UPenn

My Postdoctoral research at UPenn focused on fusing novel image acquisition techniques with the rapid advances in deep learning to revolutionize spectral CT. Areas of major clinical application include dual-contrast protocols and hepatic imaging.

Spectral CT via slow kVp switching

This project aims at infusing a novel CT image acquisition technique with the rapid advances in deep learning to revolutionize spectral CT. Despite considerable investigations on spectral CT, its clinical applications are impeded by the requirement of dedicated hardware. We investigated an alternative solution. It is inspired by fast kVp switching, but uses a much slow kVp switching rate so it can be implemented on conventional CT systems (single spectrum, energy integrating detector). This leads to several challenges in image processing. First, data is sparsely sampled in each energy level. Second, performing the material decomposition in the sinogram domain is not directly possible as the projection images of different energy levels are not angularly aligned. These challenges were overcome by deep learning. Multiple well-known (U-Net CNN, Pix2Pix GAN) and original neural network architectures were employed to perform sparse sinogram completion and material decomposition.

Cao W., Shapira N., Maidment A., Daerr H., Noël P. "Hepatic dual-contrast CT imaging: slow triple kVp switching CT with CNN-based sinogram completion and material decomposition" Journal of Medical Imaging 9.1 (2022): 014003.

Cao W., Shapira N., Noël P. "GAN-based sinogram completion for slow triple kVp switching CT" Medical Imaging 2021: Physics of Medical Imaging, 2021

Shapira N., Cao W., Liu L., Leiner T., Smits M., Noël P. "Dual-contrast decomposition of dual-energy CT using convolutional neural networks" Medical Imaging 2021: Physics of Medical Imaging, 2021

Cao W., Shapira N., Noël P. "Slow Triple kVp Switching CT with Convolutional Neural Network Based Sinogram Completion and Material Decomposition" 6th International Conference on Image Formation in X-Ray Computed Tomography (CT meeting 2020).