Skip to main content

Co-occurring diseases heavily influence the performance of weakly supervised learning models for classification of chest CT

Publication ,  Conference
Tushar, FI; Danniballe, VM; Rubin, GD; Samei, E; Lo, JY
Published in: Progress in Biomedical Optics and Imaging - Proceedings of SPIE
January 1, 2022

Despite the potential of weakly supervised learning to automatically annotate massive amounts of data, little is known about its limitations for use in computer-aided diagnosis (CAD). For CT specifically, interpreting the performance of CAD algorithms can be challenging given the large number of co-occurring diseases. This paper examines the effect of co-occurring diseases when training classification models by weakly supervised learning, specifically by comparing multi-label and multiple binary classifiers using the same training data. Our results demonstrated that the binary model outperformed the multi-label classification in every disease category in terms of AUC. However, this performance was heavily influenced by co-occurring diseases in the binary model, suggesting it did not always learn the correct appearance of the specific disease. For example, binary classification of lung nodules resulted in an AUC of < 0.65 when there were no other co-occurring diseases, but when lung nodules cooccurred with emphysema, the performance reached AUC < 0.80. We hope this paper revealed the complexity of interpreting disease classification performance in weakly supervised models and will encourage researchers to examine the effect of co-occurring diseases on classification performance in the future.

Duke Scholars

Altmetric Attention Stats
Dimensions Citation Stats

Published In

Progress in Biomedical Optics and Imaging - Proceedings of SPIE

DOI

ISSN

1605-7422

Publication Date

January 1, 2022

Volume

12033
 

Citation

APA
Chicago
ICMJE
MLA
NLM
Tushar, F. I., Danniballe, V. M., Rubin, G. D., Samei, E., & Lo, J. Y. (2022). Co-occurring diseases heavily influence the performance of weakly supervised learning models for classification of chest CT. In Progress in Biomedical Optics and Imaging - Proceedings of SPIE (Vol. 12033). https://doi.org/10.1117/12.2612700
Tushar, F. I., V. M. Danniballe, G. D. Rubin, E. Samei, and J. Y. Lo. “Co-occurring diseases heavily influence the performance of weakly supervised learning models for classification of chest CT.” In Progress in Biomedical Optics and Imaging - Proceedings of SPIE, Vol. 12033, 2022. https://doi.org/10.1117/12.2612700.
Tushar FI, Danniballe VM, Rubin GD, Samei E, Lo JY. Co-occurring diseases heavily influence the performance of weakly supervised learning models for classification of chest CT. In: Progress in Biomedical Optics and Imaging - Proceedings of SPIE. 2022.
Tushar, F. I., et al. “Co-occurring diseases heavily influence the performance of weakly supervised learning models for classification of chest CT.” Progress in Biomedical Optics and Imaging - Proceedings of SPIE, vol. 12033, 2022. Scopus, doi:10.1117/12.2612700.
Tushar FI, Danniballe VM, Rubin GD, Samei E, Lo JY. Co-occurring diseases heavily influence the performance of weakly supervised learning models for classification of chest CT. Progress in Biomedical Optics and Imaging - Proceedings of SPIE. 2022.

Published In

Progress in Biomedical Optics and Imaging - Proceedings of SPIE

DOI

ISSN

1605-7422

Publication Date

January 1, 2022

Volume

12033