Rotating gamma-camera tomography is the method of choice for ²⁰¹Tl perfusion imaging to detect myocardial ischaemia or infarction, but the optimum technique has not previously been investigated. Variables which can affect the utility of the images include acquisition time, collimator sensitivity and resolution, spatial filtering, reconstruction matrix and display parameters. An 'Iowa' design myocardial phantom containing ²⁰¹Tl, with a defect representing an unperfused area, was used to study the effect of these variables. Results were assessed by measuring image contrast and -non-uniformity and by observer grading trials. Receiver operator characteristic (ROC) curves and contrast ratios indicating that filtering the acquisition data with a count-dependent (Metz) filter before reconstruction and using a narrow slice width gave best results. No single collimator was superior for all four defect locations investigated but low-energy general purpose and high-resolution collimators were preferred to those with greater sensitivity.