Skip to main content

Table 4 Performances of the decision support system for the two patient cohorts and the two examiners

From: A decision support system based on radiomics and machine learning to predict the risk of malignancy of ovarian masses from transvaginal ultrasonography and serum CA-125

  

Examiner 1

Examiner 2

First cohort (239 women)

Sensitivity

122/123, 99.2% (95.6–100.0)

121/123, 98.4% (94.3–99.8)

Specificity

88/116, 75.9% (67.0–83.3)

91/116, 78.5% (69.9–85.5)

Accuracy

210/239, 87.9% (83.0–91.7)

212/239, 88.7% (84.0–92.4)

PPV

122/150, 81.3% (74.2–87.2)

121/146, 82.9% (75.8–88.6)

NPV

88/89, 98.9% (93.9–100.0)

91/93, 97.9% (92.5–99.7)

Second cohort (35 women)

Sensitivity

20/20, 100.0% (83.2–100.0)

20/20, 100.0% (83.2–100.0)

Specificity

12/15, 80.0% (51.9–95.7)

12/15, 80.0% (51.9–95.7)

Accuracy

32/35, 91.4% (76.9–98.2)

32/35, 91.4% (76.9–98.2)

PPV

20/23, 86.967% (66.4–97.2)

20/23, 86.967% (66.4–97.2)

NPV

12/12, 100.0% (74.5–100.0)

12/12, 100.0% (74.5–100.0)

  1. Data are given as ratios, percentages (95% confidence intervals). Note: two of the 35 ovarian masses were defined differently by the two examiners (solid for the first examiner, mixed for the second): however, the decision support system classified them with the same level of risk (medium-high risk)