Skip to main content

Table 2 Clinical image scoring from two readers of 20 whole-body FDG scans acquired on the D710 scanner. p-values were calculated using the Wilcoxon signed-rank test. The values in bold represent the best (highest) score for each metric

From: Sequential deep learning image enhancement models improve diagnostic confidence, lesion detectability, and image reconstruction time in PET

Reconstruction

Lesion detectability

Diagnostic confidence

Image quality

ToF-BSREM

3.50 ± 0.07

3.35 ± 0.08

2.50 ± 0.10

ToF-OSEM + DLE

3.27 ± 0.10

(p = 0.19)

3.33 ± 0.10

(p = 1.00)

3.98 ± 0.13

(p = 0.001)

OSEM + DLE + DLT

2.88 ± 0.17

(p = 0.03)

2.77 ± 0.16

(p = 0.04)

2.95 ± 0.11

(p = 0.04)

ToF-OSEM + DLE + DLT

4.42 ± 0.11

(p < 0.001)

4.45 ± 0.11

(p = 0.001)

3.88 ± 0.10

(p = 0.001)

ToF-BSREM + DLT

4.28 ± 0.11

(p = 0.001)

4.15 ± 0.10

(p = 0.001)

2.88 ± 0.08

(p = 0.009)

ICC

0.56

0.57

0.65