Evaluation of butterworth post-filtering effects on contrast and signal noise to ratio values for SPECT images reconstruction

Abstract

In nuclear medicine, single photon emission computed tomography (SPECT) imaging is very important for the detection or staging of variety of diseases such as myocardial perfusion, bone cancer, thyroid function and many more. To reduce the noise and improved contrast image in SPECT imaging is one of the essential tasks. This study aims to evaluate the optimum value for Butterworth filters to obtain the best physical image with high-quality contrast and reduced noise. In this study, the NEMA 2012/IEC 2008 phantom was imaged using the GE NM/CT Pro Discovery-670 system by implementing routine bone scan acquisition time. The large background compartment of the phantom was filled with 99mTc solution mixed with deionized water using a 5:1 sphere to background activity concentration ratio. The six spheres with different inner diameter and volume were filled with 99mTc solution of 300 kBq/mL with tumor background ratio. This activity concentration corresponds to a total injected activity of 500 MBq of 99mTc. Images in the trans-axial direction were reconstructed by Filter Back Projection technique. The images were examined in terms of hot regions, detectability, and contrast. The contrast and signal to noise ratio (SNR) were calculated for each sphere and COF values. Results show that the optimum parameter for Butterworth filter is at cut-off frequency (COF) of 7 cycle/mm. The hot regions from the body phantom study show that the detectability and contrast vary with the change of COF. The shape of five hot regions shows clarity and easier to draw the region of interest (ROI). Unfortunately, it is difficult to draw on the smallest ROI due to the presence of noise that caused blurring and unclear edges on the images. Statistically, a strong positive correlation was found between the contrast and SNR values for each sphere with coefficient of determination, R2 values above 0.96. The Butterworth filter at COF of 7 cycle/mm increases the contrast and SNR values at 4.4%–7.1%, and 61 to 859 counts, respectively for Sphere 2 to 6, compared to non-filtered images. In conclusion, the optimal parameter of COF for this study can be estimated from the calculation of total counts of contrast and SNR. Besides, findings from this study provides the physicians an important information in delivering the best and optimum diagnosis and treatment to the patients, thus improving the patient's well-being and health.