Abstract
The goal of this study is to exploit feedback provided by experts that daily interact with a deep learning based segmentation tool to efficiently improve its segmentation performance. A convolutional neural network (CNN) for segmenting organs at risk for head & neck cancer was implemented in clinical practice and its predicted contours were verified, corrected and approved by the radiation oncologists before being used for treatment planning. A second CNN was subsequently trained to predict how the original contours as created by the first CNN should be adapted according to the experts. Three artificial datasets were created from the clinical dataset by introducing different amounts of systematic errors in the original predictions, in addition to the clinically corrected errors. Dice score for brainstem improved from 88% to 90.5% on average for the dataset that was the least adapted, and from 66% to 89.3% on average for the dataset in which the most systematic errors were introduced. For the clinical dataset, final segmentation improved for glottic area and supraglottic larynx compared to the initial predictions. In general, the systematic errors introduced in the contours are easier to learn compared to the clinical corrections by the expert, which are more subtle and subject to observer variability in our clinical dataset.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Brouwer, C.L., et al.: Ct-based delineation of organs at risk in the head and neck region: DAHANCA, EORTC, GORTEC, HKNPCSG, NCIC CTG, NCRI, NRG oncology and TROG consensus guidelines. Radiother. Oncol. 117(1), 83–90 (2015)
Brouwer, C.L., et al.: 3D variation in delineation of head and neck organs at risk. Radiat. Oncol. 7(1), 1–10 (2012)
Cardenas, C.E., et al.: Deep learning algorithm for auto-delineation of high-risk oropharyngeal clinical target volumes with built-in dice similarity coefficient parameter optimization function. Int. J. Radiat. Oncol. *Biol.* Phys. 101(2), 468–478 (2018)
Commowick, O., Grégoire, V., Malandain, G.: Atlas-based delineation of lymph node levels in head and neck computed tomography images. Radiother. Oncol. 87(2), 281–289 (2008)
Daisne, J.F., Blumhofer, A.: Atlas-based automatic segmentation of head and neck organs at risk and nodal target volumes: a clinical validation. Radiat. Oncol. 8(1), 1–11 (2013)
van Dijk, L.V., et al.: Improving automatic delineation for head and neck organs at risk by deep learning contouring. Radiother. Oncol. 142, 115–123 (2020)
Ibragimov, B., Xing, L.: Segmentation of organs-at-risks in head and neck CT images using convolutional neural networks. Med. Phys. 44(2), 547–557 (2017)
Men, K., et al.: Fully automatic and robust segmentation of the clinical target volume for radiotherapy of breast cancer using big data and deep learning. Physica Med. 50, 13–19 (2018)
Mnih, V., et al.: Playing atari with deep reinforcement learning. arXiv preprint arXiv:1312.5602 (2013)
Mukesh, M., et al.: Interobserver variation in clinical target volume and organs at risk segmentation in post-parotidectomy radiotherapy: can segmentation protocols help? Br. J. Radiol. 85(1016), e530–e536 (2012)
Nguyen, T.T., Nguyen, N.D., Nahavandi, S.: Deep reinforcement learning for multiagent systems: a review of challenges, solutions, and applications. IEEE Trans. Cybern. 50(9), 3826–3839 (2020)
Nikolov, S., et al.: Deep learning to achieve clinically applicable segmentation of head and neck anatomy for radiotherapy. arXiv preprint arXiv:1809.04430 (2018)
Palma, D., et al.: Volumetric modulated arc therapy for delivery of prostate radiotherapy: comparison with intensity-modulated radiotherapy and three-dimensional conformal radiotherapy. Int. J. Radiat. Oncol. *Biol.* Phys. 72(4), 996–1001 (2008)
Piotrowski, T., et al.: Impact of the intra-and inter-observer variability in the delineation of parotid glands on the dose calculation during head and neck helical tomotherapy. Technol. Cancer Res. Treat. 14(4), 467–474 (2015)
Ribani, R., Marengoni, M.: A survey of transfer learning for convolutional neural networks. In: 2019 32nd SIBGRAPI Conference on Graphics, Patterns and Images Tutorials (SIBGRAPI-T), pp. 47–57. IEEE (2019)
Ronneberger, O., Fischer, P., Brox, T.: U-Net: convolutional networks for biomedical image segmentation. In: Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F. (eds.) MICCAI 2015. LNCS, vol. 9351, pp. 234–241. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-24574-4_28
Taylor, M.E.: Teaching reinforcement learning with Mario: an argument and case study. In: Proceedings of the 2011 AAAI Symposium Educational Advances in Artificial Intelligence (2011)
Van der Veen, J., Willems, S., Bollen, H., Maes, F., Nuyts, S.: Deep learning for elective neck delineation: more consistent and time efficient. Radiother. Oncol. 153, 180–188 (2020)
Van der Veen, J., et al.: Benefits of deep learning for delineation of organs at risk in head and neck cancer. Radiother. Oncol. 138, 68–74 (2019)
Willems, S., et al.: Clinical implementation of DeepVoxNet for auto-delineation of organs at risk in head and neck cancer patients in radiotherapy. In: Stoyanov, D. (ed.) CARE/CLIP/OR 2.0/ISIC -2018. LNCS, vol. 11041, pp. 223–232. Springer, Cham (2018). https://doi.org/10.1007/978-3-030-01201-4_24
Wong, J., et al.: Comparing deep learning-based auto-segmentation of organs at risk and clinical target volumes to expert inter-observer variability in radiotherapy planning. Radiother. Oncol. 144, 152–158 (2020)
Acknowledgments
Siri Willems and Heleen Bollen are supported by a Ph.D. fellowship of the research foundation - Flanders (FWO).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 Springer Nature Switzerland AG
About this paper
Cite this paper
Willems, S. et al. (2021). Learning from Mistakes: An Error-Driven Mechanism to Improve Segmentation Performance Based on Expert Feedback. In: Oyarzun Laura, C., et al. Clinical Image-Based Procedures, Distributed and Collaborative Learning, Artificial Intelligence for Combating COVID-19 and Secure and Privacy-Preserving Machine Learning. DCL PPML LL-COVID19 CLIP 2021 2021 2021 2021. Lecture Notes in Computer Science(), vol 12969. Springer, Cham. https://doi.org/10.1007/978-3-030-90874-4_7
Download citation
DOI: https://doi.org/10.1007/978-3-030-90874-4_7
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-90873-7
Online ISBN: 978-3-030-90874-4
eBook Packages: Computer ScienceComputer Science (R0)