In PET images, the intensity of each pixel should be proportional to the amount of radioactivity in the tissue, but errors called partial volume effects may occur. This study investigated the effect of ML-EM algorithms on the accuracy of tumor size detection. GATE 7.2 software was used to simulate the imaging system. A heart phantom containing a tumor was created with XCAT settings, and the binary files created with XMedCon were converted to DICOM format and combined with MATLAB. Then, 7 transaxial slices containing the tumor tissue were separated and the file format was changed to .mhd. Image reconstruction was performed with CASTOR software.
The results showed that increasing the number of iterations in the ML-EM algorithm amplifies the noise in the range of 52 to 56, while image contrast improves up to 33 iterations. The SSIM and PSNR values also increased with increasing iterations. Also, in all iterations, the tumor size in slices (which include a small volume of the tumor) was overestimated. This magnification is due to the effect of blending adjacent tissues and was more prominent than other slices.