Skip to main content

Table 4 Diagnostic performance of four models in the evaluation of severe renal function impairment

From: MRI texture-based machine learning models for the evaluation of renal function on different segmentations: a proof-of-concept study

 

Training cohort

Validation cohort

Models

AUC

Sensitivity (%)

Specificity (%)

Accuracy (%)

AUC

Sensitivity (%)

Specificity (%)

Accuracy (%)

All-K

0.974 (0.967–0.982)

78.3 (77.8–78.7)

86.6 (86.0–87.2)

85.0 (84.4–85.6)

0.938 (0.934–0.941)

54.5 (53.5–55.6)

95.1 (94.7–95.6)

73.1 (72.5–73.6)

LC-K

0.959 (0.956–0.962)

87.0 (85.8–88.1)

90.7 (89.5–92.0)

90.0 (89.2–90.8)

0.863 (0.857–0.887)

72.7 (72.1–73.4)

87.8 (86.2–89.4)

84.6 (84.0–85.3)

TLCO-K

0.970 (0.964–0.977)

73.9 (73.4–74.4)

92.8 (90.9–94.7)

89.2 (88.5–89.8)

0.314 (0.201–0.426)

0 (-0.2–0.2)

85.3 (84.4–86.2)

67.3 (66.7–67.9)

PIZZA-K

0.994 (0.986–1.00)

82.6 (81.9–83.3)

76.2 (76.1–76.5)

77.5 (76.8–78.2)

0.688 (0.673–0.704)

9.1 (8.8–9.4)

78.0 (77.5–78.6)

80.6 (79.8–81.3)

  1. Data in parentheses are 95% confidence interval