Fig. 4

Axial views of the patient’s lesion in Fig. 3 (as indicated by the red line in y in Fig. 3). a–e PET images at scan times of 6 s, 15 s, 30 s, 60 s and 150 s, respectively, are shown in axial views. f–j PET images synthesized from a–e using the NLM method. k–o PET images synthesized from a-e using 3DNet. p–t PET images synthesized from a-e using p3DNet. u–y Full-time reference images. PET images synthesized using the p3DNet method can effectively recover anatomical structures (red arrow). (1–20) are the residual maps (absolute values) obtained by subtracting the reference image from a–t. (1–5) Residual maps of a–e. (6–10) Residual maps of (f–j). (11–15) Residual maps of k–o. (16–20) Residual maps of (p–t)