Estimation of half fan-beam collimator parameters for brain SPECT with an L-shaped dual camera SPECT system

Accurate determination of collimator and imaging parameters for asymmetric fan-beam collimators is critical for accurate reconstructions. Previously the parameters were estimated by fitting data from a line source to an analytical model using the Marquardt algorithm. The authors have found with this method that (1) the parameters often cannot be determined uniquely, (2) local minima are problematic, and (3) the global minimum resides in a long narrow valley that contributes to the instability of the problem. In addition, inaccurate parameter estimates from multiple cameras can lead to misregistration problems and loss of resolution. The authors have developed a parameter estimation method to determine the unique set of parameters using multiple line sources and/or radii-of-rotation with the Powell algorithm. This new method was tested using both simulated data and data experimentally collected from a GE Optima L-shaped dual-camera SPECT system fitted with two half fan-beam collimators designed for high resolution brain SPECT. The results indicate that the authors´ new method provides stable and accurate estimates of the parameters, resulting in fewer image artifacts and higher resolution in the final reconstructed image than the previous method. The accuracy of the authors´ new method is demonstrated by reconstructions of a set of multiple line sources and a 3D Hoffman brain phantom