Table 10 Computational efficiency comparison with statistical measures.

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

Method

Parameters (M)

FLOPs (G)

Training time (hours ± SD)

95% CI training

Inference time (ms/image ± SD)

95% CI inference

GPU memory (GB ± SD)

95% CI memory

Efficiency score*

t-test p-value

DentoSMART-LDM

68.7

82.3

14.2 ± 1.8

[13.3, 15.1]

11.8 ± 1.4

[11.1, 12.5]

7.4 ± 0.8

[7.0, 7.8]

9.2/10

Reference

Enhanced-PSO-LDM

79.4

101.7

19.8 ± 2.3

[18.7, 20.9]

16.3 ± 1.9

[15.4, 17.2]

9.2 ± 1.1

[8.7, 9.7]

7.8/10

p < 0.01

GA-Diffusion

86.8

118.6

23.4 ± 2.8

[22.1, 24.7]

19.7 ± 2.3

[18.6, 20.8]

10.8 ± 1.3

[10.2, 11.4]

7.1/10

p < 0.001

DE-Enhancement

74.2

89.4

18.1 ± 2.1

[17.2, 19.0]

14.9 ± 1.7

[14.1, 15.7]

8.6 ± 0.9

[8.1, 9.1]

8.0/10

p < 0.01

Stable diffusion

91.3

127.8

28.7 ± 3.4

[27.0, 30.4]

22.4 ± 2.6

[21.2, 23.6]

12.5 ± 1.5

[11.8, 13.2]

6.3/10

p < 0.001

DALL-E 2

112.6

159.3

38.9 ± 4.2

[36.8, 41.0]

31.2 ± 3.1

[29.7, 32.7]

16.8 ± 1.8

[15.9, 17.7]

4.8/10

p < 0.001

MedDiffusion

83.7

109.2

21.6 ± 2.5

[20.4, 22.8]

18.1 ± 2.1

[17.1, 19.1]

9.9 ± 1.2

[9.4, 10.4]

7.4/10

p < 0.001

PathoDiff

81.4

105.8

20.3 ± 2.4

[19.2, 21.4]

17.4 ± 2.0

[16.5, 18.3]

9.5 ± 1.1

[9.0, 10.0]

7.6/10

p < 0.01

Traditional enhancement

8.9

12.4

1.3 ± 0.2

[1.2, 1.4]

2.1 ± 0.3

[1.9, 2.3]

1.8 ± 0.2

[1.7, 1.9]

8.5/10**

p < 0.001

Back to article page