MRI myocardium T2∗ measurement by a new PCA-based object recognition algorithm

MRI myocardium T∗2measurement is a very important task in MRI for the detection, for example, of myocardial iron overload. Generally, T∗2values are obtained by a T∗2multiecho MRI. In particular, signal intensities of selected ROIs on different TE images are evaluated and the signal -TE relation is used in order to estimate the T∗2. In order to correctly estimate the T∗2, it is important that the different selected ROIs correspond to the same anatomical pixels. In this paper, a new PCA-based recognition algorithm is presented in order to recognize and quantify the same anatomical pixels on different TE images of a multiecho sequence. The algorithm was implemented in Matlab. In order to test the algorithm and to obtain preliminary results, a group of 10 patients, referred to MRI with presumptive, clinical diagnosis of myocardial iron overload, was examined in order to test the algorithm. All patients showed no myocardial iron overload with a T∗2>20ms.To assess intra- and interobserver variability, two observers blindly analyzed the data by delimiting myocardial region. A good intra- and inter-observer reproducibility was obtained, in fact the mean difference between the two observer measurements was 0.8 ms and the 95% limits of agreement on the Bland-Altman plot were -4.8 to 6.5 ms.