Skip to main content
. 2022 Jan 11;1:796078. doi: 10.3389/fradi.2021.796078

Table 2.

Evaluation on the test set.

Baseline Wu et al. (18) Ensembles MC-drop Ours (MTL)
Overall 70.69 ± 0.36 (<0.01) 71.26 (<0.01) 73.39 ± 0.98 (<0.01) 72.82 ± 0.29 (<0.01) 80.46 ± 0.29
Density
   A 78.42 ± 0.12 (0.03) 81.34 (0.13) 79.14 ± 0.91 (0.05) 80.83 ± 0.10 (0.12) 87.26 ± 0.14
   B 68.14 ± 0.19 (<0.01) 70.32 (<0.01) 72.14 ± 1.05 (<0.01) 70.97 ± 0.17 (<0.01) 79.58 ± 0.13
   C 67.15 ± 0.16 (0.19) 68.46 (0.21) 68.23 ± 0.90 (0.23) 66.42 ± 0.16 (0.10) 74.23 ± 0.10
   D 64.81 ± 0.09 (0.07) 74.81 (0.36) 67.12 ± 0.87 (0.11) 73.33 ± 0.19 (0.23) 83.44 ± 0.07
View angle
   CC 63.72 ± 0.14 (<0.01) 69.76 (<0.01) 67.54 ± 1.01 (<0.01) 66.27 ± 0.29 (<0.01) 78.47 ± 0.13
   MLO 78.16 ± 0.21 (0.14) 69.82 (<0.01) 79.29 ± 1.12 (<0.30) 79.72 ± 0.29 (0.31) 82.44 ± 0.03

Binary cancer classification performance [“AUC score (p-value to ours)”] of the compared methods on the entire DHMItest dataset and for the subsets filtered per density class and per view angle.