Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Dec 19:29:100526.
doi: 10.1016/j.phro.2023.100526. eCollection 2024 Jan.

Harnessing uncertainty in radiotherapy auto-segmentation quality assurance

Kareem A Wahid  1   2 , Jaakko Sahlsten  3 , Joel Jaskari  3 , Michael J Dohopolski  4 , Kimmo Kaski  3 , Renjie He  2 , Enrico Glerean  5 , Benjamin H Kann  6 , Antti Mäkitie  7 , Clifton D Fuller  2 , Mohamed A Naser  2 , David Fuentes  1
Affiliations

Harnessing uncertainty in radiotherapy auto-segmentation quality assurance

Kareem A Wahid et al. Phys Imaging Radiat Oncol. .
No abstract available

PubMed Disclaimer

Conflict of interest statement

Clifton D. Fuller has received unrelated direct industry grant/in-kind support, honoraria, and travel funding from Elekta AB.

Figures

Fig. 1
Fig. 1
Comparison between conventional and approximate Bayesian deep learning approaches. Conventional deep learning methods generate point estimates and are often poorly calibrated, while approximate Bayesian methods, e.g., Monte Carlo dropout and deep ensemble, generate posterior predictive distributions that are often better calibrated. The Monte Carlo dropout method consists of randomly removing nodes from the network during the training and inference procedures. The deep ensemble method trains submodels with different random initializations of network parameters and, optionally, varying subsets of training data, then combines their predictions. This figure is loosely inspired by figures from van den Berg and Meliadò .
See this image and copyright information in PMC

References

    1. Rodríguez Outeiral R., Ferreira Silvério N., González P.J., Schaake E.E., Janssen T., van der Heide U.A., et al. A network score-based metric to optimize the quality assurance of automatic radiotherapy target segmentations. Phys Imaging Radiat Oncol. 2023;28 - PMC - PubMed
    1. Guo C, Pleiss G, Sun Y, Weinberger KQ. On calibration of modern neural networks. In: Precup D, Teh YW, editors. Proceedings of the 34th international conference on machine learning, vol. 70, PMLR; 06--11 Aug 2017, p. 1321–30.
    1. Holm A.N., Wright D., Augenstein I. Revisiting softmax for uncertainty approximation in text classification. Information. 2023;14:420.
    1. Pearce T, Brintrup A, Zhu J. Understanding softmax confidence and uncertainty. arXiv [csLG] 2021.
    1. Jaskari J., Sahlsten J., Damoulas T., Knoblauch J., Särkkä S., Kärkkäinen L., et al. Uncertainty-aware deep learning methods for robust diabetic retinopathy classification. IEEE Access. 2022;10:76669–76681.

LinkOut - more resources