The analysis tool, called KidneyIntelX has been developed by RenalytixAI and is expected to be launched in the middle of this year. It is described as a major break-through in this news article. It should help decide what treatment is best for a patient and whether they need medicine, dialysis or a transplant.
So how does this sort of tool work?
It's basically similar to those telephone systems where you get asked a series of questions on what help you want when calling a support helpline, with the answer you give defining what the next group of questions should be. This is how a doctor decides what is wrong with you, or how a research chemist analyses an unknown chemical by starting with a few standard tests, then choosing the next tests to use based on the results of the first texts. Eventually, the doctor or chemist reaches a conclusion, assuming they remember what the next step is after each preliminary test.
But it includes a few extra twists after that.
Initially, doctors give the software the important questions to ask and say what the answers mean, and what questions to ask next. The creators of the tool then fed into the system the medical records and answers given to questions for literally 3 million patient health records. Then the AI system kicks in and looks for common features that might not spring to mind as being related to the problem being studied. The more real cases the AI code (called a neural network) examines, the more likely the system is to spot the important trends such as do your symptoms match those of patients diagnosed as in need of dialysis by their doctor. Or the bit that humans struggle with, is there some common factor in your earlier medical history that hints about the likely outcome? Your medical history and the outcome of the treatment will be added to the system over time. as well.
This is described as the AI system learning how to diagnose and predict the best outcome. It's not got a mind of its own, it is looking for common factors between patients and what the result of certain treatments was, based on, as mentioned earlier, millions of medical records. It does also need a lot of records where the patient didn't have a kidney problem, to avoid false positives. And contrary to popular views on these systems, it does not decide what will happen to you, it advises the doctor on what to do, and the doctor makes the final decision.
It could take a doctor quite some time to read through your medical record, and to spot changes over time. But humans can miss these odd clues occasionally. This sort of software can do this very quickly and is usually very good at spotting the smaller important clues in amongst your entire medical history. So the doctor is presented with the most important data very quickly and can come to a more accurate conclusion. Freeing up more time, leading to more patients getting the treatment they need.
It's not some power-mad living machine trying to rule your life. The cyborgs from Terminator are not making the decisions. ;)
