Table 1 Improvement of FAMO in per-image and per-fracture level.

From: The Feature Ambiguity Mitigate Operator model helps improve bone fracture detection on X-ray radiograph

Network

Per-image sensitivity

Per-image specificity

Youden Index (%)

AUC (%)

Per-fracture AP (%)

ResNeXt101+FPN

579/935

(61.9%)

801/875

(91.5%)

53.4

74.9

76.8

CI: (58.7%, 65.0%)

CI: (89.5%, 93.3%)

CI: (48.2, 58.6)

CI: (74.1, 75.7)

CI: (76.1, 77.4)

FAMO+ResNext101+FPN

603/935

(64.5%)

813/875

(92.9%)

57.4

77.5

77.4

CI: (61.3%, 67.5%)

CI: (91.0%, 94.5%)

CI: (52.3, 62.5)

CI: (76.5, 78.5)

CI: (76.6, 78.2)

  1. Note: ResNeXt101, a network structure; FPN, feature pyramid network; FAMO, Feature Ambiguity Mitigate Operator (FAMO) method; AUC, area under the curve; AP, average precision; CI, confidence interval. All CIs were calculated at the two-tailed significance of 0.05.
Back to article page