Table 2 Top-performing multivariate models for pCR prediction.

Model	Variable			Training (N = 109)			Testing (N = 54)			P value
Model	Time point	DCE	DWI	AUC	Accuracy	AUC CI	AUC	Accuracy	AUC CI	P value
1	RD, C2/BL	Minimum Maximum 5 GLCM features	1st percentile 95th percentile Minimum Kurtosis 24 GLCM features	0.91	0.84	0.850—0.959	0.80	0.76	0.674—0.930	< 0.001
2	AD, C4/BL	5th percentile 95th percentile 11 GLCM features	Standard deviation 95th percentile Mean 28 GLCM features	0.90	0.80	0.847—0.960	0.80	0.76	0.671—0.928	< 0.001
3	C4	10 GLCM features	3 GLCM features	0.82	0.71	0.744—0.899	0.79	0.76	0.662—0.923	< 0.001
3	AD, C4/BL	–	5 GLCM	0.82	0.71	0.744—0.899	0.79	0.76	0.662—0.923	< 0.001

AD, Absolute difference; RD, Relative difference; BL, Baseline; C2, After 2 cycles of NAST; C4, After 4 cycles of NAST; GLCM, Gray level co-occurrence matrix; AUC, Area under the receiver operating characteristic (ROC) curve; CI 95% Confidence interval.

Quick links

Search