Skip to main content
. 2018 Jun 22;8:9490. doi: 10.1038/s41598-018-27764-9

Table 2.

Diagnostic performance of perfusion parameters from DCE-MRI to predict pathologic complete response.

Cut-off value* non-CR (n = 61) CR (n = 13) Sensitivity (99.8% CI**) Specificity (99.8% CI**) AUC (99.8% CI**)
2D analysis of BP CL
Ktrans ≥0.03 16 8 0.615 (0.190–0.940) 0.738 (0.560–0.915) 0.683 (0.425–0.941)
Kep ≥0.21 12 6 0.462 (0.025–0.795) 0.803 (0.643–0.964) 0.627 (0.331–0.924)
Ve ≥0.20 11 6 0.462 (0.025–0.795) 0.820 (0.664–0.975) 0.629 (0.332–0.925)
3D histogram analysis of tumor
Ktrans
  25th perc ≥0.11 29 9 0.692 (0.288–1.000) 0.525 (0.323–0.726) 0.610 (0.335–0.885)
  50th perc ≥0.22 24 9 0.692 (0.288–1.000) 0.607 (0.409–0.804) 0.624 (0.305–0.943)
  75th perc ≥0.34 19 9 0.692 (0.288–1.000) 0.689 (0.501–0.876) 0.612 (0.285–0.938)
  mean ≥0.24 23 9 0.692 (0.288–1.000) 0.623 (0.427–0.819) 0.605 (0.287–0.924)
  kurtosis ≤−0.92 28 4 0.308 (0.000–0.616) 0.541 (0.340–0.742) 0.502 (0.235–0.769)
  skewness ≤0.52 38 12 0.923 (0.690–1.000) 0.377 (0.181–0.573) 0.647 (0.388–0.905)
Kep
  25th perc ≥0.14 50 13 1.000 (1.000–1.000) 0.180 (0.025–0.336) 0.546 (0.267–0.825)
  50th perc ≥0.51 22 8 0.615 (0.190–0.940) 0.639 (0.445–0.833) 0.575 (0.257–0.893)
  75th perc ≥0.94 9 5 0.385 (0.000–0.710) 0.852 (0.709–0.996) 0.571 (0.241–0.902)
  mean ≥0.55 24 8 0.615 (0.190–0.940) 0.607 (0.409–0.804) 0.574 (0.250–0.897)
  kurtosis ≥4.53 2 2 0.154 (0.000–0.395) 0.967 (0.895–1.000) 0.508 (0.212–0.804)
  skewness <0.82 41 10 0.769 (0.400–1.000) 0.328 (0.138–0.518) 0.535 (0.267–0.802)
Ve
  25th perc ≤0.43 31 9 0.692 (0.288–1.000) 0.492 (0.290–0.694) 0.472 (0.243–0.702)
  50th perc ≤0.52 28 9 0.692 (0.288–1.000) 0.541 (0.340–0.742) 0.487 (0.236–0.738)
  75th perc ≤0.61 30 8 0.615 (0.190–0.940) 0.508 (0.306–0.710) 0.484 (0.223–0.745)
  mean ≤0.53 29 9 0.692 (0.288–1.000) 0.525 (0.323–0.726) 0.494 (0.244–0.745)
  kurtosis ≤0.28 40 6 0.462 (0.025–0.795) 0.344 (0.152–0.536) 0.449 (0.183–0.715)
  skewness ≤−0.10 23 9 0.692 (0.288–1.000) 0.623 (0.427–0.819) 0.641 (0.375–0.907)

*Optimal cut off point was obtained from Youden index on ROC curve perc, percentile.

**Confidence interval use a Bonferroni corrected (1–0.05/30) confidence level.