Vector Quantized Convolutional Autoencoder for Low-Dose CT Image Reconstruction with Perceptual and Bias Reducing Loss.