While much media hype has centered on the use of AI as a text-generative tool, there are many uses that are being tested and applied for other very specific purposes.
One interesting example comes from the Netherlands where doctors specializing in brain cancer are testing a tool called Sturgeon.
While operating, cancer surgeons are often faced with the difficult task of deciding how much tissue to remove.
In other parts of the body, this is less serious, but when talking about the brain every cell counts.
Finding the edge of a tumor and determining how many surrounding cells are affected is often based on guesswork while the patient is on the operating table.
Definitive DNA sequencing can provide clarity for these guesses, but that process can be too time-consuming to be relevant, often taking weeks for the samples to be classified.
This results in doctors being overly cautious and removing brain tissue that is actually healthy.
Sturgeon utilizes machine learning to speed up the sequencing of cells in small areas to provide doctors with a much clearer picture of the border of the diseased tissue.
The tool was able to recognize 45 out of 50 cases within 40 minutes which gave the doctors a much better chance of correctly treating the tumors.
Harvard University is also testing a system it calls CHARM that has been trained to recognize the most aggressive types of brain cancer.
That system was trained by ingesting data on over 2000 cases of different types of tumors.
In early testing, CHARM has correctly predicted cases 93% of the time.
The process for both of these systems was likened to image recognition based on a single pixel, and in a use case where every minute counts, the high accuracy of these models is very promising.