REC-NN: A reconstruction error compensation neural network for Magnetic Resonance Electrical Property Tomography (MREPT).

Electrical properties (EPs) are expected as biomarkers for early cancer detection. Magnetic resonance electrical properties tomography (MREPT) is a technique to non-invasively estimate the EPs of tissues from MRI measurements. While noise sensitivity and artifact problems of MREPT are being solved progressively through recent efforts, the loss of tissue contrast emerges as an obstacle to the clinical applications of MREPT. To solve the problem, we propose a reconstruction error compensation neural network scheme (REC-NN) for a typical analytic MREPT method, Stab-EPT. Two NN structures: one with only ResNet blocks, and the other hybridizing ResNet blocks with an encoder-decoder structure. Results of experiments with digital brain phantoms show that, compared with Stab-EPT, and conventional NN based reconstruction, REC-NN improves both reconstruction accuracy and tissue contrast. It is found that, the encoder-decoder structure could improve the compensation accuracy of EPs in homogeneous region but showed worse reconstruction than only ResNet structure for tumorous tissues unseen in the training samples. Future research is required to address overcompensation problems, optimization of NN structure and application to clinical data.

Physics-Coupled Neural Network Magnetic Resonance Electrical Property Tomography (MREPT) for Conductivity Reconstruction.

The electrical property (EP) of human tissues is a quantitative biomarker that facilitates early diagnosis of cancerous tissues. Magnetic resonance electrical properties tomography (MREPT) is an imaging modality that reconstructs EPs by the radio-frequency field in an MRI system. MREPT reconstructs EPs by solving analytic models numerically based on Maxwell's equations. Most MREPT methods suffer from artifacts caused by inaccuracy of the hypotheses behind the models, and/or numerical errors. These artifacts can be mitigated by adding coefficients to stabilize the models, however, the selection of such coefficient has been empirical, which limit its medical application. Alternatively, end-to-end Neural networks-based MREPT (NN-MREPT) learns to reconstruct the EPs from training samples, circumventing Maxwell's equations. However, due to its pattern-matching nature, it is difficult for NN-MREPT to produce accurate reconstructions for new samples. In this work, we proposed a physics-coupled NN for MREPT (PCNN-MREPT), in which an analytic model, cr-MREPT, works with diffusion and convection coefficients, learned by NNs from the difference between the reconstructed and ground-truth EPs to reduce artifacts. With two simulated datasets, three generalization experiments in which test samples deviate gradually from the training samples, and one noise-robustness experiment were conducted. The results show that the proposed PCNN-MREPT achieves higher accuracy than two representative analytic methods. Moreover, compared with an end-to-end NN-MREPT, the proposed method attained higher accuracy in two critical generalization tests. This is an important step to practical MREPT medical diagnoses.