Automated interventricular septum segmentation for black-blood myocardial T2* measurement in thalassemia.

PURPOSE: To develop and validate an automated segmentation method that extracts the interventricular septum (IS) from myocardial black-blood images for the T2* measurement in thalassemia patients. MATERIALS AND METHODS: A total of 144 thalassemia major patients (age range, 11-51 years; 73 males) were scanned with a black-blood multi-echo gradient-echo sequence using a 1.5 Tesla Siemens Sonata system (flip angle 20°, sampling bandwidth 810 Hz/pixel, voxel size 1.56 × 1.56 × 10 mm(3) and variable fields of view (20-30) × 40 cm(2) depending on patient size). The improved Chan-Vese model with an automated initialization by the circular Hough transformation was implemented to segment the endocardial and epicardial margins of the left ventricle (LV). Consequently, the IS was extracted by analyzing the anatomical relation between the LV and the blood pool of the right ventricle, identified by intensity thresholding. The proposed automated IS segmentation (AISS) method was compared with the conventional manual method by using the Bland-Altman analysis and the coefficient of variation (CoV). RESULTS: The T2* measurements using the AISS method were in good agreement with those manually measured by experienced observers with a mean difference of 1.71% and a CoV of 4.15% (P < 0.001). CONCLUSION: Black-blood myocardial T2* measurement can be fully automated with the proposed AISS method.

[Reconstruction of dynamic positron emission tomographic images by exploiting low rank and sparse penalty].

OBJECTIVE: To propose a new method for dynamic positron emission tomographic (PET) image reconstruction using low rank and sparse penalty (L&S). METHODS: The L&S reconstruction model was established and the split Bregman method was used to solve the optimal cost function. The one-tissue compartment model was used to simulate a set of PET 82Rb myocardial perfusion image. The L&S reconstruction method was compared with maximum likelihood expectation maximization (MLEM) method, low-rank penalty method and sparse penalty method. RESULTS: The L&S reconstruction method had the smallest MSE and well maintained the feature information. The polar map created by L&S method was the most similar with the reference actual polar map. CONCLUSION: L&S reconstruction method is better than the other three methods in both visual and quantitative analysis of the PET images.