RAIDER: Rapid, anatomy-independent, deep learning-based PDFF and R2* estimation using magnitude-only signals

T. J. P. Bray,G. V. Minore,A. Bainbridge,L. Dwyer-Hemmings,S. A. Taylor,M. A. Hall-Craggs,H. Zhang

2024-11-01

Abstract:Purpose: There has been substantial recent interest in magnitude-based fitting methods for estimating proton density fat fraction (PDFF) and R2* from chemical shift-encoded MRI data, since these methods can still be used when complex-based methods fail or when phase data are inaccessible or unreliable, and may also be used as a final processing step with complex-based methods. However, conventional fitting techniques are computationally expensive. Deep learning (DL)-based methods promise to accelerate parameter estimation, but previous attempts have used convolutional neural networks (CNNs), which are limited by training requirements and poor generalizability (anatomy-dependence). To address these limitations, we propose RAIDER, a voxelwise method for rapid, anatomy-independent deep learning-based PDFF and R2* estimation using multi-echo magnitude-data. Theory and Methods: RAIDER uses two multilayer perceptrons, each trained separately with simulated single-voxel multi-echo magnitude signals, to estimate PDFF and R2*. The use of two networks, each with restricted training distribution, solves the problem of degeneracy during training. During inference, the solution from one of the two networks is chosen based on likelihood. Performance and speed are investigated in a series of simulation experiments, in phantoms and in vivo. Results: RAIDER is 285-1450 times faster than conventional magnitude fitting, taking 1.4-2.3s per slice rather than 8-56minutes, and offers performance similar to conventional fitting. It produces accurate PDFF measurements in phantoms and in vivo images with different anatomies, despite having been trained only on simulation data. Conclusion: RAIDER can substantially accelerate magnitude-based PDFF and R2* estimation, whilst avoiding intrinsic limitations of CNN-based methods.

Medical Physics