Similar to conventional nuclear medicine imaging, parallel-hole collimators (Figure 13.10a) are commonly used in camera-based SPECT systems. As described earlier, the tradeoff between detection efficiency and spatial resolution of parallel-hole collimator is a limitating factor for SPECT. A means to improve SPECT system performance is to improve the tradeoff imposed by the parallel-hole collimation.

To achieve this goal, converging-hole collimator designs that increase the angle of acceptance of incoming photons without sacrificing spatial resolution have been developed. Examples are fan-beam [Jaszczak et al., 1979b; Tsui et al., 1986], cone-beam [Jaszczak et al., 1987], astigmatic [Hawman and Hsieh, 1986], and more recently varifocal collimators. As shown in Figure 13.10b, the collimator holes co nverge to a line that is oriented parallel to the axis of rotation for a fan-beam collimator, to a point for a cone-beam collimator, and to various points for a varifocal collimator, respectively. The gain in detection efficiency of a typical fan-beam and cone-beam collimator is about 1.5 and 2 times of that of a parallel-hole collimator with the same spatial resolution. The anticipated gain in detection efficiency and corresponding decrease in image noise are the main reasons for the interest in applying converging-hole collimators in SPECT.

Despite the advantage of increased detection efficiency, the use of converging-hole collimators in SPECT poses special problems. The tradeoff for increase in detection efficiency as compared with parallel-hole collimators is a decrease in field of view (see Figure 13.10). Consequently, converging-hole collimators are restricted to imaging small organs or body parts such as the head [Jaszczak et al., 1979b; Tsui et al., 1986] and heart [Gullberg et al., 1991]. In addition, the use of converging-hole collimators requires special data-acquisition strategies and image reconstruction algorithms. For example, for cone-beam tomography using a conventional single planar orbit, the acquired projection data become increasingly insufficient for reconstructing transaxial image sections that are further away from the central plane of the cone-beam geometry. Active research is underway to study special rotational orbits for sufficient projection data acquisition and 3D image reconstruction methods specific for cone-beam SPECT.