Identification of HER2-over-expression, HER2-low-expression, and HER2-zero-expression statuses in breast cancer based on 18F-FDG PET/CT radiomics

Table 4 The performances of all 12 constructed radiomic models in discriminating HER2 expression status in the training and independent validation sets from the internal institution

Models AUC (95%CI) and datasets	LR	KNN	SVM	RF
HER2-over-expression vs. others
Training set	0.766(0.690–0.831)	0.834 (0.765–0.890)	0.820(0.749–0.878)	0.843(0.774–0.897)
Independent validation set	0.756(0.634–0.854)	0.639 (0.510–0.755)	0.616(0.488–0.735)	0.785(0.665–0.877)
HER2-low-expression vs. others
Training set	0.783(0.708–0.846)	0.867(0.802–0.917)	0.836(0.767–0.892)	0.875(0.811–0.923)
Independent validation set	0.756(0.634–0.854)	0.701(0.574–0.808)	0.742(0.618–0.843)	0.733(0.609–0.835)
HER2-zero-expression vs. others
Training set	0.732(0.672–0.786)	0.929(0.890–0.958)	0.774(0.716–0.825)	0.799(0.744–0.848)
Independent validation set	0.696(0.598–0.782)	0.847(0.764–0.910)	0.706(0.610–0.791)	0.734(0.639–0.815)

ISSN: 1470-7330