Classification of multiparametric correlation MRI signals using deep neural networks

Sebastian Endt, Tobias Lachermeier, Marion Menzel

Keywords: Magnetic Resonance Imaging, MRI, multiparametric correlation imaging, multi-component relaxometry, classification, microstructure

Abstract: Correlation MRI is a promising microstructure imaging technique, but its reconstruction remains highly ill-conditioned. We propose to first classify correlation signals, and achieve high accuracy for classification into single- vs. multi-component. Further we establish a solid baseline for predicting the exact number of sub-compartments.

Submission Number: 90