Calculation of lung-heart ratios for single-photon emission computed tomography

We investigate the effectiveness of simple iterative reconstruction techniques in calculating lung-heart activity ratios (LHRs). The LHR has been shown to be an effective indicator of the severity of coronary artery disease in cardiac SPECT. A study was conducted with a mathematical cardiac torso phantom that modelled uptake of ²⁰¹Tl in the heart and lung regions. The projection data included only the effects of nonuniform photon attenuation. The data were first reconstructed with zeroth-order Chang and a variant of the Bellini method, both of which utilize information from the nonuniform attenuation map. This nonuniform (NU) Bellini method compensates exactly for attenuation in the heart region, but is incorrect for other regions in the medium. These reconstructions were then used as the initial estimates in the iterative Chang, variable step-size (VSS) Chang, and Morozumi methods, for one and five iterations. The average heart count (AHC) and average lung count (ALC) were calculated using region-of-interest (ROI) templates derived from the true activity map. The population mean LHR was tabulated as the ratio of the ALC to AHC. Using the same reconstruction procedure, we also calculated the sample mean LHR and standard deviation from 21 noisy 3D reconstructions. The results showed that the NU Bellini and zeroth-order Chang methods provided good estimates of the AHC, while both estimated the ALC poorly. The NU Bellini method yielded a negative ALC, while the zeroth-order Chang method estimated the ALC at twice its true value. Application of one iteration of Chang can correct for the negative bias in the ALC for the NU Bellini estimate, and can correct for the positive bias in the ALC for the zeroth-order Chang estimate, while retaining a low bias in the AHCs for the two methods. Thus more precise LHRs are produced. Furthermore, the sample standard deviation of the LHR estimate was shown to be small in both of these cases