Development of a novel estimation method for spatially varying noise levels in MR images

点击翻译按钮获取中文摘要

Signal-to-noise ratio (SNR) is an important metric commonly used to evaluate image quality in magnetic resonance (MR) imaging, because it is quite sensitive to any change of imaging techniques and hardware performances. The SNR is defined as the ratio of the mean signal intensity within a homogeneous region of interest to noise estimated as its standards deviation. However, poor signal uniformity is fairly common in MR images, and thus standard deviation of pixel values is more strongly influenced by the image uniformity than true random noise. In this study, we have developed an accurate and robust noise estimation method for calculating reliable SNR values.