Skip to main content

Table 2 Reliability of contrast features from the model for malignancy prediction

From: Characterization of a murine xenograft model for contrast agent development in breast lesion malignancy assessment

Contrast features (Method)   Accuracy Sensitivity Specificity PPV NPV
Molecular targeting contrast CD31 (IHC) 100 % 100 % 100 % 100 % 100 %
VEGF (IHC) 100 % 100 % 100 % 100 % 100 %
Non-targeting contrast Rim enhancement (DCE-MDCT) 93.5 % 95.7 % 91.3 % 91.7 % 95.5 %
Time-density curve 91.3 % 86.9 % 100 % 100 % 88.5 %
Washout T1/2 < 100 min (DCE-MDCT)
Endogenous contrast Angular/ill-defined margin (grey scale US) 92.6 % 85.2 % 100 % 100 % 100 %
  Angular/ill-defined margin (DCE-MDCT) 93.5 % 87 % 95.7 % 95.2 % 87 %
  Orientation width/height > 2.2 (grey scale US) 100 % 100 % 100 % 100 % 100 %
  Neovascular flow (color Doppler US) 57.4 % 14.8 % 100 % 100 % 54 %
  1. TPMDA = number of true positive (indicating malignancy) of MDA-MB 231 xenografts; TPMCF = number of true positive (indicating malignancy) of MCF-7 xenografts; TN = number of true negative (not indicating malignancy) of benign xenografts; FP = number of false positive (indicating malignancy) of benign xenografts; FNMDA = number of false negative (not indicating malignancy) of MDA-MB 231 xenografts; FNMCF = number of false negative (not indicating malignancy) of MCF-7 xenografts; accuracy = (TPMDA + TPMCF + TN)/(TPMDA + TPMCF + FNMDA + FNMCF + TN + FP) × 100 %; sensitivity = (TPMDA + TPMCF)/(TPMDA + TPMCF + FNMDA + FNMCF); specificity = TN/(TN + FP); PPV (positive predictive value) = (TPMDA + TPMCF)/(TPMDA + TPMCF + FP); NPV (negative predictive value) = (TN)/(FNMDA + FNMCF + TN); washout T1/2 = half-life of washout-like contrast pattern