Non-Local Means Denoising for Myocardial T2* Measurement

Introduction: Myocardial T2* measurement is a valuable tool for non-invasive assessment of tissue iron overload which can prevent disease in thalassemia patients (1). For the heavily iron loaded heart, T2* is substantially shortened (<10ms) and the signal to noise ratios (SNRs) of later echo images are usually very low (2). Various signal processing techniques have been proposed for denoising MRI images with varied levels of success, but also criticised for smoothed image details. Waveatom technique, which has the ability to adapt to arbitrary local directions of a pattern, demonstrated advantages over wavelet in reducing MRI noise (3). A more recent development is a non-local means (NLM) method (4), which has shown great potential in dealing with MRI noise (4). To be useful in clinical settings, a denoising algorithm must preserve fine details of the original images and should not alter the diagnostic information extracted from the images. By comparing these two techniques on myocardial T2* images from patient with cardiac overload, the purpose of this study was to evaluate denoising effects on T2* images and quantification using state-of-art signal processing techniques.