CT image reconstruction with jetson xavier agx

Hi, I am currently working on an image reconstruction piepline for x-ray ct imagaing. This is for use in a synchrotron radiation lab by an x-ray tomography beamline. I plan to do a filtered backprojection plus image denoising.

The image denoising part uses a neural net and I have been able to deploy it using pytroch and tensorrt

However the image reconstruction part has been unsuccessful so far. I tried building astra toolbox (https://github.com/astra-toolbox/astra-toolbox) on my xavier, and even after finishing the build and install I keep guetting Cuda error: no kernel image is available for execution on the device.

I tried building from source a pytorch extenstion named torch-radon (https://github.com/matteo-ronchetti/torch-radon). I encounter the same kernel image not found error even after compiling and installing it even after specifing 7.2 capability in the setup file.

Is it possible to find a tool that does filtered backprojection or any ct image reconstruction on the edge via jetson xavier? or is it only for neural network applications?


you may like to consider clara sdk medical imaging
Also discrete GPU will do better for heavy load like reconstruction maybe