Low-Dose CT with A Residual Encoder-Decoder Convolutional Neural Network RED-CNN

• Feng Lin  |  Ge Wang  |  Hu Chen  |  Jiliu Zhou  |  Mannudeep K. Kalra  |  Peixo Liao  |  Yang Chen  |  Yi Zhang

• College of Computer Science Sichuan University China

For low-dose CT imaging, the model combines the autoencoder, deconvolution network, and shortcut connections into the residual encoder-decoder convolutional neural network (RED-CNN). After patch-based training, the model achieves a competitive performance relative to the-state-of-art methods in both simulated and clinical cases. The model has been favorably evaluated in terms of noise suppression, structural preservation, and lesion detection at a high computational speed. Normal-dose CT images of patients from National Biomedical Imaging Archive (NBIA) is used for training.

Input variables : CT image
Output Variables : Denoised CT image

Visit Model : github.com

Additional links : arxiv.org