Fine-Grained Lung Cancer Classification from PET and CT Images Based on Multidimensional Attention Mechanism

Table 2

Evaluation of different attention mechanisms.

Network architecture	Data	AUC				Balanced accuracy
		Score of each category		Average score	Variance
MFSE-DenseNet (r = 4)	CT + PET	LUSC	0.867	0.860	0.03	0.62
		LUAD	0.876
		SCLC	0.837
MFSE-DenseNet (r = 16)	CT + PET	LUSC	0.809	0.851	0.09	0.63
		LUAD	0.774
		SCLC	0.969
MFSA-DenseNet	CT + PET	LUSC	0.860	0.890	0.04	0.67
		LUAD	0.961
		SCLC	0.850
MFSCA-DenseNet (r = 4)	CT + PET	LUSC	0.938	0.920	0.05	0.72
		LUAD	0.910
		SCLC	0.913