EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Towards objective and systematic evaluation of bias in artificial intelligence for medical imaging

Emma A M Stanley iD, Raissa Souza iD, Anthony J Winder, Vedant Gulve, Kimberly Amador, Matthias Wilms, Nils D Forkert iD

Journal of the American Medical Informatics Association · 01.11.2024 · 31 (11) · 2613-2621

DOI: 10.1093/jamia/ocae165 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of the American Medical Informatics Association, online erschienen am 28.06.2024. Verfasst von Emma A M Stanley und 6 weiteren Autor:innen. Beteiligte Einrichtung: Biomedical Engineering Graduate Program, University of Calgary, Calgary, Alberta, T2N 1N4, Canada. Hauptthemen laut MEDLINE: Artificial Intelligence, Bias, Neural Networks, Computer. Die Arbeit wurde bisher 29-mal zitiert (Crossref) und nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Alberta Innovates, Natural Sciences and Engineering Research Council of Canada, River Fund at Calgary Foundation, Canada Research Chairs Program, University of Calgary Department of Pediatrics.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

Artificial intelligence (AI) models trained using medical images for clinical tasks often exhibit bias in the form of subgroup performance disparities. However, since not all sources of bias in real-world medical imaging data are easily identifiable, it is challenging to comprehensively assess their impacts. In this article, we introduce an analysis framework for systematically and objectively investigating the impact of biases in medical images on AI models. Materials and Methods Our framework utilizes synthetic neuroimages with known disease effects and sources of bias. We evaluated the impact of bias effects and the efficacy of 3 bias mitigation strategies in counterfactual data scenarios on a convolutional neural network (CNN) classifier.

Results

The analysis revealed that training a CNN model on the datasets containing bias effects resulted in expected subgroup performance disparities. Moreover, reweighing was the most successful bias mitigation strategy for this setup. Finally, we demonstrated that explainable AI methods can aid in investigating the manifestation of bias in the model using this framework.

Discussion

The value of this framework is showcased in our findings on the impact of bias scenarios and efficacy of bias mitigation in a deep learning model pipeline. This systematic analysis can be easily expanded to conduct further controlled in silico trials in other investigations of bias in medical imaging AI.

Conclusion

Our novel methodology for objectively studying bias in medical imaging AI can help support the development of clinical decision-support tools that are robust and responsible.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Artificial IntelligenceBiasNeural Networks, ComputerHumansDiagnostic ImagingNeuroimaging

Stichwörter der Autor:innen

algorithmic biasartificial intelligencebias mitigationsynthetic data

Autor:innen und Institutionen

  • Biomedical Engineering Graduate Program, University of Calgary, Calgary, Alberta, T2N 1N4, Canada.
  • Department of Radiology, University of Calgary, Calgary, Alberta, T2N 4N1, Canada.
  • Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, T2N 4N1, Canada.
  • Alberta Children's Hospital Research Institute, University of Calgary, Calgary, Alberta, T2N 4N1, Canada.
  • Department of Pediatrics, University of Calgary, Calgary, Alberta, T2N 4N1, Canada.
  • Department of Community Health Sciences, University of Calgary, Calgary, Alberta, T2N 4N1, Canada.

ORCID-Profile: Emma A M Stanley, Raissa Souza, Nils D Forkert

Förderung

  • Alberta Innovates, Natural Sciences and Engineering Research Council of Canada, River Fund at Calgary Foundation
  • Canada Research Chairs Program
  • University of Calgary Department of Pediatrics
  • Alberta Children's Hospital Foundation
  • Alberta Innovates
  • Natural Sciences and Engineering Research Council of Canada
  • River Fund at Calgary Foundation, Canada Research Chairs Program

Zitieren

APA

Stanley, E. A. M., Souza, R., Winder, A. J., Gulve, V., Amador, K., Wilms, M., & Forkert, N. D. (2024). Towards objective and systematic evaluation of bias in artificial intelligence for medical imaging. Journal of the American Medical Informatics Association, 31(11), 2613-2621. https://doi.org/10.1093/jamia/ocae165

Vancouver

Stanley EAM, Souza R, Winder AJ, Gulve V, Amador K, Wilms M, et al. Towards objective and systematic evaluation of bias in artificial intelligence for medical imaging. Journal of the American Medical Informatics Association. 2024;31(11):2613-2621. doi:10.1093/jamia/ocae165

BibTeX
@article{Stanley2024,
  author = {Stanley, Emma A M and Souza, Raissa and Winder, Anthony J and Gulve, Vedant and Amador, Kimberly and Wilms, Matthias and Forkert, Nils D},
  title = {Towards objective and systematic evaluation of bias in artificial intelligence for medical imaging},
  journal = {Journal of the American Medical Informatics Association},
  year = {2024},
  volume = {31},
  number = {11},
  pages = {2613--2621},
  doi = {10.1093/jamia/ocae165}
}
Kostenlos starten →