Determination of optimal parameters for SPECT bone imaging

High spatial frequency information and relatively small discrete structures in SPECT images of bony structures preclude the indiscriminate use of smoothing and low pass filtering to improve image quality. To determine optimal parameters for this application, SPECT images of a phantom, normal pelvis, and patients with known pelvic bony abnormalities were acquired and processed using varying activity in the case of the phantom, use of 1800 or the full 3600 of acquired data, processing with and without attenuation and misalignment corrections, 12 different tomographic reconstruction filter kernels, inter- and intraimage smoothing, and projection image reregistration. The resulting images were numerically analyzed and reviewed by clinicians for diagnostic quality. Data were acquired using a Siemens Orbiter camera and a Siemens MicroDelta processor. Results indicate that image quality is relatively more sensitive to motion and mispositioning than to small differences in reconstruction filter kernel parameters