Due to radiation restrictions, Low Dose Computed Tomography (LDCT) has offered tremendous benefits in clinical applications. In this model an adversarially trained network and a sharpness detection network were trained for guiding training process. This deals with the problem of blur effect on final reconstructed denoised image. Experiment was performed on both simulated and real dataset. Simulated dataset has 239 CT images downloaded from the National Cancer Imaging Archive (NBIA). Real dataset has CT scans of a deceased piglet. The model consists of three networks, the generator, the discriminator and sharpness detector network. The generator learns a mapping from LDCT image to denoised image which is expected to be more similar with original CT image. Detector differenciates the virtual image with original one. Generator tried to create a virtual CT image that can fool Detector and Detector tries to not get foooled. The sharpness detector explicitly evaluate the denoised image's sharpness.
Input variables : Low dose CT scan image
Output Variables : Denoised CT image
Visit Model : github.com
Additional links : arxiv.org
| Model Category | : | Public |
| Date Published | : | September, 2017 |
| Healthcare Domain | : | Medical Technology |
| Code | : | github.com |
| Medical Imaging |
| Image Denoising |