Fig 7. EIT reconstructions from phantom data with different algorithms and the proposed method.

(a) One-step GN, (b) PDIPM after five iterations, (c) ANN as an inverse solver and (d) as the proposed post-processor (ANNs in (c) and (d) are trained without noise), (e) ANN as an inverse solver and (f) as the proposed post-processor (ANNs in (e) and (f) are trained with noise). The contours of the original targets are shown in green. The normalized resistivity distribution are shown on top.