Table 3 Ablation Study on ISIC 2016, ISIC 2017 & ISIC 2018 Datasets.

From: ARCUNet: enhancing skin lesion segmentation with residual convolutions and attention mechanisms for improved accuracy and robustness

Model

Accuracy

Val acc

Dice coef

Val dice coef

Jaccard

Val jaccard

ISIC 2016

 UNet

0.9399

0.9115

0.8884

0.8689

0.8566

0.8265

 UNet + RC

0.9791

0.9589

0.9293

0.9099

0.9009

0.8889

 UNet + RC + AM

0.9932

0.9812

0.9680

0.9468

0.9345

0.9114

ISIC 2017

 UNet

0.9199

0.8925

0.8642

0.8389

0.8566

0.8365

 UNet + RC

0.9491

0.9189

0.9193

0.8999

0.8989

0.8589

 UNet + RC + AM

0.9831

0.9645

0.9321

0.9121

0.9065

0.8834

ISIC 2018

 UNet

0.8429

0.8225

0.9284

0.9089

0.8890

0.8665

 UNet + RC

0.9651

0.9389

0.9593

0.9399

0.9289

0.9189

 UNet + RC + AM

0.9909

0.9819

0.9845

0.9534

0.9566

0.9353

Back to article page