Positron emission tomography (PET), on account of its non-invasiveness, exquisite sensitivity, and ability to perform longitudinal studies is becoming a very popular tool in both clinical and preclinical imaging. The increasing role of PET however, has placed increased demands on both spatial resolution and gamma-ray detection efficiency (sensitivity). Developing an efficient, yet high-resolution PET detector has been a fundamental challenge and most attempts to improve resolution make use of smaller scintillation crystals that sacrifice sensitivity. This dissertation aims to study and investigate in detail, the potential of a novel gamma-ray detector in providing a cost-effective alternative for high performance PET. The detector design comprises a single, continuous scintillator read out by large-area solid-state photosensors on both sides. In addition to possibly providing high spatial resolution and sensitivity, the detector has fewer readout elements (reduced costs) and could also help reduce parallax errors by measuring the depth-of-interaction (DOI) of the gamma-ray within the scintillator. Moreover the design is expected to be compatible with MRI and would enable simultaneous imaging with both PET and MRI techniques. An initial prototype detector was built with a 10 mm thick lutetium oxyorthosilicate (LSO) scintillator and large-area avalanche photodiodes (APDs). Due to the use of ready-made electronics, only two channels on the detector were initially instrumented; early promise was demonstrated. Subsequent efforts were devoted towards developing a fully working prototype detector. The efforts ranged from developing a thorough understanding and reduction of electronic noise, to a detailed evaluation and optimization of each detector component. In parallel, a detailed Monte Carlo model of the detector was also built. The model was used to first predict detector performance, and later, to understand in detail the influence of various detector components. To utilize the full potential of the detector, a novel Maximum Likelihood based, 3D event-positioning algorithm was also developed. Proof-of-concept is demonstrated with an improved version of the prototype detector. Thorough experimental characterization reveals an isotropic resolution of ~3 mm, in excellent agreement with predictions from Monte Carlo simulations. The validated Monte Carlo model was also used to optimize detector design and demonstrate that the above methods, when combined together could realize a high-performance PET detector with spatial resolution better than 1 mm.