Interpretation of myocardial perfusion images, produced with blood flow tracer rubidium-82 chloride (82Rb) and positron emission tomography (PET), can be affected when high tracer uptake in extra-cardiac organs adjacent to the heart (stomach) interferes with the myocardium. Since extra-cardiac organs are physically spatially distinct from the myocardium, extra-cardiac interference (ECI) in 82Rb PET images arises from limited spatial resolution, and cardiac and respiratory motion. This thesis aims to provide automated methods that detect and correct ECI. Three algorithms were developed to fulfill these aims: the first detects and ranks severity of ECI, the second attempts ECI correction based on factor analysis of dynamic image series, and the third corrects ECI with a 1D convolution-based method. All algorithms were developed, implemented and evaluated based on sets of clinical images. The detection and severity classification (DSC) algorithm was developed based on concordance of a 200 image dataset with clinical interpretation. It detected ECI with high accuracy (97% sensitivity and 82% specificity), low failure rate (<1%) and short execution time (<7s). The algorithm was used to estimate prevalence of ECI in a 4920- image dataset and to determine if simple modifications to image processing protocols could reduce ECI prevalence and/or severity. While reduced filtering showed the most promise, none of the available modifications eliminated ECI in the majority of images. Factor analysis of dynamic image series uses differences in the temporal behaviour of tracer uptake in the myocardium compared to that in the extra-cardiac organ to separate the two structures. Variations of this approach, applied to 82Rb PET images, were not able to simultaneously correct images with ECI of all severities and avoid reducing myocardial intensity in images without interference prior to correction. The 1D convolution-based correction algorithm modeled the image point spread function, including the effects of motion, as a 1D Gaussian and the underlying myocardial and stomach tracer uptake as simple 1D rectangular functions. This algorithm corrected images with ECI of all severities without reducing myocardial intensity in ECI-free images and >90% of scans examined showed visually acceptable correction. The convolution-based correction algorithm shows promise as a softwarebased ECI correction.