This is a portable, low-energy MRI machine. It costs about $20,000 USD compared to millions of dollars for conventional MRI machines which can take up the space of an entire room. One of the tradeoffs is that the images are low resolution compared to their larger brethren, but a person familiar with the kinds of infections being sought should have enough information to make a diagnosis.
The team is experimenting with running the low-res image through an AI program to up-res the scan based on a boatload of training data. Upon hearing this, the following conversation took place:
Me: “You know there’s a risk of the AI hallucinating, resulting with false diagnoses that could lead to unnecessary death, right?”
Dr. Schiff: “Yes, that is certainly a concern.”
Me: “Since you’re writing your own software, why not have the AI program scan its training set and flag any new scan that doesn’t match, saying “Insufficient information to make a diagnosis. Please try a more conventional method.”? That should cut the hallucinations to zero.”
About 10 seconds silence occurred before Dr. Schiff gave his answer: “We hadn’t thought of that!”
My work here is done. 🙂
(C) Gary L. Friedman. This image is available for licensing. Contact
info@FriedmanArchives.com