Table 2 Diagnostic Performance of Reviewers and DL-DSS.

From: Deep learning-based decision support system for the diagnosis of neoplastic gallbladder polyps on ultrasonography: Preliminary results

 

AUC

Comparison

Sensitivity

Specificity

Accuracy

F-1 Score

Step 1: Initial Image Analysis

Step 1 vs Step 2 (p value)

    

Reviewer A

0.94 [0.88–0.98]

0.49

88.6 (31/35) [73.3–96.8]

85.7 (54/63) [74.6–93.3]

86.7 (85/98) [74.1–94.6]

0.827 [0.737–0.870]

Reviewer B

0.78 [0.68–0.85]

<0.01

71.4 (25/35) [53.7–85.4]

68.3 (43/63) [55.3–79.4]

69.4 (68/98) [54.7–81.5]

0.624 [0.509–0.704]

Reviewer C

0.87 [0.79–0.93]

0.11

97.1 (34/35) [85.1–99.9]

65.1 (41/63) [52.0–76.7]

76.5 (75/98) [63.8–85.0]

0.747 [0.687–0.760]

Step 2: DL-DSS Alone

     

DL-DSS

0.92 [0.85–0.97]

 

74.3 (26/35) [56.7–87.5]

92.1 (58/63) [82.4–97.4]

85.7 (84/98) [73.2–93.9]

0.788 [0.663–0.867]

Step 3: DL-DSS Aided

Step 1 vs Step 3 (p value)

    

Reviewer A

0.95 [0.88–0.98]

0.65

85.7 (30/35) [69.7–95.2]

93.7 (59/63) [84.5–98.2]

90.8 (89/98) [79.2–97.1]

0.869 [0.770–0.921]

Reviewer B

0.91 [0.83–0.96]

<0.01

80.0 (28/35) [63.1–91.6]

93.7 (59/63) [84.5–98.2]

88.8 (87/98) [76.9–95.8]

0.836 [0.723–0.902]

Reviewer C

0.91 [0.83–0.96]

0.17

91.4 (32/35) [76.9–98.2]

71.4 (45/63) [58.7–82.1]

78.6 (77/98) [65.2–87.9]

0.753 [0.673–0.787]

  1. Note.- Data in brackets are 95% CI.