Validation of OSEM algorithm for attenuation correction in cardiac SPECT studies

Photon attenuation is believed to be one of the major causes of artefacts in cardiac SPECT studies and therefore should be corrected to improve the specificity of cardiac SPECT imaging for the detection of coronary artery disease. Today, many SPECT system manufacturers provide optional hardware (CT imaging) and software (OSEM reconstruction algorithm) technology attempting to produce attenuation corrected (artefacts-free) images and more accurate quantitative cardiac SPECT studies. However, this attenuation correction (AC) technique has not received widespread acceptance in routinely clinical practice due to the less-than-optimal validations of AC using realistic data and the potential for creating new artefacts. The purpose of this study is to validate whether AC can produce quantitatively accurate cardiac SPECT images and whether it creates more additional artefacts than it has cured. METHODS: Multiple experiments were carried out using cardiac SPECT data and the state-of-the-art AC technology available in Sheffield Teaching Hospitals (CT/SPECT hybrid system and OSEM algorithm). Cardiac SPECT data obtained from computer-simulated models, a physical phantom (anthropomorphic phantom with cardiac insert) and one patient data were used to validate and evaluate AC technique. Qualitative and quantitative analysis was performed to compare cardiac SPECT images without and with AC. RESULTS: The computer-simulated phantom showed that a realistic model can be generated and was used successfully for validation purposes. The simulation results showed that the attenuation artefacts were corrected significantly in both noiseless and noisy data. Using post-filtering with high number of iterations was found to be the reasonable criteria of stopping the reconstruction. The simulated data with 1 cm misregistration error between emission and transmission images produces a significant error in AC images. Attenuation artefacts were significantly corrected in all walls of cardiac insert in physical phantom with some non-uniformity in images reconstructed with low number of iterations. These images become more uniform with high number of iterations. Patient data demonstrated that AC appears to have worked well. CONCLUSION: OSEM algorithm was able to correct attenuation artefacts (without scatter modelling) of cardiac SPECT images using CT attenuation map with no evidence of generating additional artefacts. If such artefacts exist they must arise either from early stopping of the OSEM algorithm or other sources e. g. scatter and nusregistration. In conclusion, OSEM algorithm is recommended for AC in cardiac SPECT studies.