Table 1 Ablation study showing performance of different DL models before and after modification of network backbone.

DenseNet-121

Modified (DiaNet v2)

0.9048

0.8925

0.9172

0.9134

0.9021

0.8108

Backbone

0.8108

0.8181

0.8036

0.8068

0.8123

0.6218

p value

5.669E−154

5.888E−183

9.462E−147

7.443E−136

7.571E−173

2.677E−136

ResNet-50

Modified (DiaNet v2)

0.9213

0.9184

0.9241

0.9231

0.9205

0.843

Backbone

0.9059

0.9065

0.9053

0.9068

0.9058

0.8132

p value

1.666E−144

7.275E−139

2.175E−113

7.199E−117

3.827E−142

1.332E−144

EfficientNet

Modified (DiaNet v2)

0.8812

0.8556

0.9069

0.9011

0.8776

0.7637

Backbone

0.874

0.8735

0.8746

0.8765

0.874

0.7497

p value

3.625E−153

1.764E−106

7.987E−149

1.053E−160

1.027E−155

2.261E−152

VGG-11

Modified (DiaNet v2)

0.9263

0.9393

0.9132

0.9176

0.9281

0.8532

Backbone

0.8914

0.9446

0.838

0.8781

0.9041

0.7978

p value

2.375E−232

1.574E−235

9.616E−17

9.175E−29

2.505E−206

3.565E−187

MobileNet_v2

Modified (DiaNet v2)

0.8842

0.8918

0.8765

0.8784

0.8849

0.7687

Backbone Network

0.8822

0.8814

0.8829

0.883

0.8821

0.7646

p value

4.069E−159

6.224E−158

4.146E−74

8.861E−80

7.673E−150

9.739E−160