AI Program Could Check Blood For Signs of Lung Cancer (theguardian.com) 19
An anonymous reader quotes a report from The Guardian: Scientists have developed an artificial intelligence program that can screen people for lung cancer by analyzing their blood for DNA mutations that drive the disease. The program works by examining free-floating DNA that circulates in the blood. The majority of this genetic detritus enters the bloodstream when harmless cells in the body break down and spill their molecular innards, but tumors also shed DNA as they form and grow larger.
Writing in the journal Nature, the scientists describe how their AI program crunched data on the DNA found in the blood of lung cancer patients to learn which common cancer mutations most effectively predicted the disease. The researchers then used the trained program to distinguish lung cancer patients from healthy people in a separate group of volunteers who gave blood samples for the study. The system cannot confidently diagnose cancer, but instead flags up likely cases for further medical investigation. In tests, the program had a 2% false positive rate -- meaning that it mistakenly flagged two in every 100 healthy people as having the disease -- while rating 55% of stage 2 cancers and nearly 70% of stage 3 cancers as patients likely to have the disease.
Writing in the journal Nature, the scientists describe how their AI program crunched data on the DNA found in the blood of lung cancer patients to learn which common cancer mutations most effectively predicted the disease. The researchers then used the trained program to distinguish lung cancer patients from healthy people in a separate group of volunteers who gave blood samples for the study. The system cannot confidently diagnose cancer, but instead flags up likely cases for further medical investigation. In tests, the program had a 2% false positive rate -- meaning that it mistakenly flagged two in every 100 healthy people as having the disease -- while rating 55% of stage 2 cancers and nearly 70% of stage 3 cancers as patients likely to have the disease.
Can this AI detect... (Score:1)
...syphilitic retards that inhabit the White House? We could use that now.
What about Coronavirus? (Score:1, Informative)
COVID-19 is the real threat here, not lung cancer. Lung cancer is mostly preventable. We need to allocate all of your AI resources to the fight against COVID-19 NOW. This is potentially a civilization-ending crises and I feel like people aren't taking it seriously enough (especially Silicon Valley and the people here on Slashdot).
Re: (Score:3)
Or not.
Note that, so far, covid-19 looks a lot like the pandemic 60-odd years ago. Which is probably where Trump got that 100,000 number last week, since that's about how many people in the USA died in the pandemic 60-odd years ago. What, you didn't know we had
Re: (Score:2)
"In tests, the program had a 2% false positive rate -- meaning that it mistakenly flagged two in every 100 healthy people as having the disease "
In a couple of years we'll see if it really was a _false_ positive or if the AI is just better than the doctors.
Re: (Score:2)
I'd be more concerned about the 45% false-negative rate
Re: (Score:2)
Note that, so far, covid-19 looks a lot like the pandemic 60-odd years ago. Which is probably where Trump got that 100,000 number last week, since that's about how many people in the USA died in the pandemic 60-odd years ago. What, you didn't know we had a pandemic 60-odd years ago that killed 110K+ Americans (and comparable numbers everywhere else)?
~ CrimsonAvenger
There was a time of innocence when simply requesting a citation would suffice, but in the spirit of Robert Frost, I wish to address the Ice-9 level psyops being trotted about a track lately...
[QUOTE] Looks a lot like something 60-odd years ago?[QUOTE]
You didn't know? Well curse our darkness and spit at a grave of your neighbor.
Funny no comparable numbers of 110k+ apply to China. But Fox reports a Corona-related ethnic murder. You wouldn't encourage that, would you? But, Heyyy...Fonzi...Wouldn't somone p
Re: (Score:2)
Civilization ending? Melodrama much?
Spain, which has currently the highest per-capita infection rate (aside of tiny contries where a single infection already skyrockets them to the top) has 2 out of every 1000 people infected. 0.2% infected. With a death toll of about 10%, which is considerably high, by the way, we're looking at 0.02% people dead.
If that ends your civilization, it wasn't worth having it anyway.
AI? (Score:1)
Re: (Score:1)
It seems more and more that "AI" is just any software package the researchers don't understand (or don't want to explain to VCs).
Well, TFA does say "machine learning", which, admittedly, sound less sexy than "AI". Saying we've just used an 85-line Python script to train a model from scikit sounds even more boring. What's even more annoying: why is this article behind a $8.99 paywall??
Re: (Score:2)
The description really does sound a lot like a classic Bayesian spam filter: find the most "interesting" snippets and score them.
Re: (Score:2)
That was always AI, according to the AI nuts.
It's artificial-ly doing intelligent things.
SpamAssassin, the original AI. Wait, no, first computer program, first AI.
Why depend on AI to *do* the test? (Score:1)
I can see using AI as a tool to generate the testing algorithm/protocol, but once a step-by-step "do this and look for that" type test is developed, you don't need an AI to do the test.
Re: (Score:2)
I would assume it would be inference based.
Looking at DNA seems like a good candidate for the sort of complex data inputs that current 'AI' is good at, taking impossibly complex input and outputting a structured classified representation that can then be realistically handled by traditional programming to get it the rest of the way.
Keep in mind, inference 'AI' is no big deal and even low end phone's run inference on camera preview to do things like detect when the people in frame are smiling.
55%, 70% (Score:2)
55% and 70% means there is a lot of improvement needed. What is the rate for other technologies, and can these be used to augment that?
tap (Score:2)
Screw so-called 'AI' nonsense (Score:1)