Adapting Machine Learning for Medicine

One project I want to move into after we finally get Socrates up for all levels of service is to turn toward medicine. Google’s AI algorithm predicts heart disease by scanning your eyes. They have used machine learning to accomplish what a doctor would do by traditional methods including blood work.

Medicine suffers from the same problem as financial analysis. It is dominated by opinion and that depends entirely upon (1) your experience and (2) the current situation. For example, a small child of about two was running a high fever. She was taken to the doctor who was flooded with flu patients. She was checked for the flu, found to be negative, and sent home. Day three, the fever was still there. Again, the little girl was taken back to another doctor. The same thing took place. After day four, the fever persisted and once more she was taken back to a doctor and the same thing took place.

Finally, the mother noticed she was not urinating and the child simply responded that it hurt. A fourth trip ended up admitting her to a hospital with a urinary infection that had by now started to impact her kidneys.

Medicine, like analysis, is dominated by opinion. If the doctor does not think of anything else to test, nothing takes place.

Machine learning can be used with a full body scanner and much more accurate results can be obtained. We need a machine that will also do a full body ultrasound scan, a simple pinprick for a drop of blood, and the machine could come up with a lot more information at a far lower cost than what we see today.