Geometry of Deep Learning for Magnetic Resonance Fingerprinting

TL;DR

This paper introduces MRF-Net, a deep learning model that significantly reduces memory and computation needs for Magnetic Resonance Fingerprinting by approximating the dictionary-matching process through manifold clustering and hierarchical filtering.

Contribution

The paper presents a novel deep learning approach with a dimensionality reduction layer that replaces traditional dictionary-matching in MRF, achieving over 60 times efficiency improvements.

Findings

MRF-Net reduces memory and computation by over 60 times.

The network approximates the Bloch response manifold with a piece-wise affine model.

It clusters the manifold and learns hierarchical filters for efficient NMR parameter estimation.

Abstract

Current popular methods for Magnetic Resonance Fingerprint (MRF) recovery are bottlenecked by the heavy storage and computation requirements of a dictionary-matching (DM) step due to the growing size and complexity of the fingerprint dictionaries in multi-parametric quantitative MRI applications. In this paper we study a deep learning approach to address these shortcomings. Coupled with a dimensionality reduction first layer, the proposed MRF-Net is able to reconstruct quantitative maps by saving more than 60 times in memory and computations required for a DM baseline. Fine-grid manifold enumeration i.e. the MRF dictionary is only used for training the network and not during image reconstruction. We show that the MRF-Net provides a piece-wise affine approximation to the Bloch response manifold projection and that rather than memorizing the dictionary, the network efficiently clusters this manifold and learns a set of hierarchical matched-filters for affine regression of the NMR characteristics in each segment.