a radiologist might examine thousands of patient images a day
Shudder. As a young man, I used to inspect new cars for shipping damage. It was well known in the business that an inspector's effectiveness dropped off dramatically as the day progressed. Missed damage could cost our customer thousands of dollars, and yet even blatantly obvious stuff was missed from time to time. I once missed frame damage on a pickup truck that had been chained down a little too enthusiastically. The 45k truck was a write off.
Our inspections numbered in the "hundreds per day". I can't even begin to imagine the mental haze at the end of thousands.
Stating it as "thousands of patient images a day" is misleading. It would be the same as saying you inspected "thousands of parts each day". As the radiologist further down notes, CT scans contains many slices.
While computers don't get tired, they also have a really hard time solving stuff like annotation tasks automatically. One thing is getting a good enough general performance, another is to never make critical errors. I see a huge potential for ML approaches in health care, but primarily as an aid for the health care professionals and not as a full replacement.
Maybe not right now, but there is nothing to say that it can't eventually surpass humans in effectiveness and critical error rate. 10-15 years ago people would have said "yeah self-driving cars are good, but as an aid for the driver, never as a full replacement".
Yes, it's called decision fatigue and it's a real thing. Don't make important decisions in your life at the end of a long day where you've already made many, many decisions.
I don't think radiologists can examine and report on thousands of images. Have you seen a radiology report? It's a detailed 1-2 page writeup of findings. I am sure they have software that may autocomplete some stuff, or they dictate the findings to someone else. Still, the math doesn't check out.
Radiologist here. A chest abdomen pelvis cat scan I routinely read as about 500 axial images at 2.5 mm slices. Then there is a comparison or two or more. Then there are different phases of contrast for the liver, kidneys. Then there are coronal and Sagittal reconstructions. I easily look at an average of 2000 images per patient.
Voice recognition software autocomplete is about as good as your phone's autocorrect...
Is this a general comment or is there a specific service that you're having problems with? There are IBMers around here (for example, me) that really want to create a great platform. If it's a specific service that's causing problems, let me know and we'll figure out what's going on.
I'm not who you are replying to, but I am a developer at Merge who is excited to try out your toolkits! I think there are a ton of opportunities for medical image analysis in the PACS world. I really hope that IBM will challenge us with the task of integrating Watson-technology into the PACS workflow. We've got some really smart employees here and I am hopeful this represents a huge opportunity for us.
(Secretly I can't help but think the smart business decision is to "black box" the image analysis as much as possible, but I can't help but being giddy about it even so!)
It is a general comment about 3-4 specific services we have been using. I'm actually coming to give feedback in IBM in two weeks. If you'd like I can gladly send you a copy of it if you tell me how.
Maybe I'm too cynical, but my first thought when reading this was, let's see how highly paid professionals react when technology is coming for their jobs.
But of course, a few phone calls, the passing of some additional legislation if needed, and they will be completely protected from any financial ramifications. At least the world will still be a better place if this technology actually works.
That may not work out. If radiologists block its use, then the next logical customer is malpractice lawyers. Use it to sift potential clients for ones where early indications were missed. After a few of these suits succeed then the hospital industry will have to institute automated analysis as a defensive measure.
Radiologist here. Not too worried about watson reading cat scans anytime soon. image segmentation is not a solved problem. Deep belief networks have been doing image classification, not segmentation. I welcome more competition in the radiology Pacs business though; which I think is all this is.
Thanks for the link. I myself have attempted using deep learning for segmentation with limited success with and without help from people with background in deep learning. I am not saying it's impossible but there is a lot of work left to do.
Yeah, with the renaissance that image analysis is undergoing in the last few years with DNNs, to say anything is impossible is showing a lack of imagination.
anyways, that's not my point. i'm saying deep learning for image analysis has been a huge success, and people should explore ways to apply this success to medical imaging.
but just for your info, neural networks are slowly being generalized to a lot of other challenges. look around for topics on recurrent neural networks, memory networks, reinforcement learning, etc etc. i don't think we've fully finished exploring the many ways neural networks can help solve life's challenges just yet.
Wow! Good. Thanks for this also. Last time I searched was a while ago, and I didn't find much. Glad to see progress. I will look into it "deeper" again.
Not everything is classification and there are problems better solved with other methods. For example I also haven't seeing anyone playing chess or go with DBN's. Also Compression, general programming...
Classification is a huge part of chess programs. You need to quickly evaluate a millions of boards to decide which side has the advantage. However, do to the well understood rules we can write efficient classifiers by hand.
Go is a much harder problem in large part because it's really hard to accurately classify which board is better off.
Teaching Deep Convolutional Neural Networks to Play Go: "Our convolutional neural networks can consistently defeat the well known Go program GNU Go... It is also able to win some games against state of the art Go playing program Fuego while using a fraction of the play time."http://arxiv.org/abs/1412.3409
You can say that about almost anything, and the world is still full of factory workers.
As a PhD student in medical imaging, you must also know that getting fully automating segmentation methods to work to the standard required in the clinic is really hard. And once you solve it for one clinic you will likely not be able to transfer the trained model to another clinic, because scan parameters, patients and workflow are different.
But when we solve the segmentation task, I think most radiologist will clap their hands and move on.
I follow the progress in deep learning too, Radiologists have been close to out of a job since the 80's, but this time may be different. Deep conv nets have basically allowed computers to surpass us in identifying visual patterns. The bottleneck is just the training data.
Are there applications of this technology in other areas of radiology as an automated second opinion? Would the culture of Radiology welcome such checks or would it be seen as a threat?
My brother is a lawyer, working in some large corporate contracts mill. Their division just got sold to a huge conglomerate and the labor outsourced overseas. This is happening and it is nothing to cheer. The squeeze will slowly move up the food chain until only the billionaires remain.
No, it certainly isn't something to cheer, sorry if it sounded that way. I just have this general sense that the huge gap in public opinion on outsourcing & similar matters is almost completely aligned with each individuals income or place on the financial hierarchy - while the lower and middle class are raising an alarm, the well compensated professionals generally have a "tsk tsk, should have got a real education" attitude, or at the very least "there has always been new jobs created, and there will this time as well".
I'm fortunate enough (so far) to still have one of those high paying jobs, but I worry a lot about not just myself but for society in general. To me a very likely natural progression is that as the power of technology increases, it will be the well connected monied class who either harness it to make themselves even more wealthy, or at the very least use the courts to protect themselves from harm. So in that sense, while it is unfortunate to read about your brother, at the same time it is encouraging in a strange way that the one profession that you'd think would use the legal system is also being hurt by technology and outsourcing. But it is also discouraging in that it shows once again that those in power (and already have more money than they know how to spend) are willing to even throw "their own" over the side of the ship in the pursuit of even more money to add to their pile.
The world is becoming increasingly bizarre every day from a financial perspective. Yesterday I was reading a community bulletin board and see a handwritten "for sale" ad for some household items (plates, scarves, various knick knacks) for like $2 to $10 per item - the whole lot might be $50. That'll buy you like two days of survival, then what? Of course it's good to recycle, and I'm not saying the seller needs this money, but when you can probably go to the dollar store and buy all of those items brand new for the same or lower price, it's kinda like what's the point. Last night I figured I'll sell my old cell phone that I paid $700 for a few years back, spent over an hour wiping it, etc and then check craigslist for going rates, I'll be lucky if I can get $75 for it.
All these things kind of invoke a feeling of cognitive dissonance in me, know what I mean? To me the world just seems in many ways to be in almost a logically impossible state of equilibrium at the moment. From everyone freaking out about "peak oil" 5 years ago, suddenly we have massive oversupply of oil and commodities, and it seems plausible that it may remain that way for a very long time, if not indefinitely - which again, could be good or bad depending on your lot in life.
It's much more profitable to replace high wage labor than low skill labor so it makes sense that this segment of the workforce would be replaced sooner than later. I never considered the angle of the additional political power this group has which may end up stalling that process. Excellent point.
Maybe not a great example: dentistry. While it is surely not "simple", I don't think it is nearly as complex as something like medicine, and it is fundamentally important basic need in everyone's life. But is also extremely expensive, and is often not covered by universal health even in countries where medicine is.
So, from time to time you'll hear, typically in ethnic communities in western countries, of non-certified people practicing dentistry off the books. I have no way of knowing what skill level they tend to operate at, but I'd be surprised if it wasn't somewhat equivalent to where mainstream dentistry was at say 20 years ago - so certainly not the best, but far better than nothing. But when it leaks out this is happening, they are shut down immediately.
As a comparison, auto mechanics is also quite complicated, but if you literally can't afford to have it dealer serviced, at least you have less-better but affordable alternatives. Not so with dentistry, because they have a powerful industry lobby that sees that any competition is formally illegal, and any freelancers have the full force of the law brought down to stop their practice. The argument for this I imagine would be "public safety", yet auto mechanics also has a public safety aspect, a much larger one I would argue, yet I can buy tools and work on my own car in my garage without breaking any laws.
And so it will go as technology progresses, industries and professions with government connections will often be untouchable.
The older I get, the more I see a similar pattern recurring everywhere (but in different forms), such as the "Why is rent so high and pay so low" article on the front page of HN today: https://news.ycombinator.com/item?id=10024681
Professional organizations, especially in medicine and law, are a disgusting selfish cabal enriching themselves at the expense of the sick and poor who are then priced out of their services.
Fortunately these lobbies are only national. You can always go to Vietnam / India / Whatever and get your non-emergency medical and dental care done there. I got my wisdom teeth out for $30 in Saigon; the only difference (besides $300 in price) was the tools were a little older and I was in an open-floor cubicle instead of a closed office.
Very good comparison!! I just pulled my gf's attention away from her ipad to read this to her. My only fear is that rather than loosen up the practice of dentistry, the government will begin tighter licensing on hand tool usage.
we see this all the time in our industry. CI/CD has replaced a lot of the functions of "Release Engineering", but not all of them, and these folks have moved up the the value chain to solve other issues.
Microsoft Azure has been expanding its analytics capabilities into this space but more for general Fortune 500 enterprise needs rather than specific verticals.
I can see the two giants on a collision course in the future.
I recently had the privilege of meeting one of the engineers on the Watson project. From what he has shared, I believe that it will be challenging to replace radiologists with this technology. Sure, Watson can be trained to identify medical imagery, but that would be the only thing its training model would be capable of. I believe human radiologists will still be superior.
Huh. I don't have a source handy, but I recall hearing about field trials that showed Watson used as a diagnostic assistant was able to identify some hard to identify medical problems, where Watson's human counterparts were unable.
What does it mean to "bring" medical images to Watson? Does Watson actually do anything, or is it a name for a bunch of services that IBM relies on other companies to construct?
Anyway, if you ever used Watson on their Blue-thing you know how horrible the experience can be. Just pay someone to make some documentation or something and an API..
Shudder. As a young man, I used to inspect new cars for shipping damage. It was well known in the business that an inspector's effectiveness dropped off dramatically as the day progressed. Missed damage could cost our customer thousands of dollars, and yet even blatantly obvious stuff was missed from time to time. I once missed frame damage on a pickup truck that had been chained down a little too enthusiastically. The 45k truck was a write off.
Our inspections numbered in the "hundreds per day". I can't even begin to imagine the mental haze at the end of thousands.
I don't see Watson getting tired.