Table 1 A summary of the model comparisons

From: Development and multicenter validation of an AI driven model for quantitative meibomian gland evaluation

Algorithms

IoU/% (95% CI)

Dice (95% CI)

Recall (95% CI)

ACC (95% CI)

UNet42

81.67 (81.03–82.31)

89.72 (89.30–90.14)

90.48 (98.34–98.57)

97.49 (97.38–97.62)

Unet++43

78.85 (76.83–80.87)

87.9 (86.59–89.21)

88.27 (85.76–90.78)

97.38 (97.05–97.71)

Unet3+33

75.39 (73.25–77.53)

85.55 (84.08–87.02)

86.02 (83.84–88.20)

96.9 (96.57–97.23)

U2Net44

79.69 (78.97–80.41)

88.48 (88.02–88.94)

88.68 (87.91–89.46)

97.21 (97.07–97.35)

AttentionUnet45

79.64 (78.08–81.22)

88.4 (87.30–89.50)

89.4 (86.05–92.75)

97.47 (97.16–97.78)

HRnet46

79.24 (77.34–81.14)

88.16 (86.89–89.43)

88.63 (87.45–89.81)

97.43 (97.08–97.78)

DenseUNet47

79.61 (78.32–80.90)

88.39 (87.39–89.39)

88.51 (87.40–89.62)

97.49 (97.25–97.71)

Swim-Unet48

39.5 (36.02–42.98)

55.87 (52.24–59.51)

85.79 (82.74–88.84)

83.73 (82.74–88.84)

CE-net49

80.96 (80.19–81.73)

89.3 (88.88–89.72)

89.23 (88.21–90.25)

97.4 (97.28–97.52)

MNet50

37.80 (37.34–42.27)

56.4 (53.96–59.03)

67.2 (59.99–74.58)

87.0 (86.48–87.68)

U-net333

80.88 (80.13–81.64)

89.22 (88.74–89.70)

89.75 (88.26–91.24)

97.39 (97.29–97.48)