Conference Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 361 - Using Statistical Foundations to Demonstrate Effectiveness of ML/AI Algorithms for Clinical Utility
Type: Invited
Date/Time: Wednesday, August 10, 2022 : 8:30 AM to 10:20 AM
Sponsor: Biopharmaceutical Section
Abstract #320354
Title: Deep Learning for Image Analysis - Initial Development of Digital Diagnostic/Prognostic Algorithms: Analogies and Lessons Learned from Drug Development
Author(s): Chong Duan*
Companies: Pfizer
Keywords: Medical Imaging; Deep Learning; Drug Development
Abstract:

While the development of digital diagnostic/prognostic algorithms (DDPA) for medical imaging has gained tremendous visibility, most DDPAs failed to be successfully deployed in clinical practice. We believe this could be improved if DDPA development followed the same evaluation process as other pharmacologic interventions through phased clinical trials. In this presentation, we will discuss analogies between the initial DDPA development and the discovery phase of drug development. Phase 0 studies assess the unmet medical need, explore intervention approaches and identify best candidates to advance to human trials. Similarly, DDPA development should assess the need for improved diagnostic/prognostic performance, investigate appropriate algorithms and data sources, and create the initial model. Phase 1 dose escalation and safety investigations are analogous to estimating model performance metrics and assessing the cost/benefit of true/false positives. We will give an example on the development of a cardiac MRI-based diagnostic algorithm for transthyretin amyloid cardiomyopathy to illustrate how these Phase 0/1 principles can be applied to improve DDPA development and implementation.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2022 program