Uncertainty-aware self-supervised neural network for liverT1ρmapping with relaxation constraint

An uncertainty aided framework for learning based liverT1ρmapping and analysis

Objective. QuantitativeT1ρimaging has potential for assessment of biochemical alterations of liver pathologies. Deep learning methods have been employed to accelerate quantitativeT1ρimaging. To employ artificial intelligence-based quantitative imaging methods in complicated clinical environment, it is valuable to estimate the uncertainty of the predicatedT1ρvalues to provide the confidence level of the quantification results. The uncertainty should also be utilized to aid the post-hoc quantitative analysis and model learning tasks.Approach. To address this need, we propose a parametric map refinement approach for learning-basedT1ρmapping and train the model in a probabilistic way to model the uncertainty. We also propose to utilize the uncertainty map to spatially weight the training of an improvedT1ρmapping network to further improve the mapping performance and to remove pixels with unreliableT1ρvalues in the region of interest. The framework was tested on a dataset of 51 patients with different liver fibrosis stages.Main results. Our results indicate that the learning-based map refinement method leads to a relative mapping error of less than 3% and provides uncertainty estimation simultaneously. The estimated uncertainty reflects the actual error level, and it can be used to further reduce relativeT1ρmapping error to 2.60% as well as removing unreliable pixels in the region of interest effectively.Significance. Our studies demonstrate the proposed approach has potential to provide a learning-based quantitative MRI system for trustworthyT1ρmapping of the liver.

Uncertainty-weighted Multi-tasking for T1ρ and T2 Mapping in the Liver with Self-supervised Learning

Multi-parametric mapping of MRI relaxations in liver has the potential of revealing pathological information of the liver. A self-supervised learning based multi-parametric mapping method is proposed to map T1ρ and T2 simultaneously, by utilising the relaxation constraint in the learning process. Data noise of different mapping tasks is utilised to make the model uncertainty-aware, which adaptively weight different mapping tasks during learning. The method was examined on a dataset of 51 patients with non-alcoholic fatter liver disease. Results showed that the proposed method can produce comparable parametric maps to the traditional multi-contrast pixel wise fitting method, with a reduced number of images and less computation time. The uncertainty weighting also improves the model performance. It has the potential of accelerating MRI quantitative imaging.