CerebNet: A fast and reliable deep-learning pipeline for detailed cerebellum sub-segmentation
Jennifer Faber,
David Kügler,
Emad Bahrami,
Lea-Sophie Heinz,
Dagmar Timmann,
Thomas M. Ernst,
Katerina Deike-Hofmann,
Thomas Klockgether,
Bart van de Warrenburg,
Judith van Gaalen,
Kathrin Reetz,
Sandro Romanzetti,
Gulin Oz,
James M. Joers,
Jorn Diedrichsen,
Martin Reuter,
Paola Giunti,
Hector Garcia-Moreno,
Heike Jacobi,
Johann Jende,
Jeroen de Vries,
Michal Povazan,
Peter B. Barker,
Katherina Marie Steiner,
Janna Krahe
Affiliations
Jennifer Faber
German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany; Department of Neurology, University Hospital Bonn, Germany
David Kügler
German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany
Emad Bahrami
German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany; Computer Science Department, University Bonn, Bonn, Germany
Lea-Sophie Heinz
German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany
Dagmar Timmann
Department of Neurology, Center for Translational Neuro, and Behavioral Sciences (C-TNBS), University Hospital Essen, University of Duisburg-Essen, Essen, Germany
Thomas M. Ernst
Department of Neurology, Center for Translational Neuro, and Behavioral Sciences (C-TNBS), University Hospital Essen, University of Duisburg-Essen, Essen, Germany
Katerina Deike-Hofmann
Department of Neuroradiology, University Hospital Bonn, Germany
Thomas Klockgether
German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany; Department of Neurology, University Hospital Bonn, Germany
Bart van de Warrenburg
Department of Neurology, Donders Institute for Brain, Cognition, and Behaviour, Radboud university medical center, Nijmegen, The Netherlands
Judith van Gaalen
Department of Neurology, Donders Institute for Brain, Cognition, and Behaviour, Radboud university medical center, Nijmegen, The Netherlands
Kathrin Reetz
Department of Neurology, RWTH Aachen University, Germany; JARA-Brain Institute Molecular Neuroscience and Neuroimaging, Forschungszentrum Jülich, Germany
Sandro Romanzetti
Department of Neurology, RWTH Aachen University, Germany
Gulin Oz
Center for Magnetic Resonance Research, Department of Radiology, University of Minnesota, Minneapolis, MN, USA
James M. Joers
Center for Magnetic Resonance Research, Department of Radiology, University of Minnesota, Minneapolis, MN, USA
Jorn Diedrichsen
Departments of Computer Science and Statistical and Actuarial Sciences, Western University, London, ON, Canada
Martin Reuter
Corresponding author.; German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany; A.A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA, USA; Department of Radiology, Harvard Medical School, Boston, MA, USA
Paola Giunti
Ataxia Centre, Department of Clinical and Movement Neurosciences, UCL Queen Square Institute of Neurology & National Hospital for Neurology and Neurosurgery, University College London Hospitals NHS Foundation Trust, London, UK
Hector Garcia-Moreno
Ataxia Centre, Department of Clinical and Movement Neurosciences, UCL Queen Square Institute of Neurology & National Hospital for Neurology and Neurosurgery, University College London Hospitals NHS Foundation Trust, London, UK
Heike Jacobi
Department of Neurology, University Hospital of Heidelberg, Heidelberg, Germany
Johann Jende
Department of Neuroradiology, Heidelberg University Hospital, Heidelberg, Germany
Jeroen de Vries
Department of Neurology, Expertise Center Movement Disorders Groningen, University Medical Center Groningen, University of Groningen, The Netherlands
Michal Povazan
Johns Hopkins University School of Medicine, Baltimore, MD, U.S.
Peter B. Barker
Johns Hopkins University School of Medicine, Baltimore, MD, U.S.
Katherina Marie Steiner
Department of Neurology, Center for Translational Neuro, and Behavioral Sciences (C-TNBS), University Hospital Essen, University of Duisburg-Essen, Essen, Germany
Janna Krahe
Department of Neurology, RWTH Aachen University, Germany
Quantifying the volume of the cerebellum and its lobes is of profound interest in various neurodegenerative and acquired diseases. Especially for the most common spinocerebellar ataxias (SCA), for which the first antisense oligonculeotide-base gene silencing trial has recently started, there is an urgent need for quantitative, sensitive imaging markers at pre-symptomatic stages for stratification and treatment assessment. This work introduces CerebNet, a fully automated, extensively validated, deep learning method for the lobular segmentation of the cerebellum, including the separation of gray and white matter. For training, validation, and testing, T1-weighted images from 30 participants were manually annotated into cerebellar lobules and vermal sub-segments, as well as cerebellar white matter. CerebNet combines FastSurferCNN, a UNet-based 2.5D segmentation network, with extensive data augmentation, e.g. realistic non-linear deformations to increase the anatomical variety, eliminating additional preprocessing steps, such as spatial normalization or bias field correction. CerebNet demonstrates a high accuracy (on average 0.87 Dice and 1.742mm Robust Hausdorff Distance across all structures) outperforming state-of-the-art approaches. Furthermore, it shows high test-retest reliability (average ICC >0.97 on OASIS and Kirby) as well as high sensitivity to disease effects, including the pre-ataxic stage of spinocerebellar ataxia type 3 (SCA3). CerebNet is compatible with FreeSurfer and FastSurfer and can analyze a 3D volume within seconds on a consumer GPU in an end-to-end fashion, thus providing an efficient and validated solution for assessing cerebellum sub-structure volumes. We make CerebNet available as source-code (https://github.com/Deep-MI/FastSurfer).
