Table 4. Quantitative evaluation of the results of the different methods from Figure 10 for 50 K, 500 K, and 5 M counts.

Methods	50 K				500 K				5 M
	PSNR	SSIM	rRMSE		PSNR	SSIM	rRMSE		PSNR	SSIM	rRMSE
MLEM + filter	18.639	0.816	0.281		22	0.923	0.22		25.137	0.977	0.153
MLEM + NLM	19.99	0.811	0.277		25	0.929	0.151		27.853	0.969	0.112
CNN	17.539	0.841	0.367		27	0.956	0.12		30.034	0.982	0.087
GAN	19.916	0.877	0.279		28	0.963	0.114		30.679	0.983	0.068
LCPR-Net	21.956	0.887	0.224		29	0.965	0.102		32.203	0.986	0.061

PSNR, peak signal-to-noise ratio; SSIM, structural similarity index; rRMSE, relative root mean square error; MLEM, maximum-likelihood expectation-maximization; NLM, nonlocal means; CNN, convolutional neural network; GAN, generative adversarial network; LCPR-Net, low-count positron emission tomography image reconstruction network.