Table 32 Performance under natural variability conditions.

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

Variability factor

Prevalence in independent data (95% CI)

Performance impact (mean ± SD)

Mitigation effectiveness (95% CI)

Baseline accuracy

Enhanced accuracy

Recovery rate

χ² statistic

p-value

Equipment aging (> 10 years)

34.7% (32.1–37.3)

-3.2 ± 1.1% accuracy

87.3% (84.8–89.8)

82.4%

88.7%

89.2%

156.34

< 0.001

Operator inexperience (< 2 years)

23.8% (21.6–26.0)

-4.1 ± 1.4% accuracy

89.6% (87.2–92.0)

80.9%

89.1%

91.4%

142.78

< 0.001

Patient positioning errors

41.2% (38.7–43.7)

-2.8 ± 0.9% accuracy

91.4% (89.1–93.7)

83.7%

89.8%

92.1%

167.92

< 0.001

Exposure variations

29.6% (27.2–32.0)

-3.7 ± 1.2% accuracy

88.9% (86.4–91.4)

81.6%

88.4%

88.7%

134.56

< 0.001

Motion artifacts

18.3% (16.4–20.2)

-5.2 ± 1.6% accuracy

85.7% (82.8–88.6)

78.3%

86.9%

85.9%

123.45

< 0.001

Environmental interference

12.4% (10.8–14.0)

-2.3 ± 0.8% accuracy

93.1% (90.9–95.3)

84.9%

90.2%

94.1%

89.67

< 0.001

Mixed pathological presentations

67.8% (65.4–70.2)

-1.9 ± 0.7% accuracy

94.8% (93.1–96.5)

85.6%

91.3%

95.2%

198.23

< 0.001