Loss Function Influence on Uncertainty Estimation for White Matter Lesions 3D Segmentation in a Shifted Domain Setting
Data
2023
Autorzy
Tytuł czasopisma
ISSN czasopisma
Tytuł tomu
Wydawca
Wydawnictwo Politechniki Łódzkiej
Lodz University of Technology Press
Lodz University of Technology Press
Abstrakt
The aim of this study is to address the problem of distributional
shift for white matter Multiple Sclerosis lesion segmentation models. The
impact of loss function on models performance and uncertainty estimation
is evaluated. The evaluation is performed on two in-domain and one out-ofdomain
dataset consisting of 3D FLAIR Magnetic Resonance images. Our
experiments show that application of segmentation losses (eg. Dice) translate
into reduced models robustness and poorer uncertainty estimation compared
with classification losses (eg. CE). The source code is publicly available.
Opis
Słowa kluczowe
white matter multiple sclerosis lesions, multiple sclerosis, 3D segmentation, magnetic resonance imaging, zmiany w istocie białej w stwardnieniu rozsianym, stwardnienie rozsiane, segmentacja 3D, rezonans magnetyczny
Cytowanie
Kaczmarska M., Majek K., Loss Function Influence on Uncertainty Estimation for White Matter Lesions 3D Segmentation in a Shifted Domain Setting. W: Progress in Polish Artificial Intelligence Research 4, Wojciechowski A. (Ed.), Lipiński P. (Ed.)., Seria: Monografie Politechniki Łódzkiej Nr. 2437, Wydawnictwo Politechniki Łódzkiej, Łódź 2023, s. 245-250, ISBN 978-83-66741-92-8, doi: 10.34658/9788366741928.39.