Skip to main content
EURASIP Journal on Image and Video Processing

Table 1 Experimental results regarding the KiTS19 data set and our four Thai patients

From: Automatic kidney segmentation using 2.5D ResUNet and 2.5D DenseUNet for malignant potential analysis in complex renal cyst based on CT images

Model architecture

Averaged Dice score

 

KiTS train

KiTS val

Thai patients

Google Colab Pro:

   

   2.5D ResUNet (slice stack of 5)

0.9801

0.9373

0.6900

NVIDIA DGX A100:

   

   2.5D ResUNet (slice stack of 3)

0.9735

0.9554

0.7977

   2.5D ResUNet (slice stack of 5)

0.9772

0.9567

0.7335

   2.5D DenseUNet (slice stack of 3)

0.9779

0.9595

0.8367

   2.5D DenseUNet (slice stack of 5)

0.9769

0.9582

0.8760

Comparative results from [30]:

   

   3D U-Net

0.9734

-

-

   Residual 3D U-Net

0.9736

-

-

  1. The best result regarding each column is highlighted in bold
Back to article page