TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

AI model detects asymptomatic Covid-19 infections through phone-recorded coughs

228 点作者 agsamek超过 4 年前

27 条评论

hdbsjdk848j超过 4 年前
A few caveats...<p>1. They trained and tested on a balanced dataset, which is very unlike the data distribution this algorithm would see “in the wild”. Under real world prescreening conditions the data would likely be extremely unbalanced toward the negative class, and also be subject to drift over time.<p>2. They seem to have identified positive subjects through a questionnaire not via clinical chemistry diagnostics; so (a) it is unclear whether their training labels are correct, and (b) they may have completely missed the asymptomatic population.<p>3. As mentioned in another comment ca. 5000 patients and 250K samples is not a lot considering the size and diversity of the population(s) where this would be deployed.<p>Disclaimer: I gave the article the brief high level scan treatment so I could be wrong about any or all of these. Please correct me if I am mistaken.
评论 #24952501 未加载
评论 #24952227 未加载
评论 #24953887 未加载
评论 #24952407 未加载
评论 #24952139 未加载
krajzeg超过 4 年前
One argument for not getting excited just yet:<p>This was not a blinded clinical trial. The subjects all knew whether they have COVID-19 or not and knowing how strong psychological effects can be, what&#x27;s detectable in their cough might be their <i>knowledge</i> they&#x27;re sick. The researchers even acknowledge in the paper that &quot;sentiment&quot; is a big part of how a forced cough sounds.<p>What&#x27;s worrying is also how little of the data was from a diagnostic test (over half of &quot;positive&quot; samples were &quot;self-diagnosed&quot; COVID-19, whatever that means).<p>I don&#x27;t think FDA or any other regulatory body would accept such an app as a screening tool without a proper trial being done.<p>If it works, that would be the most practical and coolest application of ML I&#x27;ve seen - but it still feels like something from the &quot;too good to be true&quot; category at the moment.
评论 #24955297 未加载
graeme超过 4 年前
Notably this works on asymptomatic forced coughs. 17% false positive rate but 98.5% true positive rate. Extremely interesting.
评论 #24951874 未加载
评论 #24953216 未加载
评论 #24954082 未加载
burlesona超过 4 年前
If this really works, this could be the biggest news of the year. Assuming they get FDA approval for this app, we would suddenly have a free, instantly scalable, instantly available test that catches the vast majority of true positives. You could very easily set these up at all public places and require people to check before entry. If that happens you don’t even have to require people to self-test at home, they’ll do it voluntarily to know in advance whether they’ll be let in at their destination.<p>The final and most difficult step would be effective quarantine of infected individuals, some of whom are likely to try and go to work anyway etc.<p>But even if you assume nothing more than voluntary self-quarantine etc, I would expect this to drive R0 below 1 very quickly, as the vast majority of infected would stay home and thus cease to spread the disease.<p>Finally, if all of the above where to come true, I think this could go down in history as the first truly life-changing AI discovery, and potentially one of the biggest watershed moments in recent history.<p>Obviously we’re not there yet, but I am very optimistic and excited after reading the story.
评论 #24951823 未加载
评论 #24953018 未加载
heinrichf超过 4 年前
Discussion on &#x2F;r&#x2F;MachineLearning with caveats here: <a href="https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;MachineLearning&#x2F;comments&#x2F;jkrzlt&#x2F;d_about_the_covid19_artificial_intelligence" rel="nofollow">https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;MachineLearning&#x2F;comments&#x2F;jkrzlt&#x2F;d_a...</a>
elil17超过 4 年前
Reading the headline, I got really nervous imagining how many people would assume that a negative on the cough app means they are COVID-free and then go out and infect everyone.<p>However, reading the article, it seems like the false negative rate is really low. It sounds like this could be an incredibly effective screening tool.
评论 #24951360 未加载
评论 #24953192 未加载
评论 #24952988 未加载
piker超过 4 年前
Relevant paper seems to be: <a href="https:&#x2F;&#x2F;ieeexplore.ieee.org&#x2F;document&#x2F;9208795" rel="nofollow">https:&#x2F;&#x2F;ieeexplore.ieee.org&#x2F;document&#x2F;9208795</a><p>Looks incredible. Hopefully the weights and code are released quickly.
curiousllama超过 4 年前
There&#x27;s a reason they published this in an engineering journal, not medical one. It&#x27;s very cool, but in terms of practical use, is in the realm of &quot;maybe it&#x27;s worth doing a real medical study to see if this works&quot; - a valuable contribution, no doubt, but a not really a contribution to diagnostics in itself (yet).<p>That said, I like that they&#x27;re thinking outside the box on this one. A free digital test with a low false negative rate would be a game changer.
评论 #24955416 未加载
Jabbles超过 4 年前
This sounds extremely useful, if it&#x27;s as accurate as it sounds - scaling this up to test literally billions of people a day would be a major factor in controlling the pandemic, and far easier than scaling physical tests.
dariosalvi78超过 4 年前
I tried to propose this idea at the beginning of the pandemic to a bunch of pulmonologists, some also very active in research, but received no interest at all.<p>Anyway, happy that someone else is doing it.
mamon超过 4 年前
Minor nitpick: if there is cough involved, then those infections are not really asymptomatic, right?
评论 #24952235 未加载
评论 #24952257 未加载
nostrademons超过 4 年前
Also relevant: look at the Google Trends graph for &quot;loss of taste&quot;, both over time and the geographic distribution:<p><a href="https:&#x2F;&#x2F;trends.google.com&#x2F;trends&#x2F;explore?q=loss%20of%20taste&amp;geo=US" rel="nofollow">https:&#x2F;&#x2F;trends.google.com&#x2F;trends&#x2F;explore?q=loss%20of%20taste...</a>
评论 #24951544 未加载
FrancisOfAssisi超过 4 年前
Could not find examples of a &quot;Covid-19 cough&quot; on the WWW! Every link I clicked on was spammed to death and I couldn&#x27;t find any sound files.<p>Does anyone have a link to a sound file that is a good example of a &quot;Covid-19 cough&quot;? I would appreciate it!
dsukhin超过 4 年前
&gt; A user could log in daily, cough into their phone, and instantly get information...<p>While this model is indeed extremely useful and interesting work, this seemingly casual quote gives new meaning to how unsanitary our phones really are&#x2F;can be.
评论 #24951435 未加载
Kreotiko超过 4 年前
I am no expert but isn’t 200k sample a low number? I am worried about false positive
评论 #24951303 未加载
m463超过 4 年前
Now that you or your employer have agreed to the terms and conditions of facetime&#x2F;zoom&#x2F;teams&#x2F;webex&#x2F;skype&#x2F;etc, we now have an opportunity to identify and detai... market to coronavirus sufferers.
Areading314超过 4 年前
The way this is reported doesn&#x27;t give that much information. It says it has a high true positive rate, but what is the false-positive rate? A test isn&#x27;t useful if there is a high rate of these.
dwheeler超过 4 年前
This sounds too good to be true. I want to see a double-blind test done ASAP.<p>I&#x27;m rather skeptical, but the potential upsides are so huge that it&#x27;s worth rapid additional investigation.
verroq超过 4 年前
Now we just need to install public safety microphones† in every home and public space to prevent the spread of disease.<p>You are not against public health are you citizen? You aren’t hiding any covid patients are you?<p>† AKA telescreens
评论 #24951165 未加载
评论 #24953127 未加载
sieste超过 4 年前
Once this becomes widely available as a free app, people will figure out quickly how to cough in a way that gets the desired result.
gfodor超过 4 年前
If this works out this will go down as the most incredible application of ML ever.
blackrock超过 4 年前
This sounds like a bad usage of machine learning. Unless the Covid cough sounds distinctly the same, then there might just be too many false positives that can come from this. Or even a false negative.
Tycho超过 4 年前
I remember back in March a lot of people (within the field) saying AI folk should stand back and let the medical&#x2F;virology community deal with Covid. This sort of thing puts paid to that attitude, IMO.
vaccinator超过 4 年前
Sounds like a job for the NSA
asdev超过 4 年前
how do you select the features for a model like this?
propogandist超过 4 年前
what a perfect way to make the ongoing domestic spying seem like it&#x27;s for &quot;public saftey&quot;.
IgorPartola超过 4 年前
I want an Amazon Echo skill to screen anyone before they come to my house :)<p>Seriously, this is incredibly and if this is verified to work could be a game changer. The real thing we need to do is all get tested at once on the same day at the same time. Then those who are positive need to isolate for 2-3 weeks until they are negative again. That would completely reset us back to nearly zero. Then do this again 2-3 times and we could shove this demon back into the bottle.
评论 #24951488 未加载