Table 4 P-values (p) resulting from McNemar’s test to evaluate differences when predicting the margin status for 31 invasive carcinomas of no special type (NST) with the following three strategies: majority voting of seven physicians, prediction with Residual U-Nets (ResU-Nets), and prediction with intensity thresholding (Intens thresh) models

From: Supporting intraoperative margin assessment using deep learning for automatic tumour segmentation in breast lumpectomy micro-PET-CT

BCa type

Margin assessment strategy

p

NST (N = 31)

7 Physicians (Majority vote) vs. ResU-Net

1.0000

 

7 Physicians (Majority vote) vs. Intens thresh

0.0078

 

ResU-Net vs. Intens thresh

0.0078

  1. Significant performance differences following Bonferroni-Holm correction (α = 0.05, m = 3) are indicated in bold.