Table 2.

Area under the curve from the ROC analysis of pathologic complete response prediction using different MRI measures

AUC	Baseline	After 2 cycles	After 3-4 cycles	After 5 cycles
Size measure: number of lesions in the analysis (pCR)	41 (7)	13 (3)	18 (3)	36 (6)
Lesion size
Maximal diameter	0.563			0.821
2D diameter	0.580	0.667	0.889	0.761
3D diameter	0.626	0.633	0.933	0.761
3Dseg	0.527	0.600	0.889	0.788
Size ratio (TP/BS)
2D diameter		0.833	0.889	0.817
3D diameter		0.833	0.933	0.811
3Dseg		0.700	0.867	0.727
ADC measure: number of lesions in the analysis (pCR)	42 (7)	14 (3)	19 (3)	34 (6)
2D ADC values
Mean	0.637	0.667a	0.542	0.800
3D ADC values
Mean	0.633	0.697a	0.500	0.743
Median	0.635	0.788a	0.563
15th percentile	0.669	0.697a	0.583
25th percentile	0.673	0.636a	0.521
75th percentile	0.527	0.697a	0.500
90th percentile	0.531	0.636a	0.521
Δ 3D ADC (TP-BS)
Mean		0.606	0.729	0.577
Median		0.788	0.750
15th percentile		0.667	0.771
25th percentile		0.788	0.750
75th percentile		0.515	0.646
90th percentile		0.576	0.729

Note: pCR pathologic complete response, Δ difference of, 3Dseg 3D segmented tumour size, TP time-point, BS baseline; the numbers in parenthesis correspond to the number of pCR patients in that group

^ain ROC analysis, using real ADC values, smaller values were considered positive for pCR prediction in BS, after three to four cycles and after five to eight cycles; however, higher values were considered positive for pCR prediction in data measured after two cycles