Skip to main content
. 2022 Feb 11;32(7):4868–4878. doi: 10.1007/s00330-022-08558-0

Table 2.

Confusion matrices comparing BI-RADS 2/3 and BI-RADS 4/5 classification results of dCNN (a), human reader 1 (b), and reader 2 (c) to ground truth (actuals) based on single images and inter-reader agreement (d), intra-reader reliability for reader 1 (e) and reader 2 (f)

graphic file with name 330_2022_8558_Tab2_HTML.jpg