson's—this could be a big problem.

(I'm also irritated by the glib "99 percent success rate" but I just ranted about that on a different HN post so I won't go into it here.)

[0] Why do this? So the phone company can compress and send a lot more data over the same amount of internal bandwidth. Come to think of it, it's kind of related to how wavelet-based compression works.


8 bits, 8 kHz. There's only so much you can do with that.

http://en.wikipedia.org/wiki/DS0

Sure, we can do more now, but this isn't exactly new. Consider the age in which it was conceived and it may seem more reasonable.


The sampling frequency doesn't directly effect the audio frequencies it can encode. Telephones do PCM encoding (meaning it has data representing the graph of the sound wave) at 8 kh/z. Following the nyquist sampling theorem (cut your rate by 2), this can allow frequencies up to 4 kh/z (as you said). It's not a hard cutoff though, you can still get most of the sounds above that pitch, they'll just sound pretty weird (as if you were talking on the telephone!)


No, components above the Nyquist frequency are filtered out before the ADC because you'd otherwise get aliasing.

https://en.wikipedia.org/wiki/Aliasing


> So the phone company can compress and send a lot more data

Actually, the limits date back to analog phone lines with circuit switching of a century ago. Back then there was a wire going through switches from one phone to another. The quality requirements were that those lines had to pass 300 Hz to 3.4 kHz or so. That often required "pupinization" - adding inductive coils to tune the line's frequency response.

If you look at a spectrogram of your voice, there's very little power above 1.5 or 2 kHz. However, it seems the high frequency part is important to understandability, including perception of emotional overtones.

(Just the other day, playing with modems, we found a weird case - voice being pumped through before the call was considered completed - which I suspect is the persistence in digital protocols of the analog behavior of a century ago.)


It's a very interesting field. Can insurance companies detect the early stages of Parkinson's when you call their call-centers and change your rates accordingly? Here's a related dissertation on detecting mental health condition by using voice recordings http://www.eecs.berkeley.edu/Pubs/TechRpts/2012/EECS-2012-55... (haven't read it yet, but was meaning to for quite a while)


The NPR story has a non-obvious link to the site. Here it is: http://www.parkinsonsvoice.org/

That's pretty amazing if they can detect it within the poor 8khz of a phone. I wonder how awful cell phone reception changes the accuracy.


While 8 kHz is low if you look at the carrier of the speech signal, which is something like the pitch of the speech, it is pretty high compared to the envelope frequency of the speech signal. You do not move your jaw, tounge and lips at anywhere near 8000 times per second. The algorithm looks at things like how the jaw, tounge and lips move during speech, and I guess 8 kHz is quite enough for those.


I've never had poor cell phone reception change the frequency of the sound, only silence, choppiness, or complete garbage being inserted - all of which should be detectable via typical DSP algos.


If this works successfully then it would be great but as a 28 year old with Parkinson's, I'm skeptical.

It took approximately 12 months, numerous blood tests, MRIs and doctor visits to diagnose me as having YOPD, so I'm finding it hard to believe that this could be replaced with a telephone call.


If you don't mind answering, what was your path to the Parkinson's diagnosis? I mean, what were some of the first signs that something was wrong?


I developed a slight tremor in my right hand and had trouble moving my arm as fluidly as my other.


I wonder what the false-negative rate of this algorithm is. 99% (which I assume is the true-positive rate) is certainly impressive for such a simple test, though.

It would be all kinds of awesome if this turns into something that can be done reliably and routinely!

Edit: seems like 99% is only for later stages of Parkinson's, and the accuracy number is just off a 50-person sample. Less impressive, but still cool. http://www.forbes.com/sites/singularity/2012/07/03/new-softw...


How do you get 99% accuracy for a 50 person sample? Has to be 100% or 98% surely?


Interesting that I got cut off during the "Ahh" test and asked to redo it. I'm sure I don't have exceptional lung capacity.


I did too the first time. Maybe that was on purpose.


shameless plug of some ideas i've had about parkinson's disease; i'd be interested to hear what others think about my reasoning:

Michael J Fox's Parkinson's Disease may have been caused by amphetamine use and/or sleep deprivation http://www.nathanwailes.com/forum/viewtopic.php?f=3&t=22...

Sleep Deprivation May Be a Common Cause of Parkinson's Dis. http://www.nathanwailes.com/forum/viewtopic.php?f=3&t=24...


Yeah, no offense, but all you've written is either widely known (Amphetamine-induced DA depletion inducing Parkinson's-like motor symptoms and so on), based on misunderstandings (some of your graph readings are horrifically beside the point), or simply wrong.

Grab a book on neuropharmacology, or even just a neuroscience intro ("Fundamental Neuroscience" by Squire and colleagues is a good one), and accept that armchair-bullshitting is not how progress is made.


Hey apl, thanks for considering my ideas! I hope you won't mind if i try to defend them against your objections:

re: "it's widely known that amph. can induce Parkinson's-like motor symptoms" This may be widely known among neuroscientists, but it's certainly not widely known among the general public, which is my audience for the essay.

re: "some of your graph readings are horrifically beside the point" There's really only one graph I read from, so I'm going to assume you're talking about the "Loss of DA Neurons After Heavy Meth Use". Why, exactly, reading that graph beside the point? The study I cite immediately prior to that graph (from a Michael J Fox source, no less) states my point for me: using amphetamines can lead to fewer dopamine neurons.

re: "some of what you've written is simply wrong" Please be more specific! I want to know what you found wrong.

re: "grab a book on neuropharmacology" I was actually a psychology major in college and took several classes on neuroscience and neuropharmacology, so I'm not totally new to this field.

re: "armchair bullshitting is not how progress is made" This wasn't armchair bullshitting; I was combining the results of several different published studies and other concrete information to arrive at a novel hypothesis. I agree my reasoning isn't enough to remove a great deal of uncertainty about whether the hypothesis is true, but all I'm trying to do is establish that it's plausible enough to warrant people's attention. Many scientific advances are made this way; for example, in physics there are often speculative theories made that need to be verified through experiment: Einstein's theories and the idea of a Higgs Boson, the idea of dark matter and dark energy. All of them started off as unproven possible explanations of observed phenomena, and all needed to be verified. But that doesn't mean they weren't worth giving attention before they were verified.


I'll give you a couple of examples.

  > is sleep a cause or a result of some other underlying 
  > factor? - Not sure what this objection means, exactly, yet.
It's simple, really: Say most PD patients present with sleep dysfunctionalities. What conclusions can we draw from that? Three, actually. That PD causes sleep dysfunction, that PD is caused by sleep irregularities, or that both are caused by some unknown cause X. Based on correlative data, all three are equally likely. You, however, immediately jump to the sleep causes PD conclusion -- critically, without mentioning a mechanism for how that could work.

If you investigated PD a bit further, you'd realize that there's dozens of confirmed comorbidities (e.g., depression, cognitive issues, hyposmia, etc.). That doesn't tell us anything about anything.

  > Amphetamines are drugs associated with pushing your 
  > brain to operate without sleep, and they're also 
  > associated with causing Parkinson's Disease
Well, you know what else amphetamines are associated with? Interference with the dopaminergic system at a massive scale. What's the key neuronal pathology in PD? Right, SN-focused DA cell degeneration. So it's exceedingly more likely that DA function plays the role of causal link here than lack of sleep. Again, common cause.

Regarding genetics and Fox: We know that PD has a genetic component; there's promising mouse models and solid epidemiological data. Fox is one data point. His lack of family history doesn't suggest anything at all.

In summary, your whole argument rests on a flawed logical strategy. You're connecting three semi-confirmed factoids (increased PD risk among amphetamine users, amphetamines lessen the need for sleep and are taken by traditionally sleep-deprived groups) by positing a direct causal connection; namely, amphetamines -> no sleep -> PD. All this while ignoring much more sensible connections (e.g., via dopaminergic systems).

Sure, one could test this. But that's faint praise -- it's still far away from what cutting-edge PD research does right now and for good reason. BTW, Higgs and Einstein developed their ideas from deep knowledge of and appreciation for state of the art research at their time; both knew their stuff and went from there. You're displaying very basic knowledge, tendency for random sourcing, and ignorance of fairly basic scientific principles (experimental logic esp.).


re: "Say most PD patients present with sleep dysfunctionalities."

I understand that idea, but the problem with it is that I'm not just saying that PD is correlated with coexisting sleep deprivation; I'm saying that evidence suggests PD is associated with a history of pre-existing sleep deprivation, where the sleep deprivation is itself correlated with the demands of a person's job/lifestyle. It seems obvious to me that the most plausible cause-and-effect relationship there is that the sleep deprivation is causing the PD, rather than vice versa or some third factor causing both.

re: "If you investigated PD a bit further, you'd realize that there's dozens of confirmed comorbidities. That doesn't tell us anything about anything."

I've actually read the literature reviews that neuroscientists use to learn about this stuff. I even include links to those journal articles. Here's the most recent one I found: "2011.04 - Epidemiology and etiology of Parkinson’s disease: a review of the evidence." I mention this journal article in the essay on PD and sleep.

You're right that comorbidities don't necessarily tell us much. But the evidence I'm talking about seems different, because the factor I'm discussing (sleep dep / job demands) predates the PD symptoms. You could argue that an underlying cause of PD was latent in these people earlier in life (their 20s-40s) and causing them to work longer hours and get less sleep, but it just seems much simpler and more sensible to me to say that people were working harder, getting less sleep, and paying a price as a result.

"it's exceedingly more likely that DA function plays the role of causal link here than lack of sleep."

I think you misunderstood the point I was making in that section; I was writing that for people who did not know much about how the brain works. I agree it seems clear it's the DA neuron death that's causing the PD symptoms; my hypothesis is that lack of sleep over a lifetime is causing gradual DA neuron death.

re: "Fox's lack of family history doesn't suggest anything at all."

I'm not trying to establish that it's lock-tight that there was no genetic component; I'm trying to explain to the average person out there that Fox's situation was not a case where everyone in his family had PD and so he got PD too. For a long time I assumed that's what had happened. Again, it isn't a lock-tight piece of evidence, but I do think it's relevant and adds weight to the environment-side of the scale (I'm imagining a scale where genetics is on one side and environment is on the other). It just seems more likely to have been caused by his environment if his family doesn't have a history of PD. That seems totally reasonable to me, and I think it would to most people.

re: "You're connecting three semi-confirmed factoids (increased PD risk among amphetamine users, amphetamines lessen the need for sleep and are taken by traditionally sleep-deprived groups) by positing a direct causal connection; namely, amphetamines -> no sleep -> PD. All this while ignoring much more sensible connections (e.g., via dopaminergic systems)."

Again, I think you misunderstood what I was trying to get across in my essay: I was saying that the amphetamines AND / OR sleep were causing damage to the DA neurons. So it seems plausible to me that amphetamines are causing the DA damage on their own and sleep dep. has no effect, or that amph. have no effect and only cause DA damage by causing sleep dep, or that both amph. and sleep dep. are able to independently cause DA neuron damage.

Thank you again for your feedback! There are some complicated facts here to process (complicated for those who haven't studied this subject) and so it's helpful for me to see when I'm not communicating my ideas as clearly as I would like.


Wonder whether it's possible for them to crowdsource this via web instead of the phone lines, which would make it easier for more people to participate.


Average onset of Parkinson's happens at 60 years old.


That is given the poor tools we have to date to diagnose the disease. I think the really promising thing about this development is the possibility of earlier warning and being able to test new treatments before the disease progress's so far that someone is going in to ask the doctor what is wrong.

My father has Parkinson's, it is an awful disease.


Possibly just use something like Twillo or Skype to make this global, simple and cheap/free?


This will come as no surprise to the orthodontists in the UK and US who treat some cases of Parkinson's as a disorder of the mandible.

http://www.youtube.com/watch?v=VYKBPd2Kowo


would be cool to see somebody build a twilio app to link on the article and collect voice samples right off the page




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: