Table 8 Ablation study of window size in GCA module.

From: Multi-branch CNN and grouping cascade attention for medical image classification

Window size

Params (M)

FLOPs (G)

Acc

F1

Precision

Recall

Auc

BUSI

5

25.2M

6.5G

0.9200

0.9050

0.9194

0.8948

0.9217

7

25.2M

6.4G

0.9333

0.9261

0.9326

0.9226

0.9404

9

25.2M

6.5G

0.9200

0.9049

0.9522

0.8730

0.9072

COVID19-CT

5

25.2M

6.5M

0.9054

0.9054

0.9710

0.8481

0.9096

7

25.2M

6.4G

0.9257

0.9317

0.9146

0.9494

0.9240

9

25.2M

6.5G

0.9122

0.9182

0.9125

0.9241

0.9113

BUSI

5

25.2M

6.5G

0.8569

0.8023

0.8113

0.7950

0.8734

7

25.2M

6.4G

0.8635

0.8090

0.8191

0.8012

0.8776

9

25.2M

6.5G

0.8560

0.8047

0.8122

0.7985

0.8748

  1. Bold indicates the optimal metric values among all compared methods.