Table 11 Few-shot learning performance with statistical measures.

From: Novel metaheuristic optimized latent diffusion framework for automated oral disease detection in public health screening

Samples per pathology

DentoSMART-LDM (mean ± SD)

95% CI

Enhanced-PSO-LDM (mean ± SD)

95% CI

GA-Diffusion (mean ± SD)

95% CI

Traditional enhancement (mean ± SD)

95% CI

No enhancement (mean ± SD)

95% CI

t-test p-value

2 Samples

89.2 ± 1.7%

[88.5, 89.9]

78.6 ± 2.3%

[77.7, 79.5]

75.3 ± 2.5%

[74.3, 76.3]

61.4 ± 2.8%

[60.3, 62.5]

52.7 ± 3.1%

[51.5, 53.9]

p < 0.001

5 Samples

92.8 ± 1.4%

[92.3, 93.3]

84.7 ± 1.9%

[84.0, 85.4]

81.9 ± 2.1%

[81.1, 82.7]

69.8 ± 2.5%

[68.8, 70.8]

62.1 ± 2.7%

[61.0, 63.2]

p < 0.001

10 Samples

95.1 ± 1.2%

[94.7, 95.5]

89.3 ± 1.6%

[88.7, 89.9]

86.7 ± 1.8%

[86.1, 87.3]

75.6 ± 2.2%

[74.7, 76.5]

68.9 ± 2.4%

[67.9, 69.9]

p < 0.001

20 Samples

96.4 ± 1.0%

[96.1, 96.7]

91.8 ± 1.4%

[91.3, 92.3]

89.4 ± 1.6%

[88.9, 89.9]

79.3 ± 2.0%

[78.5, 80.1]

73.2 ± 2.2%

[72.3, 74.1]

p < 0.001

50 Samples

96.9 ± 0.9%

[96.6, 97.2]

93.2 ± 1.2%

[92.7, 93.7]

91.1 ± 1.4%

[90.6, 91.6]

82.7 ± 1.8%

[82.0, 83.4]

76.8 ± 2.0%

[76.0, 77.6]

p < 0.001

Full dataset

97.3 ± 0.8%

[97.0, 97.6]

93.8 ± 1.1%

[93.4, 94.2]

91.4 ± 1.3%

[90.9, 91.9]

87.6 ± 1.7%

[86.9, 88.3]

81.4 ± 1.9%

[80.6, 82.2]

p < 0.001

Back to article page