Deep Image Prior for Spatio-temporal Fluorescence Microscopy Images: DECO-DIP
Standard
Deep Image Prior for Spatio-temporal Fluorescence Microscopy Images: DECO-DIP. / Meyer, Lina; Woelk, Lena-Marie; Gee, Christine; Lohr, Christian; Arcot Kannabiran, Sukanya; Diercks, Björn-Philipp; Werner, Rene.
Bildverarbeitung für die Medizin 2024: Proceedings, German Conference on Medical Image Computing, Erlangen, March 10–12, 2024. Hrsg. / A Maier; TM Deserno; H Handels; K Maier-Hein; C Palm; TM Tolxdorff. 1. Aufl. Springer Vieweg, 2024. S. 322-327 (Informatik Aktuell).Publikationen: SCORING: Beitrag in Buch/Sammelwerk › SCORING: Beitrag in Sammelwerk › Forschung › Begutachtung
Harvard
APA
Vancouver
Bibtex
}
RIS
TY - CHAP
T1 - Deep Image Prior for Spatio-temporal Fluorescence Microscopy Images: DECO-DIP
AU - Meyer, Lina
AU - Woelk, Lena-Marie
AU - Gee, Christine
AU - Lohr, Christian
AU - Arcot Kannabiran, Sukanya
AU - Diercks, Björn-Philipp
AU - Werner, Rene
PY - 2024
Y1 - 2024
N2 - Image deconvolution and denoising is a common postprocessing stepto improve the quality of biomedical fluorescence microscopy images. In recentyears, this task has been increasingly tackled with the help of supervised deeplearning methods. However, generating a large number of training pairs is, if at all possible, often laborious. Here, we present a new deep learning algorithm called DECO-DIP that builds on the Deep Image Prior (DIP) framework and does not rely on training data. We extend DIP by incorporating a novel loss function that, in addition to a standard 퐿2 data term, contains a term to model the underlying image generation forward model. We apply our framework both to synthetic data and Ca2+ microscopy data of biological samples, namely Jurkat T-cells and astrocytes. DECO-DIP outperforms both classical deconvolution and the standard DIP implementation. We further introduce an extension, DECO-DIP-T, which explicitly utilizes the time dependence in live cell microscopy image series.
AB - Image deconvolution and denoising is a common postprocessing stepto improve the quality of biomedical fluorescence microscopy images. In recentyears, this task has been increasingly tackled with the help of supervised deeplearning methods. However, generating a large number of training pairs is, if at all possible, often laborious. Here, we present a new deep learning algorithm called DECO-DIP that builds on the Deep Image Prior (DIP) framework and does not rely on training data. We extend DIP by incorporating a novel loss function that, in addition to a standard 퐿2 data term, contains a term to model the underlying image generation forward model. We apply our framework both to synthetic data and Ca2+ microscopy data of biological samples, namely Jurkat T-cells and astrocytes. DECO-DIP outperforms both classical deconvolution and the standard DIP implementation. We further introduce an extension, DECO-DIP-T, which explicitly utilizes the time dependence in live cell microscopy image series.
U2 - 10.1007/978-3-658-44037-4_82
DO - 10.1007/978-3-658-44037-4_82
M3 - SCORING: Contribution to collected editions/anthologies
SN - 978-3-658-44036-7
T3 - Informatik Aktuell
SP - 322
EP - 327
BT - Bildverarbeitung für die Medizin 2024
A2 - Maier, A
A2 - Deserno, TM
A2 - Handels, H
A2 - Maier-Hein, K
A2 - Palm, C
A2 - Tolxdorff, TM
PB - Springer Vieweg
ER -
