Deep Learning for Magnetic Resonance Fingerprinting

Solene Girardeau, Ilkay Oksuz, Gastao Cruz, Claudia Pietro Vasquez, Andrew King, James Clough

Keywords: Deep Learning, Magnetic Resonance Fingerprinting, Quantitative MRI, Magnetic Resonance Imaging

Abstract:

Standard methods for quantitative MRI are quite time consuming whereas techniques based on deep learning have the potential to be significantly faster while also improving parameter estimation accuracy.
The presented models aim to explore the different aspects of MR data, notably the spatial and temporal correlations in and between the signal evolutions.
The models developed include purely temporal-focused and spatial-focused models as well as a model trained in both domains.
The importance of pre-selecting important features prior to training was also studied and tested.

Code Of Conduct: I have read and accept the code of conduct.