Skip to main content

Table 2 Mean ± standard deviation of performance metrics across 100 within-patient series pair for manual and automated registrations in the small-scale substudy

From: Fully automated convolutional neural network-based affine algorithm improves liver registration and lesion co-localization on hepatobiliary phase T1-weighted MR images

Registration method

Liver overlap score

Image correlation

Observation distance (annotated by reader 1)

Observation distance (annotated by reader 2)

Manual registration

Reader 1

0.74 ± 0.10

0.36 ± 0.14

14.73 ± 8.20

/

Reader 2

0.73 ± 0.12

0.35 ± 0.14

/

16.09 ± 8.49

Automated registration (with masking)

Binary mask

0.86 ± 0.06

0.42 ± 0.16

8.40 ± 4.62

8.42 ± 5.78

Intensity mask

0.86 ± 0.06

0.43 ± 0.16

8.26 ± 4.64

8.49 ± 5.67

Automated registration (without masking)

Whole image

0.73 ± 0.15

0.39 ± 0.17

22.95 ± 24.42

18.11 ± 16.06

  1. Statistical comparisons between registration methods were performed using paired t tests (Additional file 1: Table S1)