It's official – Google AI gives you cancer ...diagnosis in real time: Neural net can spot breast, prostate tumors Google Health's so-called augmented-reality microscope has proven surprisingly accurate at detecting and diagnosing cancerous tumors in real time. The device is essentially a standard microscope decked out with two extra components: a camera, and a computer running AI software with an Nvidia Titan Xp GPU to accelerate the …
There's an interesting example of innovation inside the NHS on inside health:
https://www.bbc.co.uk/programmes/m0006ls6
A Scottish trust developed new ways of automating tests with the equipment they already had. Invest the money in stuff like this (and hopefully money to roll it out as well).
> A Scottish trust developed new ways of automating tests with the equipment they already had. Invest the money in stuff like this (and hopefully money to roll it out as well).
I think that would require someone who knows what they're doing to be in charge, and I'm not sure that this is in the current set of current govmt policies
NHSScot is falling apart just like NHS England, waiting lists are up all over the place, its just we don't do private contractors. Shame the result is the same though....locally try 5+ month backlog on top of normal 12 week wait time to see physio therapy and its seemingly only going up and up and up
Was it included in funding to tackle waiting times? No, despite it being a useful way to get people back to work quickly / prevent condition worsening and ergo someone having to be off work due to the injury getting worse.
To me, the fact that it is real time is less important than the fact that it is accurate.
I'm less hot on doctors being able to ask the machine what it is they're looking at. Replacing doctor's knowledge with a machine's knowledge just feels wrong to me somehow. If I am to be treated, I want to be treated by a doctor, not a guy who's googling the problem.
I'm also all for well trained experienced doctors being being pointed at items of interest by clever software while carefully viewing slides for anything interesting.
My concern is that human nature tends to the easiest path and beancounters will replace doctors with 'specialists' and up the workload to the point where there's only time to agree with the softwear.
I'm less hot on doctors being able to ask the machine what it is they're looking at. Replacing doctor's knowledge with a machine's knowledge just feels wrong to me somehow. If I am to be treated, I want to be treated by a doctor, not a guy who's googling the problem.
Medicine is becoming too complex for a human to understand; some kind of system is needed to produce recommendations. For example, breast cancers in research are often classified with "PAM50", you can guess how many categories there are :)
"To me, the fact that it is real time is less important than the fact that it is accurate".
I Agree, but realize that sometimes the patient may already be in surgery under anesthesia, and the surgeon is waiting on the pathology report to determine how to proceed. In those circumstances, time may be critical.
ARM is also promising for another reason. It’s cheaper than “conventional whole-slide scanners” by about one or two magnitudes, apparently.
Although this is much cheaper than a slide scanner, they're not really a replacement. A slide scanner basically has a load of slides loaded in a rack which it then scans at a high resolution automatically; this solution has the scientist/doctor take each slide one by one and then move the slide around the "stage" and then look at the bits they think they're interested in. If you have lots of slides to look at then you'd really prefer to have them scanned automatically, you can then zoom into the bits of tissue that you're interested in like you would with something like Google Earth.
As an aside, I was involved in setting up the infrastructure for the first one of these the a research site had. We almost pooed ourselves because they thought the scanner would generate 1TB a year :-D
"The algorithms had a performance score of 0.92"
Is that like 92% accuracy as compared with human microscopists? The comparison is with sets of slides already "scored" by humans but have those scores been "corrected" by taking into consideration subsequent outcomes for the patients. i.e. if an "all clear" patient did develop cancer in the next year or two and if on further testing or surgery a "cancer positive" diagnosis was found to be incorrect.
Do we know how successful humans are at reading the slides correctly?
When the AI scan is deemed inaccurate that can be false positive of false negative. If they're largely false positive then what happens next becomes important - a further level of diagnosis:good, unnecessary surgery:bad. False negatives are more of a problem patient is wrongly given "all clear" so no further action (until/unless more evidence surfaces).
Speaking as an elderly man... They whipped my prostate out just in time (had already spread but that was excised too). If you're a bloke then the risk becomes significant as we age, usually not till at least 60 unless there's a family history. You will find that some GPs are not keen to make what should ideally become a regular check (if there are signs of a possible problem that will be followed up with other tests starting with a blood test and, where indicated, tissue samples and slide analysis). That initial check is a DRE: digital rectal examination, just a 5 minute task. This being an essentially IT forum I should point out that in this case "digital" does not imply some kind of high-tech device but involves a rubber glove and lubricant.
Side effects of radical prostatectomy are not great (ED, incontinence, developing boobs) but better than letting an aggressive cancer run its course (other treatments may be indicated depending on individual circumstances). What was never explained was that the result of removing an inch or two of the pipeline means something has to "give ground". The bladder, being firmly fixed in its habitual location, isn't going to shift so the other end has to take up the slack, the effect is not unlike that of taking a swim in freezing water. The only (somewhat questionable) benefit is that the stream, no longer impeded by the troublesome prostate, might be usefully employed as an alternative to purchasing a Karcher (Pressure Washer).
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
