Evaluation of noise equivalent count per line of response: new parameter for count statistics for PET systems

Abstract

The noise equivalent count (NEC) is a useful, widely accepted method of evaluating image quality in positron emission tomography (PET) from the standpoint of effective count statistics. However, NEC cannot be used when different types of PET scanners are compared owing to the differences in slice thickness and field of view. Moreover, NEC should be treated differently depending on whether the 2D or 3D mode is used for a given PET scanner. A new parameter "Specific NEC," which is NEC per line of response (LOR) was devised to compare image quality between different PET scanners and acquisition modes. Two PET scanners were employed, the CTI-Siemens ECAT EXACT HR(+) and ECAT EXACT 47. Images of a cylindrical (68)Ge phantom were scanned in 2D and 3D modes using various acquisition times ranging from 15 secx20 frames to 120 secx20 frames in order to examine the effect of count statistics on the quality of image reconstruction. The data were reconstructed using a ramp filter with a cutoff frequency of 0.5 cycles/pixel, corrected for dead time, random, attenuation, and scatter. The quality of the reconstructed images was evaluated with the coefficient of variation (COV; SD/average for the pixels within a 16 cm region of interest). Specific NEC was defined as NEC divided by the number of LOR for the entire scanner at detector level. COV showed a linear relationship with Specific NEC in double logarithmic plot within a given experiment. When the Specific NEC was used, all 2D and 3D mode showed the same relationship. The slight difference between the two scanners was attributed to the difference in slice thickness. Image quality was dependent on effective count statistics per number of LOR. Our method was considered effective for evaluating image quality in both 2D and 3D modes.