Evaluating the Impact of Intensity Normalization on MR Image Synthesis

摘要

Image synthesis learns a transformation from the intensity features of an input image to yield a different tissuecontrast of the output image. This process has been shown to have application in many medical image analysistasks including imputation, registration, and segmentation. To carry out synthesis, the intensities of the inputimages are typically scaled-i.e., normalized-both in training to learn the transformation and in testing whenapplying the transformation, but it is not presently known what type of input scaling is optimal. In this paper,we consider seven different intensity normalization algorithms and three different synthesis methods to evaluatethe impact of normalization. Our experiments demonstrate that intensity normalization as a preprocessing stepimproves the synthesis results across all investigated synthesis algorithms. Furthermore, we show evidence thatsuggests intensity normalization is vital for successful deep learning-based MR image synthesis.