Improved Lesion Detection in SPECT using MLEM Reconstruction

Receiver Operating Characteristic (ROC) analysis was applied to determine if the Maximum Likelihood Expectation Maximization (MLEM) reconstruction algorithm improves lesion detectability in SPECT when compared with a Filtered Backprojection (FBP) algorithm. The performance of the reconstruction algorithms was evaluated based on cold lesion detection by human observers. Simulated SPECT projection data with added Poisson noise were reconstructed using the MLEM and the FBP algorithms. The parameters of the study, including lesion size and total counts, were chosen to maximize the statistical power of the test. Three hundred images were generated, and nine observers were employed for the study. The average ROC curves for both reconstruction methods were measured. Diagnostic performance was quantified using the area under the ROC curves. MLEM showed improved detectability for all the observers. Observer P-values were calculated, and Student's t-test was performed in order to verify the statistical significance of the observed improvement.