A hybrid maximum likelihood-least squares reconstruction algorithm for emission computed tomography

A novel efficient hybrid maximum-likelihood/weighted least squares (ML/WLS) reconstruction algorithm for emission computed tomography is presented. The ML/WLS algorithm constructs optimized linear combinations of prior and current iterative image and projection estimates using a weighted least squares objective function. The optimized projections are backprojected with a maximum likelihood-expectation maximization (ML-EM) procedure. The ML/WLS method, including attenuation compensation, was used to reconstruct experimental single-photon emission computed tomography projection data obtained from a Hoffman brain phantom using the emission-transmission computed tomography system at the University of California, San Francisco. Results indicate that the ML/WLS procedure generates images qualitatively similar to those reconstructed with a conventional ML-EM algorithm, with significantly fewer iterations