Abstract
Reliably detecting diseases using relevant biological information is crucial for real-world applicability of deep learning techniques in medical imaging. We debias deep learning models during training against unknown bias – without preprocessing/filtering the input beforehand or assuming specific knowledge about its distribution or precise nature in the dataset. Control regions are used as surrogates that carry information regarding the bias; subsequently, the classifier model extracts features, and biased intermediate features are suppressed by our custom, modular DecorreLayer. We evaluate our method on a dataset of 952 lung computed tomography scans by introducing simulated biases w.r.t. reconstruction kernel and noise level and propose including an adversarial test set in evaluations of bias reduction techniques. In a moderately sized model architecture, applying the proposed method to learn from data exhibiting a strong bias, it near-perfectly recovers the classification performance observed when training with corresponding unbiased data.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Mühlberg A, Katzmann A, Heinemann V, et al. The technome - a predictive internal calibration approach for quantitative imaging biomarker research. Sci Rep. 2020;10(1103).
Taubmann O, Berger M, Bögel M, et al. Computed tomography. Medical imaging systems. Ed. by Maier A, et al. Springer, 2018. Chap. 8:147–89.
Choe J, Lee SD, Do K, et al. Deep learning-based image conversion of CT reconstruction kernels improves radiomics reproducibility for pulmonary nodules or masses. Radiol. 2019;292 2:365–73.
Kim B, Kim H, Kim K, et al. Learning not to learn: training deep neural networks with biased data. Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit. 2019:9004–12.
Amini A, Soleimany A, Schwarting W, et al. Uncovering and mitigating algorithmic bias through learned latent structure. Proc Conf AAAI/ACM AI, Ethics, and Society. 2019:289– 95.
Fortin J, Sweeney E, Muschelli J, et al. Removing inter-subject technical variability in magnetic resonance imaging studies. Neuroimage. 2016;132:198–212.
Anderson AE, Foraker AG. Centrilobular emphysema and panlobular emphysema: two different diseases. Thorax. 1973;28:547–50.
He K, Zhang X, Ren S, et al. Deep residual learning for image recognition. Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit. 2016:770–8.
Krizhevsky A. Learning multiple layers of features from tiny images. Tech. rep. University of Toronto, 2009.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 Der/die Autor(en), exklusiv lizenziert an Springer Fachmedien Wiesbaden GmbH, ein Teil von Springer Nature
About this paper
Cite this paper
Langer, S., Taubmann, O., Denzinger, F., Maier, A., Mühlberg, A. (2023). Mitigating Unknown Bias in Deep Learning-based Assessment of CT Images DeepTechnome. In: Deserno, T.M., Handels, H., Maier, A., Maier-Hein, K., Palm, C., Tolxdorff, T. (eds) Bildverarbeitung für die Medizin 2023. BVM 2023. Informatik aktuell. Springer Vieweg, Wiesbaden. https://doi.org/10.1007/978-3-658-41657-7_38
Download citation
DOI: https://doi.org/10.1007/978-3-658-41657-7_38
Published:
Publisher Name: Springer Vieweg, Wiesbaden
Print ISBN: 978-3-658-41656-0
Online ISBN: 978-3-658-41657-7
eBook Packages: Computer Science and Engineering (German Language)