PISCO: Self-supervised k-space regularization for improved neural implicit k-space representations of dynamic MRI

PISCO: Self-supervised k-space regularization for improved neural implicit k-space representations of dynamic MRI.

Veronika Spieker¹, Hannah Eichhorn², Wenqi Huang³, Jonathan K Stelter⁴, Tabita Catalan⁵, Rickmer F Braren⁶, Daniel Rueckert⁷, Francisco Sahli Costabal⁸, Kerstin Hammernik³, Dimitrios C Karampinos⁹, Claudia Prieto¹⁰, Julia A Schnabel¹¹

¹Institute of Machine Learning in Biomedical Imaging, Helmholtz Munich, Munich, Germany; School of Computation, Information and Technology, Technical University of Munich (TUM), Munich, Germany; Millenium Institute for Intelligent Healthcare Engineering, Santiago, Chile. Electronic address: v.spieker@tum.de.

²Institute of Machine Learning in Biomedical Imaging, Helmholtz Munich, Munich, Germany; School of Computation, Information and Technology, Technical University of Munich (TUM), Munich, Germany.

³School of Computation, Information and Technology, Technical University of Munich (TUM), Munich, Germany.

⁴School of Medicine and Health, Klinikum rechts der Isar, Technical University of Munich (TUM), Munich, Germany.

⁵Millenium Institute for Intelligent Healthcare Engineering, Santiago, Chile; School of Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile.

⁶School of Medicine and Health, Klinikum rechts der Isar, Technical University of Munich (TUM), Munich, Germany; Department of Diagnostic and Interventional Radiology and Department of Nuclear Medicine, University Medical Center Hamburg Eppendorf, Hamburg, Germany.

⁷School of Computation, Information and Technology, Technical University of Munich (TUM), Munich, Germany; School of Medicine and Health, Klinikum rechts der Isar, Technical University of Munich (TUM), Munich, Germany; Department of Computing, Imperial College London, London, UK; Munich Center for Machine Learning (MCML), Munich, Germany.

⁸Millenium Institute for Intelligent Healthcare Engineering, Santiago, Chile; School of Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile; Institute for Biological and Medical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile.

⁹School of Medicine and Health, Klinikum rechts der Isar, Technical University of Munich (TUM), Munich, Germany; Laboratory of Magnetic Resonance Imaging Systems and Methods, Ecole Polytechnique Federale de Lausanne (EPFL), Lausanne, Switzerland; Center for Biomedical Imaging (CIBM), Lausanne, Switzerland.

¹⁰Millenium Institute for Intelligent Healthcare Engineering, Santiago, Chile; School of Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile; School of Biomedical Imaging and Imaging Sciences, Kings College London, London, UK.

¹¹Institute of Machine Learning in Biomedical Imaging, Helmholtz Munich, Munich, Germany; School of Computation, Information and Technology, Technical University of Munich (TUM), Munich, Germany; School of Medicine and Health, Klinikum rechts der Isar, Technical University of Munich (TUM), Munich, Germany; Munich Center for Machine Learning (MCML), Munich, Germany; School of Biomedical Imaging and Imaging Sciences, Kings College London, London, UK.

Abstract

Neural implicit k-space representations (NIK) have shown promising results for dynamic magnetic resonance imaging (MRI) at high temporal resolutions. Yet, reducing acquisition time, and thereby available training data, results in severe performance drops due to overfitting. To address this, we introduce a novel self-supervised k-space loss function LPISCO, applicable for regularization of NIK-base…

Abstract

Create a free account to keep reading