Predicting the efficacy of neoadjuvant chemotherapy in breast cancer patients based on ultrasound longitudinal temporal depth network fusion model

Table 2 Performance of three single-modality models constructed using five machine learning algorithms for predicting the efficacy of NAC in the training cohort and the test cohort

Model	Features	Training cohort				Test cohort
Model	Features	Mean AUC (95% CI)	ACC	SEN	SPE	Mean AUC (95% CI)	ACC	SEN	SPE
XGBoost	Pre	0.773(0.735–0.855)	0.708	0.714	0.703	0.726(0.660–0.732)	0.645	0.775	0.847
	Post	0.799(0.752–0.880)	0.753	0.676	0.891	0.776(0.664–0.881)	0.752	0.700	0.792
	Delta	0.785(0.727–0.846)	0.763	0.703	0.805	0.710(0.667–0.767)	0.697	0.650	0.798
RF	Pre	0.768(0.733–0.845)	0.685	0.660	0.773	0.721(0.690–0.737)	0.699	0.750	0.660
	Post	0.814(0.753–0.863)	0.735	0.626	0.812	0.713(0.650–0.779)	0.688	0.750	0.642
	Delta	0.759(0.722–0.843)	0.772	0.659	0.852	0.707(0.657–0.817)	0.690	0.625	0.830
SVM	Pre	0.744(0.732–0.824)	0.644	0.668	0.784	0.708(0.678–0.720)	0.624	0.688	0.817
	Post	0.705(0.677–0.752)	0.717	0.643	0.883	0.784(0.737–0.808)	0.710	0.750	0.792
	Delta	0.832(0.806–0.890)	0.749	0.648	0.820	0.724(0.669–0.750)	0.677	0.626	0.792
logistics	Pre	0.698(0.617–0.755)	0.653	0.605	0.775	0.679(0.606–0.703)	0.602	0.675	0.736
	Post	0.797(0.772–0.872)	0.753	0.659	0.820	0.771(0.735–0.827)	0.731	0.700	0.755
	Delta	0.783(0.749–0.862)	0.703	0.627	0.828	0.700(0.670–0.733)	0.691	0.641	0.749
DT	Pre	0.696(0.678–0.829)	0.676	0.835	0.797	0.640(0.500–0.730)	0.634	0.875	0.736
	Post	0.740(0.656–0.781)	0.753	0.659	0.820	0.727(0.607–0.727)	0.731	0.650	0.755
	Delta	0.695(0.657–0.755)	0.671	0.825	0.655	0.614(0.579–0.704)	0.602	0.700	0.728

AUC, the area under curve; ACC, accuracy; SEN, sensitivity; SPE, specificity; 95% CI, 95% confidence interval; XGBoost, eXtreme Gradient Boosting; RF, random forest; SVM, support vector machine; logistics, multivariate logistic regression analysis; DT, decision tree

ISSN: 1465-542X