If you're developing AI to screen for diseases in health care, it would probably be better to think analog instead of digital.
Instead of making the AI respond "yes" or "no", have it rate the likelihood of having the disease, on a scale such as 1-5.
If your AI only responds "yes" or "no", it would probably be useless because of false positives and negatives.
With using a scale, however, the screening could actually prove useful. For example, after running clinical trials, it might be found that:
1. Ratings of 1 don't need follow-up.
2. Ratings of 2-3 require follow-up, but not urgently.
3. Ratings of 4-5 should be a priority follow-up.
© 2024 Praveen Puri
