Skip to main content

Table 6 Ablation Study of Feature Fusion Strategies (tested based on scenario 1)

From: Fusion-driven semi-supervised learning-based lung nodules classification with dual-discriminator and dual-generator generative adversarial network

Feature Fusion Strategy

Accuracy (%)

Precision (%)

Recall (%)

F1-Score (%)

No Fusion (Separate Outputs)

85

83

86

84

Simple Averaging

88

86

90

88

Concatenation + Dense Layer

91

89

94

91

Weighted Feature Averaging

90

88

93

90

Proposed Fusion (Attentive Map)

92

90

96

93