Algorithm 1: Generating LDCT from existing retrospective CT scans

1:$HU$: Dequantise Hounsfeld unit (HU) values of original DICOM scan by adding uniform noise [0, 1) 2:$\mu$: Compute the linear attenuations values $\mu =HUx0.001x({\mu }_{water}-{\mu }_{air})+{\mu }_{water}$ using $\mu$ values of the deiserd dose tube voltage, then normalize values in range [0, 1] 3:R: Project the data using the ray transform (Radon transform in 2D) for CT 4:T:Convert the high-dose sinogram data to the transmission one $T=e^{(-R)}$ 5:$T_{LD}$: Generate the simulated low-dose transmission data by injecting Poisson and Gaussian noise. The Poisson noise models the quantum noise that stems from the process of photon generation, attenuation and detection, while Gaussian noise models the detector noise that stems from the electronic data acquisition system 6:$R_{LD}$: Convert the low-dose transmission data to sinogram data $R_{LD}=-ln\left(T_{LD}\right)$ 7:${\mu }_{LD}$: Ray Data back Projection to obtain linear attenuations $\mu$ for the low-dose CT 8:$H{U}_{LD}$: Compute the new HU value for the low-dose CT $H{U}_{LD}=1000x(\mu -{\mu }_{water})/({\mu }_{water}-{\mu }_{air})$