Table 2 Performance metrics of the proposed dual-stream approach on the HAM10K dataset. The results are reported over 10 trials, including precision, recall, F1 score, accuracy, and specificity. The maximum, mean, standard deviation (Std), and confidence interval (CI) values are also provided.

From: A dual-stream deep learning framework for skin cancer classification using histopathological-inherited and vision-based feature extraction

Trials

Precision

Recall

F1

Accuracy

Specificity

Trial 1

93.16%

93.21%

93.19%

95.85%

91.37%

Trial 2

93.80%

93.88%

93.84%

96.29%

92.08%

Trial 3

94.20%

94.21%

94.20%

96.39%

91.87%

Trial 4

93.43%

93.48%

93.45%

96.30%

92.33%

Trial 5

94.20%

94.21%

94.21%

96.60%

92.66%

Trial 6

93.97%

93.88%

93.93%

96.16%

91.10%

Trial 7

94.12%

94.15%

94.13%

96.57%

92.37%

Trial 8

94.27%

94.21%

94.24%

96.55%

92.54%

Trial 9

93.62%

93.61%

93.62%

95.73%

90.52%

Trial 10

93.07%

93.15%

93.11%

96.10%

91.39%

Max

94.27%

94.21%

94.24%

96.60%

92.66%

Mean

93.78%

93.80%

93.79%

96.25%

91.82%

Std

0.00444

0.00415

0.00428

0.00298

0.00704

CI

0.00275

0.00257

0.00266

0.00184

0.00437