ULTRA-HIGH-RESOLUTION BRAIN SPECT IMAGING - SIMULATION RESULTS

The spatial resolution in a reconstructed SPECT image is obviously influenced by the intrinsic resolution of the detector, but it is not generally recognized that the photon-counting efficiency of SPECT systems is also determined by the intrinsic resolution. In fact, it is often stated that increased intrinsic detector resolution is of little use since the overall resolution is limited by the collimator rather than the detector, and that collimator resolution cannot be increased without an unacceptable sacrifice in efficiency. In this paper we attempt to demonstrate that improvements in detector resolution can lead to both improved spatial resolution in the image and improved counting efficiency compared to conventional systems. This paradoxical conclusion results from optimizing the geometry of a multiple-pinhole coded-aperture system when detectors of very high resolution are available. Suitable semiconductor detectors, with sub-millimeter pixels, are currently under development in our laboratory and are discussed elsewhere in this volume. In this paper we report simulation studies that demonstrate the image quality that is attainable with such detectors. Reconstructions were performed using an iterative search algorithm on a custom-designed parallel computer. The imaging system was described by a calculated system matrix relating all voxels in the object space to all pixels on the detector. We found a resolution close to 2 mm on the reconstructed images obtained from these computer simulations with clinically reasonable exposure times. This resolution may be even further improved by optimization of the multiple-pinhole aperture. Thus the novel semiconductor modular gamma-camera design should provide a large improvement not only in detector resolution but also in reconstructed resolution.