Meeting Banner
Abstract #3195

How to train a Deep Convolutional Neural Network for Quantitative Susceptibility Mapping (QSM)

Thomas Jochmann1, Jens Haueisen1, and Ferdinand Schweser2,3
1Department of Computer Science and Automation, Technische Universität Ilmenau, Ilmenau, Germany, 2Buffalo Neuroimaging Analysis Center, Dept. of Neurology, Jacobs School of Medicine and Biomedical Sciences, University at Buffalo, The State University of New York, Buffalo, NY, United States, 3Center for Biomedical Imaging, Clinical and Translational Science Institute, University at Buffalo, The State University of New York, Buffalo, NY, United States

Deep convolutional neural networks have recently gained popularity for solving the ill-posed dipole inversion problem in Quantitative Susceptibility Mapping (QSM). The training of the neural networks is performed with examples of χ and f that can either be obtained from physical simulations on synthetic source distributions, or through “classical” QSM methods on real data. For both choices, there is a plethora of decisions to make and parameters to set. Here we seek to present best practices regarding the modelling of synthetic source distributions and data augmentation.

This abstract and the presentation materials are available to 2020 meeting attendees and eLibrary customers only; a login is required.

Join Here