Table 2 Review of the papers that applied deep learning for kidney vessel segmentation.

From: Deep learning for 3D vascular segmentation in hierarchical phase contrast tomography: a case study on kidney

References

Modality

Data source

No. of subj.

ML model

Input

DSC

Add’l perf. metrics

Karpinski et al.60

WSI

Publicly available

35

2D UNet with

Resnet34 backbone

2D images

–

Acc: 0.893

He et al.61

CT

Private

170

DPA-DenseBiasNet

3D volumes

0.861

MCD:1.976

Taha et al.62

CT

Private

99

Kid-Net (a 3D CNN)

3D patches

–

F1 score: 0.72 (artery);

0.67 (vein)

He at al.63

CTA

Private

122

EnMcGAN

3D patches

0.89 ± 0.6 (artery);

0.77 ± 0.12 (vein)

–

Zhang et al.64

CT

Publicly available

392

DPA-DenseBiasNet

2D images

0.884

–

Xu et al.65

Micro-CT

Private

8

CycleGAN

3D patches

0.768 ± 0.3

Acc: 0.992

Li et al.66

CT

Publicly available

35

DUP-Net

3D patches

0.883

Precision:0.911;

Recall:0.858

  1. CNN convolutional neural network, GAN generative adversarial network, WSI whole slide imaging, DPA deep priori anatomy, MCD mean centerline distance, EnMcGAN ensemble multi-condition GAN, DUP-Net double UPoolFormer networks, DSC Dice similarity coefficient.
Back to article page