Table 1 Summary of components and loss functions
From: Disentangled generative adversarial network for low-dose CT
Components | Loss function | Description |
|---|---|---|
DNSN-DF | minâ„’dfn | Dynamic filter network with MSE loss |
DNSN-1 | min(ℒdfn + ℒc) | Oriented-PSNR model with single input for RDN |
DNSN | min(ℒdfn + ℒc) | Oriented-PSNR model with multi-inputs for RDN |
DNSGAN-CS | \( \underset{G}{\min}\underset{D_k}{\max}\sum \limits_{k=1,2..}^K{\mathrm{\mathcal{L}}}_{GAN}^{Std}\left(G,{D}_k\right)+\lambda {\mathrm{\mathcal{L}}}_{dfn}+{\mathrm{\mathcal{L}}}_c+\eta {\mathrm{\mathcal{L}}}_{fea} \) | DNSGAN with cross entropy loss |
DNSGAN-NF | \( \underset{G}{\min}\underset{D_k}{\max}\sum \limits_{k=1,2..}^K{\mathrm{\mathcal{L}}}_{GAN}^{Ra}\left(G,{D}_k\right)+\lambda {\mathrm{\mathcal{L}}}_{dfn}+{\mathrm{\mathcal{L}}}_c \) | DNSGAN without improved feature loss |
DNSGAN-1 | \( \underset{G}{\min}\underset{D_k}{\max}\sum \limits_{k=1,2..}^K{\mathrm{\mathcal{L}}}_{GAN}^{Ra}\left(G,{D}_k\right)+\lambda {\mathrm{\mathcal{L}}}_{dfn}+{\mathrm{\mathcal{L}}}_c+\eta {\mathrm{\mathcal{L}}}_{fea} \) | DNSGAN with single-input for RDN module |
DNSGAN | \( \underset{G}{\min}\underset{D_k}{\max}\sum \limits_{k=1,2..}^K{\mathrm{\mathcal{L}}}_{GAN}^{Ra}\left(G,{D}_k\right)+\lambda {\mathrm{\mathcal{L}}}_{dfn}+{\mathrm{\mathcal{L}}}_c+\eta {\mathrm{\mathcal{L}}}_{fea} \) | DNSGAN with multi-input for RDN module |