Comparison of Rotationally Symmetric and Oriented Channels for the Hotelling Observer for Myocardial SPECT Images

Abstract

This study compared results given by the channelized Hotelling observer (CHO) with rotationally symmetric (RS) channels and the CHO with rotationally oriented (RO) channels for a myocardial SPECT defect detection study using simulated SPECT images. The 4D NCAT phantom was used to model a patient population with variations in anatomy, defects (size, location and contrast) and /sup 99m/Tc-sestmaibi uptake in the organs. Low noise, parallel projection data were generated using Monte Carlo simulation. Poisson noise was added to generate noisy realizations. Data were reconstructed using OS-EM at 1 & 4 subsets/iteration and at 1, 3, 5, 7 & 9 iterations and were then converted to 2D short-axis images. The CHO-RS used 3 RS frequency channels and the CHO-RO used 12 channels (3 frequency channels at 4 orientations) to form image vectors. For each OS-EM parameter setting, the detectability index was estimated from 864 defect-present & 864 defect-absent vectors. The CHO-RS and CHO-RO models predicted the same optimum OS-EM parameter setting of 4 subsets/iteration and 5 iterations and gave nearly the same ranking of all the parameter settings. The CHO-RO gave high absolute detectability values. In conclusion, for defect detection in simulated myocardial SPECT images, the CHO-RO model does not offer an advantage over the CHO-RS model.