19. Designing an FDA Validation Plan Exercise Solution

Designing an FDA Validation Plan Exercise Solution

ND320 C2 L4 15 Designing An FDA Validation Plan Exercise Solution Video

Summary

For both algorithms, I would want to collect a validation set that was made up of screening mammography studies only for women between the ages of 40 and 80.

To validate algorithm A, however, I would want to make sure that there were no implants, and I would also want to make sure that the distribution of breast densities A, B, C, and D was reflective of the distribution of those densities that are seen in the real world.

To validate algorithm B, I would want to make sure that my validation data set did not contain any women who had a prior history of breast cancer.

The silver standard approach of using several radiologists would be more optimal for Algorithm A because I gave you the hint that it’s really hard for radiologists to agree on breast density labels.

For Algorithm B, a single radiologist’s labels would probably suffice, because I gave you the hint that they’re really good at labeling ‘normal’ v. ‘abnormal.’