Hybrid wavelet shrinkage (HWS) filter suppress the noise propagation in ordered subset expectation maximization (OS-EM) algorithm
| Autor: | Shiraishi, Takahiro, Kanno, Iwao, Ito, Hiroshi, Yoshikawa, Kyosan, Tanimoto, Katsuyuki, Ishii, Noriyuki, Kimura, Taku, Omatsu, Mika, Ohashi, Seiya, Toubaru, Sachiko, Takano, Harumasa, Tsuji, Hiroshi, Ando, Yutaka, Kamada, Tadashi, Shibayama, Kouichi |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2012 |
| Popis: | Purpose Image reconstruction for most nuro-transmission PET scan data is performed by FBP because of a positive bias caused by the noise propagation in OS-EM algorithm. A positive bias affects the estimation of binding potential (BP) calculated from radioactive concentrations of reference regions and target region in nuroreceptor binding assay. To suppress the noise propagation in OS-EM algorithm, the hybrid wavelet shrinkage (HWS) filter is developed. In this study, we performed the phantom study to evaluate the effect of HWS filter on noise propagation in OS-EM algorithm. \nMaterials and Methods Phantom studies are performed by SET-3000GCT/X (shimadzu corp). To evaluate the effect of the HWS filter, we planned the some phantom study. IEC body phantom filled with 18F (activity concentration ratios relative to background of 4:1) are used for evaluation of image quality. A pool phantom filled with 18F (11.4kBq/ml) is scanned by dynamic scan mode for evaluation the quantification of PET image. Prior to emission scans, a transmission scan (6min) for attenuation correction with a 137Cs point source with a BGO transmission detector ring coaxially attached to the GSO emission detector ring. Reconstruction parameters of PET images are follows; FBP with Gaussian filter 5.0mm, OS-EM, OS-EM with HWS filter (OS-EMw). Iteration numbers in OS-EM algorithm are varied from 1 to 100 to evaluate the noise propagation. \nResults In pool phantom study, a positive bias at low count rate is shown in PET images reconstructed by OS-EM algorithm, while any biases are not shown in images reconstructed by FBP and OS-EMw algorithm. The N10 values are diverged due to noise propagation in OS-EM algorithm while N10 values calculated from PET image reconstructed by OS-EMw algorithm are convergence at the point of 10 iterations in IEC body phantom study. Similarly, the Q10 values are diverged depends on iteration number in OS-EM algorithm while that of OS-EMw has fast convergence. Furthermore, these results indicate that OS-EMw has the fast convergence of pixel number and suppress the noise propagation. \nConclusion According to our study, OS-EMw has fast convergence and suppresses the noise propagation in OS-EM algorithm. These results show that HWS filter can be a standard de-noising method for nuro-recptor PET imaging with OS-EM algorithm. Annual Congress of the European Association of Nuclear Medicine |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|