NeCT: Neural Computed Tomography

Paper Code

Official implementation for the paper Implicit neural representation for fast 4D computed tomography of multiphase flow in porous media

NeCT leverages deep learning to improve computed tomography (CT) image quality, supporting both static and dynamic CT reconstruction. The project is a collaboration between the Norwegian Univeristy of Science and Technology (NTNU) and the CT lab at Equinor. The project is based on the INR framework tiny-cuda-nn.

• User guide
• Installation
• Quick start
• Geometry configuration
• Configuration
• Demo
• Licensing and Citation

Showcase: Bentheimer Sandstone Imbibition

Rendering of spontaneous imbibition in a Bentheimer sandstone reconstructed using NeCT. The brine flowing into the sample is shown in light blue, while the salt grains dissolving are presented in red. The video recording was started when brine entered the core.

Rendering of the dissolution of a salt grain. Three orthogonal slices visualize its temporal evolution. In the xz slice, it is possible to observe the brine coming into contact with the salt before it starts to dissolve.