Optimized Bowsher Model for PET Reconstruction

A convolutional neural network for anatomically-guided PET reconstruction.

We are sharing convolutional neural network parameters for anatomically guided reconstruction of positron emission tomography (PET) images. The reconstruction is tracer-agnostic and performed entirely in image space (i.e. without the need to access raw PET data), requiring only DICOM MR and PET images as input.

The anatomically guided reconstruction uses a Bowsher prior and relies on high-resolution MR images—typically T1—to improve the resolution of conventional PET images and significantly reduce partial voluming effects. The method has the potential to improve depiction of lesions or accuracy of quantification.

For a detailed description of the approach, see the related publication.

Related Publication

Schramm G, Rigie D, Vahle T, Rezaei A, Van Laere K, Shepherd T, Nuyts J, Boada F.
Approximating anatomically-guided PET reconstruction in image space using a convolutional neural network.
Neuroimage. 2021 Jan 1;224:117399. doi: 10.1016/j.neuroimage.2020.117399

Please cite this work if you are using the optimized Bowsher model in your research.

Get the Code

For Any Platform (pyapetnet)

A platform-independent version of the Bowsher-based anatomically guided reconstruction written in Python and Tensorflow is available on Github.

For Siemens mMR Platforms

This resource requires the Siemens work-in-progress (WIP) 1431 package, available from Siemens Healthineers.

The software available on this page is provided free of charge and comes without any warranty. CAI²R and NYU Grossman School of Medicine do not take any liability for problems or damage of any kind resulting from the use of the files provided. Operation of the software is solely at the user’s own risk. The software developments provided are not medical products and must not be used for making diagnostic decisions.

The software is provided for non-commercial, academic use only. Usage or distribution of the software for commercial purpose is prohibited. All rights belong to the author (Fernando Boada) and NYU Grossman School of Medicine. If you use the software for academic work, please give credit to the author in publications and cite the related publications.

Please spell out your affiliation (e.g. "New York University" rather than "NYU").


Questions about the Siemens-specific resource may be directed to Fernando Boada, PhD. Questions about the platform-independent resource (pyapetnet) may be raised as issues on Github.