Historically medical diagnosis has been in simplest terms largely a combination of observation, imaging and tests of blood and urine. The underlying logic of such an approach is the checklist. If certain boxes are checked you have bronchitis and one or two more boxes congestive heart failure. This type of approach does not require higher order thinking such as pattern recognition or metaphor, which is why it is used for routine things like pre-flight checks.
This type of approach has not changed in hundreds of years but the machines to gather data get better and better. No one has thought about how to use pattern recognition to do medical diagnosis. However, with advances in computing and AI, perhaps we should. I saw a pattern recognition-based medical diagnostic prototype this morning. The premise is that a particular disease has a particular color pattern when patient blood is put on a non-chemical proprietary slide. The disease pattern can only be identified given all the color permutations using pattern recognition and AI. The implications are staggering if we just consider cost and investment savings and timing and speed of diagnosis. Innovation perhaps at the scale of the transistor or the computer.
No one disputes the value of good doctors and a lot of money is invested each year on medical research and development. Perhaps equally beneficial will be the exploration of new diagnostic methods using AI in medicine.