This paper presents the design and experimental results obtained with a prototype adaptive SPECT that implements a recently developed adaptive angular sampling approach. With this system, the hardware system configuration can be adjusted in real time based on the imaging information being collected to offer an optimum sampling of an unknown object. This system consists of two major components. First, the detection system consists of two independent pinhole cameras based on recently developed energy-resolved photon-counting (ERPC) CdTe detectors. During an imaging study, both the detectors and pinhole apertures can move back and forth independently against the rotation object. This system also include a CT system that provides the physical profile of the object, and helps to determine the feasible sampling strategy for the given source geometry. Second, the adaptive angular sampling is controlled by a software package that consists of three key components: (a) a set of close-form equations for evaluating image variance and resolution attainable with given sampling strategy, (b) a gradient-based algorithm to search the optimum sampling strategy, (c) an efficient computation approach for speeding up the searching process. During an imaging study, the corresponding computation is performed on a PC to determine the optimum sampling and to guide the hardware configuration to be adjusted accordingly. We carried a series of experimental studies to demonstrate the benefit using this self-optimized adaptive angular sampling strategy. The SPECT imaging performance obtained is compared to those from conventional uniform sampling strategies.