Table 3 Quantitative evaluation results depending on architectures.

From: Dual-encoder architecture for metal artifact reduction for kV-cone-beam CT images in head and neck cancer radiotherapy

No. of encoders

Encoders

Feature fusion method

Evaluation metrics

Encoder 1

Encoder 2

PSNR

MSE

SSIM

PCC

Single-encoder

UNet

N/A

N/A

41.003026

0.000107

0.991796

0.992719

Fusion Net

N/A

N/A

39.921392

0.000142

0.990347

0.991159

UNet/L

N/A

N/A

39.301266

0.000147

0.990335

0.990836

FusionNet/L

N/A

N/A

39.991436

0.000130

0.991024

0.991741

Dual-encoder

UNet

FusionNet

Simple addition

40.430324

0.000118

0.991504

0.992012

UNet

FusionNet

Spatial attention

40.726800

0.000113

0.992119

0.992379

UNet

FusionNet

Spatial and channel wise attention

41.859071

0.000086

0.993072

0.993681

Back to article page