Translating AI Algorithms for Clinical Settings with the FDA

Back to Home

01. Translating Algorithm into a Clinical Setting with the FDA
02. Lesson Outline
03. How medical imaging fits in clinical settings
04. Intended Use and Indications for Use
05. Intended Use and Clinical Impact Quizzes
06. Intended Use and Clinical Impact Exercise
07. Intended Use and Clinical Impact Exercise Solution
08. Algorithmic Limitations
09. Algorithmic Limitations Quizzes
10. Algorithmic Limitations Exercise
11. Algorithmic Limitations Exercise Solution
12. Translate Performance into Clinical Utility
13. Translate Performance into Clinical Utility Quizzes
14. Translate Performance into Clinical Utility Exercise
15. Translate Performance into Clinical Utility Solution
16. Designing an FDA Validation Plan
17. Designing an FDA Validation Plan Quizzes
18. Designing an FDA Validation Plan Exercise
19. Designing an FDA Validation Plan Exercise Solution
20. Final Review
21. Lesson Conclusion

Back to Home

17. Designing an FDA Validation Plan Quizzes

Designing an FDA Validation Plan Quizzes

Validation Dataset

How many times is your algorithm run on your clinically-obtained FDA Validation Dataset?

Once

At the end of every training epoch

This dataset is used for algorithm training

Your algorithm is never run on this dataset

SOLUTION:

Once

Performance Standard

Given the following radiologist F1 scores that you obtained from the literature, what would be a reasonable F1 score to hold your algorithm to on the Validation Dataset for submitting to the FDA?

Study 1: 0.35
Study 2: 0.50
Study 3: 0.45

0.35

0.43

0.6

0.25

SOLUTION:

0.43

【udacity2.0 】If you need more courses, please add wechat：udacity6