Quantification of two radionuclides in simultaneous 123I/99mTc SPECT with artificial neural networks

This study was carried out to reconstruct quantitative images in simultaneous ¹²³I/^99mTc SPECT. In the ¹²³I /^99mTc SPECT study the image reconstructed with the counts of primary photons emitted from the isotope with low photo-peak energy (^99mTc) is distorted by Compton scattered photons originating in the isotope with high photo-peak energy (¹²³I). To quantify each radioisotope in the simultaneous data acquisition, we have proposed a method with an artificial neural network, in which ten energy windows were used. But the method is not commercially available except for a specially designed gamma camera system. In this paper we proposed a new method, which uses only three energy windows. The performance of our method was investigated with Monte Carlo simulations, experiments with a brain phantom and a cardiac phantom, and clinical studies. The results showed that despite the fact that this method used only three windows its performance was comparable to that obtained with our previously reported method with ten windows. Our proposed method could separate ^99mTc and ¹²³I with an error of only about 5%. The results will help make feasible clinical use with dual-isotope SPECT.