Skip to main content

Table 3 Diagnostic accuracy of for differentiating benign and malignant calcifications on spot magnification mammogram of in-house dataset

From: A YOLO-based AI system for classifying calcifications on spot magnification mammograms

 

AUC

Sen %

Spec %

Acc %

PPV %

YOLOv4-baseline on single view

0.766 (0.732–0.800)

78.6 (74.9–82.3)

70.4 (63.4–77.4)

74.5 (71.2–77.8)

77.1 (73.7–80.5)

YOLO-AMDF on single view

0.830 (0.810–0.850)

78.7 (76.1–81.3)

80.2 (77.9–82.5)

79.4 (77.1–81.7)

80.0 (77.8–82.2)

Proposed AI system on single view

0.867 (0.851–0.883)

83.0 (80.2–85.8)

81.0 (77.9–84.1)

82.0 (79.8–84.2)

82.3 (79.9–84.7)

Proposed AI system on two views

0.888 (0.868–0.908)

88.4 (86.9–89.9)

80.8 (77.6–84.0)

84.6 (81.8–87.4)

83.3 (80.8–85.8)

  1. The numbers in parentheses indicate 95% confidence intervals
  2. AUC, areas under the receiver operating characteristics curve; Sen, sensitivity; Spec, specificity; Acc, accuracy; PPV, positive predictive value; YOLOv4, You Only Look Once version 4; YOLO-AMDF, YOLO-Adaptive Multiscale Decision Fusion