Premium
Signal to noise ratio based filter optimization in triple energy window scatter correction
Author(s) -
Blokland Koos J. A. K.,
Winn R. D. R.,
Pauwels E. K. J.
Publication year - 2000
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.1287286
Subject(s) - imaging phantom , filter (signal processing) , energy (signal processing) , image resolution , noise (video) , signal to noise ratio (imaging) , optics , physics , signal (programming language) , window (computing) , photon , data acquisition , pixel , computer science , artificial intelligence , computer vision , image (mathematics) , quantum mechanics , programming language , operating system
Triple energy window (TEW) scatter correction estimates the contribution of scattered photons to the acquisition data by acquiring additional data through two narrow energy windows placed adjoined to the main (photopeak) energy window. The contribution is estimated by linear interpolation and then subtracted. Noise amplification is reduced by filtering both the photopeak scintigram and the scatter estimate. We have studied the filter settings of each filter using a physical phantom filled with a201 Tl ‐ solution resulting in count densities comparable to clinical studies. The performance of order‐8 Butterworth filters at different cut‐off frequencies (CoFs) were compared based on signal to noise ratios (SNRs). The highest SNRs were obtained when the noisy scatter information was strongly filtered with the CoF less than or equal to 0.07 cycles/pixel (cpp). The best CoF for the filter of the photopeak image is object size dependent; smaller objects require a higher CoF. For objects with a size near the SPECT spatial resolution (∼15 mm) the optimal CoF is equal to 0.18 cpp. For larger objects (31.8 mm) the highest SNR was obtained with a CoF equal to 0.13 cpp. A CoF equal to 0.16 cpp is a good compromise for all objects with a diameter equal to the spatial resolution or larger. These results depend on the initial signal to noise ratio of the acquisition data and so on the count density.