Table 2 Reconstruction performance comparison between FIN and RH-M

From: Fourier Imager Network (FIN): A deep neural network for hologram reconstruction with superior external generalization

 

Lung (int.)

Pap smear (ext.)

Prostate (ext.)

Salivary gland (ext.)

Pap smear (transfer, int.)

Prostate (transfer, int.)

Salivary gland (transfer, int.)

FIN (M = 3)

Amp. SSIM

0.799 ± 0.034

0.780 ± 0.022

0.808 ± 0.026

0.785 ± 0.039

0.826 ± 0.024

0.822 ± 0.016

0.763 ± 0.046

Phase SSIM

0.710 ± 0.080

0.637 ± 0.067

0.526 ± 0.079

0.664 ± 0.067

0.818 ± 0.053

0.554 ± 0.064

0.596 ± 0.109

RH-M (M = 3)

Amp. SSIM

0.568 ± 0.020

0.635 ± 0.031

0.527 ± 0.033

0.526 ± 0.080

0.750 ± 0.037

0.591 ± 0.026

0.576 ± 0.073

Phase SSIM

0.497 ± 0.072

0.647 ± 0.067

0.325 ± 0.055

0.504 ± 0.151

0.779 ± 0.072

0.462 ± 0.090

0.515 ± 0.148

FIN (M = 4)

Amp. SSIM

0.812 ± 0.031

0.817 ± 0.022

0.815 ± 0.027

0.790 ± 0.047

0.813 ± 0.026

0.817 ± 0.024

0.757 ± 0.044

Phase SSIM

0.739 ± 0.075

0.732 ± 0.063

0.555 ± 0.074

0.701 ± 0.096

0.826 ± 0.054

0.564 ± 0.059

0.584 ± 0.123

RH-M (M = 4)

Amp. SSIM

0.567 ± 0.026

0.640 ± 0.037

0.518 ± 0.032

0.514 ± 0.087

0.723 ± 0.034

0.586 ± 0.025

0.569 ± 0.099

Phase SSIM

0.549 ± 0.071

0.613 ± 0.067

0.338 ± 0.062

0.473 ± 0.172

0.785 ± 0.056

0.444 ± 0.098

0.509 ± 0.153

  1. Mean and standard deviation values were calculated on test sets consisting of ~50 unique FOVs for each sample type. Int. internal generalization, ext. external generalization
Back to article page