Slashdot Mirror
Algorithm Brings Speedier, Safer CT Scans
kenekaplan writes "Standard CT scanners can generate images of patient's body in less than five minutes today, but the radiation dose can be equal to about 70 chest X-rays. Lower-powered CT scans can be used in non-emergency situations, but it can take more than four days to produce those images. Intel and GE created an algorithm that speeds up a computer's ability to process the low radiation dose scans by 100x, from 100 hours per image to one hour."
1. I hate 'news articles' that are chock full of hyperbole and mis information. TFA implies that most CT scanning is done in the ER for life or death reasons which is hardly true. It oversells the current radiation dose of modern 16+ slice scanners and attempts to lower the radiation doses for all CTs.
2. Current gen CT scanners cut the dose of most tests by at least half from the second and third gen scanners. Of course, TFA doesn't mention how good the new dose regimens are in terms of decreasing dose.
3. It appears that this new tech has a significant price tag. TFA quoted 1.5 million for a 128 slice scanner with the "new algorithm". More slices = faster and more resolution, but mostly faster. The current 'top of the line' is 64 slice. "Standard" CTs are 16 slice and cost anywhere from $150 - 250K.
4. At least the GE scanners run Linux!
Faster! Faster! Faster would be better!
Technology has been making some huge leaps and bounds over the past few years, it is almost hard to believe a few years ago I was playing Ultima Online lol. This is huge for the medical field in treating and diagnosing people without adding another problem to the list. Keep these algorithms coming! Save mankind computers and prove those terminator movies wrong!
Sounds like a job for OpenCL. A GPU cluster would be much more scalable than using expensive Xeons. Which also makes this article sound like an add for Intel CPUs.
-- Home is where you eat your heart out.
How to obtain a 100x speedup: consider an architecture with 112 cores, wait for 2 CPU generations to pass, and put 3 engineers on the task for several years to parallelize the algorithm by hand. Of course, giving it to the research community would have been impossible because, you know, it may have worked faster with more general solutions.
Inquiring minds want to know, since this will substantially reduce the needed resources for a scan, how much cheaper will they be?
You can stop laughing now!
I had a CT scan two weeks ago. I didn't fully realize until after I was done about the amount of radiation I was exposed to. My arm where the IV was injected with radiation hurt like hell for about 18 hours and of course I now have a higher risk of cancer. There's enough radiation in the injection that it makes you feel like you're generating heat from the inside. It's quite a weird feeling. I guess it was needed for the procedure I had to have done, but here's hoping for improvements with lowered radiation exposure.
Seismology uses similar model-based reconstruction algorithms. the danger is you can force the data into something that looks a lot like the input model if you are not careful. Technically this get stuck on on a false local minimum in an optimization problem.
A while back I had several full-body CT scans on an emergency basis. They found what they were looking for in my liver and it was treated. But I was forbidden to have any X-Rays of any kind for two years after that. So when I came down with bronchitis and pneumonia, the doctor had to play it by ear (literally, he just listened to my chest). All is well, now. But lowering the X-Ray dosage of CT scans is very worthwhile.
I18N == Intergalacticization
Is it really a new algorithm, or is it just that they hand-tuned the code to run iterative reconstruction quickly? There's a world of difference. There are some great algorithms out there to speed up calculation of large images where you expect them to be compressible in some basis, but from this article it looks like they didn't invent a new fancy algorithm, they just heavily optimized an existing one. Anybody have a link to a technical paper so we can find out for sure?
Expected time to finish is 1 hour and 60 minutes.
This is old news: http://www.genewscenter.com/Press-Releases/GE-Healthcare-Unveils-Ultra-Low-Dose-CT-Technology-with-Profound-Image-Clarity-3367.aspx And if you read through the Intel and GE press releases, you'll find numbers all over the map as to how much this actually decreases radiation exposure. It might be a 4x reduction (GE scientist quoted in Intel article), it might be a 10x reduction (Intel article), or it might be a 100x reduction (GE article). It might just depend on the specific scan being done, but you won't find that in either article. This is quite a breakthrough and is fantastic news for anyone who needs regular scans, but it is a bit overstated (as mentioned by other commenters, CT scans had greatly improved from the doses quoted by GE/Intel as baseline figures), is nearing 2 years of availability in some areas, and is going to see slow adoption as hospitals a) aren't willing to part with the expensive, working machines they have and b) the time requirement of an hour is still significant enough to rule this out in many environments. There is still a key issue that isn't mentioned here. It might take 15 minutes to get a prepped patient in the room, to scan them, and then to do all necessary work to prepare for another patient, but it takes an hour to analyze a scan. So, we either get 4 servers to process the scans (I think Intel would like this very much) or we run under capacity or we leave a long queue of scans to be analyzed overnight and read the next day. The 1st option is expensive, the 2nd option is not going to happen if there is a living accountant in the building, and the third option would still require an extra server (assuming we do more than 24 scans each day) and is only workable if there can be a day+ delay in sending results. Perhaps some Slashdotter will make a fortune with their CT analysis "cloud." And not to sound like a party-pooper, but for customers in the U.S. GE/Intel still have no power over the insurance approval process which can delay your scan by another order of magnitude. Let's celebrate this for what it is, a nice improvement for a small niche of healthcare consumers who require several CT scans in their lifetime, and not for the miracle-machine that the GE/Intel marketing department would have us believe.
Except for some vague references to parallelizing (alleged) essentially single threaded code and peephole optimization, what have they done? Where's the great advance in software that the headline promises? How are the new algorithms difference from the old ones? Where's the statement of the underlying difficulties? TFA is inexcusable puffery.
Contribute to civilization: ari.aynrand.org/donate
On the other hand, there's quite a lot of difference betwen a hand, a head, a foot, a chest..
If the initial model used for the reconstruction is just some generic approximative shape, it will still provide some speed up, but won't be affected that much by the actual pathology which is being imaged.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]