New magnetic resonance electrical impedance tomography algorithm based on neuro-fuzzy network

A new magnetic resonance electrical impedance tomography (MREIT) algorithm was developed to image the conductivity distribution within human head. Only one component of magnetic flux densities was utilized. The target conductivities were estimated by minimizing the dissimilarity between the measured and calculated magnetic flux densities based on adaptive neuro-fuzzy inference system (ANFIS). Single-variable simulation on the sphere and realistic-geometry model estimated the skull-to-brain conductivity ratio. The relative error (RE) between the target and the estimated conductivity distribution was less than 1.10% on the sphere model and less than 0.25% on the realistic-geometry head model; when the electrode's excursion was concerned, the RE was less than 0.35% on the realistic-geometry head model. Simulation shows that ANFIS-MREIT outperforms other MREIT algorithms in estimating the head volume conductivities for piece-wise homogeneous head volume-conductor models.