AI Improves Multi-Frequency MR Elastography

The article “Multifrequency Neural Network-based Wave Inversion in MR Elastography with Uncertainty Quantification” is part of the MICCAI 2025 Special Issue of the journal Medical Image Analysis. This special issue features extended versions of selected papers from MICCAI 2025.

AI Supports the Analysis of Tissue Stiffness

Magnetic Resonance Elastography (MRE) is a non-invasive imaging technique that can be used to investigate the mechanical properties—and in particular the stiffness—of tissue. To do this, shear waves generated by vibrations are detected using MRI and then mathematically analyzed.

The new study further develops this approach using a neural network. The developed method, MF-ElastoNet, processes measurement data from multiple excitation frequencies directly together, rather than first evaluating the individual frequencies separately. This is intended to make better use of the information from the different frequencies and enable the characterization of tissue properties over a broader frequency range.

Another distinctive feature is the quantification of the uncertainty in the predictions. The neural network thus not only provides an estimate of tissue stiffness but also offers information on how reliable the respective prediction is. The method was investigated in simulations, on phantoms, and using MRE data from the liver, spleen, and kidneys of 37 healthy subjects. Measurements with excitation frequencies between 20 and 100 Hz were evaluated.

The results show that MF-ElastoNet can achieve higher accuracy in reconstructing tissue stiffness compared to approaches that evaluate individual frequencies separately. Furthermore, the stability analysis showed that as few as three independent frequencies may be sufficient to achieve consistent results. The authors view this as another step toward reliable AI-based analysis of multifrequency MR elastography.

This publication builds on the previous work “ElastoNet: Neural network-based multicomponent MR elastography wave inversion with uncertainty quantification”, which was also published in Medical Image Analysis.

To the journal article:
Multifrequency Neural Network-based Wave Inversion in MR Elastography with Uncertainty Quantification

To the MICCAI 2025 paper:
Multifrequency Neural Network-based Wave Inversion in MR Elastography