Objectives This study aims to evaluate nonlinear diffusion (NLD) processing to smoothen images while suppressing resolution degradation in single-photon emission computed tomography (SPECT) images. Phantom data were used for NLD method optimization. The resultant optimal settings were used for NLD processing of clinical images.Materials and methods99mTc was used to simulate tumors and normal soft tissues. Using the data collected, images were reconstructed. Images were processed using various k values and iteration. The background region's coefficient of variation (CV) was determined, and the effects of parameters on image properties were examined. NLD-processed images with optimal parameters were compared with Butterworth (BW)-filtered and nine-point smoothing (SM)-processed images to evaluate smoothing filter properties in real and frequency space. Receiver operating characteristic curve analysis was carried out on NLD-processed and BW048-processed bone SPECT images.ResultsFrom CVs in background, with NLD, increased k value and iteration led to a low CV, indicating enhanced smoothing effect. At k=0.9, a strong noise-reducing effect with less iteration was achieved. Contrasts and recovery coefficients of NLD were the highest. The visual score for SPECT image quality was significantly higher with NLD than with BW048, BW090, and SM. In the low-frequency and high-frequency ranges, BW048, BW090, and NLD showed similar signal strengths and NLD and BW090 showed high signal strength, respectively. SM processing reduced the signal strength at all frequency ranges. On receiver operating characteristic analysis, noise reduction using NLD processing enhanced diagnostic performance than with the use of BW processing.ConclusionNLD processing of bone SPECT images using optimized parameters enabled smoothing with less resolution degradation.
ASJC Scopus subject areas