**3. Results**

Figure 3 demonstrates Igamma, IGa68, and Ga-68 PET images after positron range correction based on CNN1 (ICNN1 PRC ), CNN2 (ICNN2 PRC ) and CNN3 (ICNN3 PRC ) of PHANbrain. With naked eye observation, boundaries in Ga-68 PET images became sharper after correction. Figure 4 demonstrates Igamma, IGa68, ICNN1 PRC , ICNN2 PRC , ICNN3 PRC of PHAN5rod with TBR = 0 and 10. The 2-mm rod in PHAN5rod with TBR = 0 that was barely seen in IGa68 became visible in ICNN2 PRC and ICNN3 PRC . The RMSE and PSNR between Igamma and IPRC are shown in Figure 5a for PHANbrain and Figure 5b for PHAN5rod with TBR = 10. Lower RMSEs and higher PSNRs were observed in Figure 5b than those in Figure 5a. For either phantom configuration, the lowest RMSE and the highest PSNR were found in CNN3-based correction, followed by CNN2- and CNN1-based correction.

**Figure 3.** (**a**) Igamma, (**b**) IGa68, (**c**) ICNN1 PRC , (**d**) ICNN2 PRC , (**e**) ICNN3 PRC of PHANbrain at 2 different axial slices (**top row** and **bottom row**).

**Figure 4.** (**a**) Igamma, (**b**) IGa68, (**c**) ICNN1 PRC , (**d**) ICNN2 PRC , (**e**) ICNN3 PRC of PHAN5rod with TBR = 0 (**top row**) and TBR = 10 (**bottom row**).

**Figure 5.** RMSE (solid line, left axis) and PSNR (dashed line, right axis) between Igamma and IPRC for (**a**) PHANbrain and (**b**) PHAN5rod with TBR = 10 (**top row**: CNN1-based correction; **mid row**: CNN2-based correction; **bottom row**: CNN3-based correction).

The RCs of Igamma, IGa68, ICNN1 PRC , ICNN2 PRC , ICNN3 PRC are shown in Figure 6a for PHAN5rod with TBR = 0 and Figure 6b for PHAN5rod with TBR = 10. The RCs of positron range corrected images were higher in TBR = 0 than those in TBR = 10 for 4- and 6-mm rods, while the difference was less obvious in 8- and 10-mm rods. Among three corrected images, the highest RCs were observed in ICNN3 PRC for 6-, 8-, and 10-mm rods in TBR = 0, while the highest RC for 4-mm rod was observed in ICNN2 PRC . With regard to TBR = 10, the highest RCs were observed in ICNN3 PRC for 4-, 6-, and 8-mm rods, while the highest RC for 10-mm rod was observed in ICNN2 PRC . The CVRCs of Igamma, IGa68, ICNN1 PRC , ICNN2 PRC , ICNN3 PRC were shown in Figure 7a for PHAN5rod with TBR = 0 and Figure 7b for PHAN5rod with TBR = 10. The CVRCs of positron range corrected images were lower in TBR = 0 than those in TBR = 10 for 4-, 6-, 8-, and 10-mm rods. Among three corrected images, the CVRCs of ICNN1 PRC were slight lower than those from ICNN2 PRC and ICNN3 PRC for either phantom configuration.

With regard to the spillover effect determined by using PHAN1sphere, the SORs of Igamma, IGa68, ICNN1 PRC , ICNN2 PRC , ICNN3 PRC were 0.017, 0.026, 0.021, 0.020, and 0.020, respectively, while the corresponding CVSORs were 0.454, 0.424, 0.406, 0.416, and 0.441. Figure 8 demonstrates IGa68, ICNN1 PRC , ICNN2 PRC , ICNN3 PRC of PHAN20rod and comparison of intensity profiles through the dashed line. Sharper boundaries were observed in 4- and 5-mm rods after positron range correction. On the other hand, the image quality improvement was limited in 2- and 3-mm rods, except for the 3-mm rod with 8.44 × 10<sup>5</sup> Bq/mL in ICNN2 PRC and ICNN3 PRC .

**Figure 6.** RC of Igamma, IGa68, ICNN1 PRC , ICNN2 PRC , ICNN3 PRC using PHAN5rod with (**a**) TBR = 0 and (**b**) TBR = 10.

**Figure 7.** CVRC of Igamma, IGa68, ICNN1 PRC , ICNN2 PRC , ICNN3 PRC using PHAN5rod with (**a**) TBR = 0 and (**b**) TBR = 10.

**Figure 8.** (**a**) IGa68, (**b**) ICNN1 PRC , (**c**) ICNN2 PRC , (**d**) ICNN3 PRC of PHAN20rod and (**e**) comparison of intensity profiles through the dashed line in (**a**).
