zax0s/ctlesspet

Synthetic CT from NAC PET images

CTlessPET

Install

pip install CTlessPET

Use

CTlessPET only requires an NAC-PET dataset and a CT dataset. The CT is used as the container for the synthetic CT, and can be an empty CT acquired before the patient enters the scanner. The NAC-PET should be reconstructed using OSEM with Time-of-Flight enabled but no PSF modeling. Reconstruction at should be at 440x440 matrix size with a 4 mm Gaussian post filter. Only Siemens Bigraph Vision scanners (including the Quarda) are supported.

Dicom data

Using a folder containing both NAC and CT data:

CTlessPET -i <input_folder> --output <output_folder>

or in seperate folders:

CTlessPET -i <input_NAC_folder> --CT <input_CT_folder> --output <output_folder>

Nifti data

CTlessPET -i <input_NAC_nii> --CT <input_CT_nii> --output <output_nii>

Choice of model

The network has been trained for FDG-PET (adult and pediatric) as well as H20-PET.

The type is automatically selected when running the model with dicom data. You can overwrite the choice of the model using the --model flag, e.g. --model FDG_Pediatric.

Optional arguments

You can change the batch size using --batch_size as well as overwrite the dose (--dose) and weight (--weight) given to the patient. This is otherwise automatically read from the dicom file (if supplied).

Installation with Cuda12

Install ONNX Runtime GPU (CUDA 12.x): pip install onnxruntime-gpu --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/

https://onnxruntime.ai/docs/install/

Citation

If you are using CTlessPET, please cite the following Diganostics paper:

Montgomery ME, Andersen FL, d’Este SH, Overbeck N, Cramon PK, Law I, Fischer BM, Ladefoged CN. Attenuation Correction of Long Axial Field-of-View Positron Emission Tomography Using Synthetic Computed Tomography Derived from the Emission Data: Application to Low-Count Studies and Multiple Tracers. Diagnostics. 2023; 13(24):3661. https://doi.org/10.3390/diagnostics13243661