IBM's Watson To Be Used For Cancer Treatment 46
Beeftopia (1846720) writes "The New York Genome Center and IBM will investigate whether Watson can be used to parse cancer genome data and then recommend treatments. The trial involves 20 to 25 glioblastoma patients with poor prognoses. The article states, 'It should theoretically be possible to analyze [genomic] data and use it to customize a treatment that targets the specific mutations present in tumor cells. But right now, doing so requires a squad of highly trained geneticists, genomics experts, and clinicians. It's a situation that can't scale to handle the [number of] patients with glioblastoma, much less other cancers. Instead, that gusher of information is going to be pointed at Watson... Watson will figure out which mutations are distinct to the tumor, what protein networks they effect, and which drugs target proteins that are part of those networks. The net result will be a picture of the biochemical landscape inside the tumor cells, along with some suggestions on how clinicians might consider intervening to change the landscape.'"
Faster, but smarter? (Score:1)
Re: (Score:2)
It's not particularly important, honestly. Sure, yes, accuracy is important and nobody will use the thing if it can't be accurate a good bit of the time. However, I can't imagine it would be anything other than a fast consultant. A source for a second opinion. I don't think for a minute that we'll be relying on digital diagnosis for a very long time.
How many Sci-Fi shows have you seen where something like this happens:
Nurse: Doctor, the patient's condition is declining rapidly.
Doctor: I don't understand
Re: (Score:2)
Hey, thanks for eating my angle brackets, Slashdot. You're the only comment system on the Internet that still thinks that's a good idea to parse user content as tags.
Re: (Score:2)
So Slashdot is *ALSO* the only commenting system around that doesn't allow editing either! Two distinctions for the site!
Though, oddly, I've stuck with HTML formatting for my commenting since I created an account. It's a bit of a pain at times but it keeps me involved in the process and more focused on the preview than I would be otherwise, I think.
Re: (Score:3)
So Slashdot is *ALSO* the only commenting system around that doesn't allow editing either!
That's always been by design, along with the inability to delete comments.
A user recently suggested allowing editing, but providing a version history (like Wikipedia does for edits). I thought that was a cool idea. Once we get some spare engineering time, we're going to do another pass on the comment system, and maybe that's something we can implement.
Re:Faster, but smarter? (Score:5, Insightful)
Lets say hypothetically, just pulling numbers out of no where, that Watson is 1/10th the price of a specialized team but only half as accurate. I think those numbers are probably a bit on the pessimistic side... yes Watson is an expensive system but each query will be completed in minutes or at most hours, the marginal cost of each additional patient just isn't very high when compared to a multidisciplinary team of geneticists, oncologists, toxicologists, and general practitioners. As for accuracy, well to put it simply this kind of network analysis is what Watson was designed from the ground up to do and it does it shockingly well.
But I digress, back to the example. Lets say there are enough specialized teams to treat 1,000 glioblastoma patients per year and they successfully treat 80% of patients. 800 saved hurray! But, for the same price, Watson could develop treatments for 10,000 patients, saving 4,000 of them. The of course there is the fact that Watson is not a build once and done kind of system. Every year there'd be 10,000 new pieces of information to be entered into the system, refining the probabilities further and further. Sooner or later, Watson will be not just equal to the human team, but will far surpass it.
Re: (Score:1)
I know the VA was looking to reduce costs by analyzing treatments. They would never tell a Dr. you can't perform treatment X, instead if he favors things considered outside the norm by the path determined by the "experts" in the field, they would analyze why. They found a good amount of the time a Dr. didn't know about a treatment, new drug, etc so this was a good way of doing continuing education. Other times, the Dr. was right, so they then considered if the program should be updated. I know it was op
Re: (Score:3)
... Lets say there are enough specialized teams to treat 1,000 glioblastoma patients per year and they successfully treat 80% of patients. 800 saved hurray! But, for the same price, Watson could develop treatments for 10,000 patients, saving 4,000 of them.
The idea of "saved" in this case is overly optimistic.
From Glioblastoma Multiforme [wikipedia.org] and Glioblastoma [abta.org]:
The median survival time from the time of diagnosis without any treatment is 3 months, but with treatment survival of 1–2 years is common.
For adults with more aggressive glioblastoma, treated with concurrent temozolamide and radiation therapy, median survival is about 14.6 months and two-year survival is 30%. However, a 2009 study reported that almost 10% of patients with glioblastoma may live five years or longer.
My wife died of this in 2006 just 7 weeks after diagnosis Remember Sue... [tumblr.com].
Though perhaps with Watson many more people could be, at least, helped.
Re: (Score:2)
Sooner or later, Watson will be not just equal to the human team, but will far surpass it.
And then that multidisciplinary team of geneticists, oncologists, toxicologists, and general practitioners can retrain to get jobs fixing robots right alongside the truck drivers and warehouse workers.
Re:Faster, but smarter? (Score:5, Insightful)
I wonder if it can come up with more accurate answers over a longer term vs. a team of researchers and scientists.
Well, the point of TFA is that a "team of researchers and scientists" is not actually available or affordable for the vast majority of patients. The possibility that they might be better in theory, doesn't really help with reality. Also, Watson should improve with feedback, more data, and increased computing power.
Re: (Score:2)
"While Watson can come up with answers much faster than a human, as demonstrated by the Jeopardy showcase, I wonder if it can come up with more accurate answers over a longer term vs. a team of researchers and scientists."
Sure. It also has not the slightest nightmares about malpractice insurance, also no hangovers, ex-wives lawyers harassing them and it works 24/7.
So long-term, no problem.
Re: (Score:1)
Re: (Score:2)
Neurosurgery Research (Score:5, Insightful)
This is good news for the future of medical AI, but in a frank discussion with a neurosurgeon about research in this area, he admitted that glioblastoma patients are the first place we see a lot of experimental treatments because their prognosis is so poor they'll try anything. If you come up with a mildly reasonable excuse to hit them in the head with a brick, they'll jump at the opportunity to use brickotherapy to cure their cancer.
The point is, the extremely poor prognosis means that there's a low bar for something to work, not that this is the area Watson is most needed or could make the greatest impact.
I still view it as an extremely positive development, just trying to temper the enthusiasm.
Much better than Watson's previous job (Score:3)
Re: (Score:2)
AI is heeeeeeeeeere (Score:2)
It's replacing highly skilled people in specific domains. It's coming to get you specialist medical technicians. You're next taxi drivers and truckers. Whose head is on the chopping block after that?
I still suspect programmers, but it could be someone else.
Re: (Score:1)
If you don't get kicked by machines then you get kicked by offshoring. The future's so bright, my robot wears shades.
Re: (Score:2)
It's replacing highly skilled people in specific domains. It's coming to get you specialist medical technicians. You're next taxi drivers and truckers. Whose head is on the chopping block after that?
I still suspect programmers, but it could be someone else.
Honestly, I'm a bit surprised that most programming isn't automated these days.
Considering how much sci-fi I've seen regarding AI gone wild, perhaps we should be thankful the machines cannot yet give themselves new instructions.
Re: (Score:2)
Honestly, I'm a bit surprised that most programming isn't automated these days.
Most programming has been automated for decades. It's computers that generate dozens of machine instructions for each line of code you write. You don't have to figure out a vast majority of that stuff anymore.
not a very good tool (Score:3, Funny)
How long until... (Score:1)
How long until it comes to the conclusion that Humanity is the only cancer that needs erradication?
Why is Watson needed for this? (Score:2)
Re:Why is Watson needed for this? (Score:4, Insightful)
There are a couple of possibilities. It is entirely possible that the very smart people working on Watson recognized some overlap between the two problem sets and recognized that they could apply lessons learned to this new goal. It is also entirely possible that marketing decided to "cure cancer" and paid very little heed to what Watson did in the past.
Re: (Score:2)
It's not a matter of whether cancer research needs Watson, it's a matter of IBM having dumped a large amount of R&D dollars into a solution in search of a problem. This is merely a last-ditch effort to make it pay off.
Re: (Score:2)
Think a bit more about what it means to communicate and what that means for a computer.
Watson is developed to do just that, it doesn't matter if the data itself is natural speech or genetic data. Data i
Re: (Score:2)
Disclosure: I work at the NY Genome Center.
The reason the language processing is important is because this application of Watson ingests journal articles to "learn" that tumors with mutation X affect pathway P and cross reference that with other articles that might say drug D is effective for regulating pathway P.
This creates a knowledge base watson can then use to apply to a sequenced tumor that has mutations that may include X and can suggest the possible effectiveness of drug D.
The Watson team has done a
Re: (Score:2)
What Watson did to win at Jeopardy doesn’t seem to have that much of a connection to decoding which genome sequences affect protein pathways
Consider that Watson is programmed to learn and then tutored by experts in a particular domain. The are two types of experts, experts in the subject under question and experts in the behaviour of Watson. Also consider that Watson has been at "medical school" for the past year or so, and if allowed, could in all likelihood pass the test for a license to practice medicine in the US.
Watson is first and foremost a pattern seeking engine and both very dissimilar tasks can be boiled down to seeking an arbitrar
BS (Score:1)
watson is a highly overrated marketing tool. is it possible it will find a "link" between a treatment modality and a particular genotype? sure. so would any number of dedicated machine learning techniques run on some HPC. its about quality of data input, quantity, clinical parameters etc etc. and then clinical trial after trial to show that machine aided decision making works better (or help supplement) a medical doctors opinion. there is nothing magical about watson except maybe what kind of rainbows and p
Diagnosis (Score:2)
I have cancer (Score:1)