Extending Probabilistic U-Net Using MC-Dropout to Quantify Data and Model Uncertainty

Bhat, Ishaan; Kuijf, Hugo J.

doi:https://doi.org/10.1007/978-3-031-09002-8_48

Extending Probabilistic U-Net Using MC-Dropout to Quantify Data and Model Uncertainty

DSpace/Manakin Repository

Extending Probabilistic U-Net Using MC-Dropout to Quantify Data and Model Uncertainty

Bhat, Ishaan; Kuijf, Hugo J.

(2022)
Brainlesion, volume 12963 LNCS, pp. 555 - 559
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), volume 12963 LNCS, pp. 555 - 559
7th International Brain Lesion Workshop, BrainLes 2021, held in conjunction with the Medical Image Computing and Computer Assisted Intervention, MICCAI 2021, volume 12963 LNCS, pp. 555 - 559

(Part of book)

Abstract

We extend the Probabilistic U-Net using MC-Dropout to estimate model uncertainty in addition to the data uncertainty in order to improve the overall predictive uncertainty estimate. We use this model on the datasets present in the QUBIQ21 challenge and achieve a mean score of 0.719.

Download/Full Text

The full text of this publication is not available.

Version on publisher website for UU-students and staff

Version on publisher website

Keywords: Theoretical Computer Science, General Computer Science

DOI: https://doi.org/10.1007/978-3-031-09002-8_48

ISSN: 0302-9743

ISBN: 9783031090011

Publisher: Springer Science and Business Media Deutschland GmbH

(Peer reviewed)