Comments on "Data truncation artifact reduction in MR imaging using a multilayer neural network"

A recent paper by Yan and Mao (see ibid., vol.12, no.1, p.73-7, 1993) provided the results of using a neural network based nonlinear prediction algorithm to extrapolate truncated magnetic resonance data. The extrapolation is intended to reduce the truncation artifacts that result when reconstructing an image from a limited k-space magnetic resonance data set using the discrete Fourier transform. When attempting to quantitatively compare Yan and Mao´s method with the authors´ own existing constrained modeling algorithm, the authors discovered a systematic error in Yan and Mao´s analysis. With the error corrected, it was found that Yan and Mao´s approach worked significantly better than they have reported and was more stable in the presence of noise.<>