Impact of Imaging Acquisition and Protocol Variability on Artificial Intelligence Model Performance: A Secondary Analysis of the ASFNR Artificial Intelligence Competition
This secondary analysis demonstrates that imaging acquisition and protocol variability may significantly affect AI model performance. Scanner manufacturer, slice thickness, and the developed AI model were significantly associated with model accuracy, whereas scan mode had no significant impact. Among these, the developed AI model consistently proved most influential, reflecting the importance of training data, model architecture, and preprocessing methods.
Authors
- Daniel P. Barboriak (ORCID: https://orcid.org/0000-0002-2536-6266)
- Adam E. Flanders (ORCID: https://orcid.org/0000-0002-4679-0787)
- Sahil Nalawade (ORCID: https://orcid.org/0000-0002-5440-8357)
- Jason W. Allen (ORCID: https://orcid.org/0000-0002-5274-1105)
- Micah Etter
- H. Sair
- Christopher G. Filippi (ORCID: https://orcid.org/0000-0002-4565-2614)
- Joseph A. Maldjian (ORCID: https://orcid.org/0000-0002-6384-1072)
- Max Wintermark (ORCID: https://orcid.org/0000-0002-6726-3951)
- Kirk M. Welker
- Yvonne W. Lui
- Patrik Michel
- Ruchir Chaudhari
- Burak Berksu Ozkara
- Christopher Hess
- Bin Jiang
- Ryan Godwin
- Christopher T. Whitlow
- Syed Hashmi
- Guangming Zhu
- Kevin Hsu
- Anjeza Chukus
- Hui Chen
- Greg Zaharchuk
Institutions
- Indiana University – Purdue University Indianapolis (US)
Publication Details
- Journal
- American Journal of Neuroradiology
- Published
- 2026-02-27
- DOI
- https://doi.org/10.3174/ajnr.a9266
- Primary Topic
- Adversarial Robustness in Machine Learning
- Type
- article
- Field-Weighted Citation Impact
- 0.00