X-ray Computed Tomography Sinogram Data Generation from Microwave Tomography Measurements Using Deep Neural Networks Conference

Istiak, MA, Hasnine, IM, Kiourti, A et al. (2022). X-ray Computed Tomography Sinogram Data Generation from Microwave Tomography Measurements Using Deep Neural Networks . 164-167. 10.1109/ICECE57408.2022.10088739

cited authors

  • Istiak, MA; Hasnine, IM; Kiourti, A; Alam, MS; Islam, MA

abstract

  • X-ray computed tomography (CT) has been a widespread medical imaging modality for decades mainly due to its high imaging resolution. Compared to X-ray CT, microwave imaging (MWI) has several advantages, such as avoiding ionizing radiation, offering lower-cost portable hardware, etc. However, MWI lacks in terms of spatial resolution and robustness. In this work, we aim to bridge this gap by exploiting the strengths of MWI measurements and X-ray CT spatial resolution and combine them using a deep neural network (DNN). To our knowledge, for the first time, the trade-off between these two modalities is being addressed simultaneously. Using a customized Shepp-Logan phantom to emulate a simplified healthy human brain, we generated MWI data employing realistic dielectric properties of the head. Concurrently, the corresponding parallel beam X-ray CT data was generated. Next, a DNN comprised of several sub-networks was trained to successfully estimate the CT sinograms directly from the MWI measurements. Finally, the efficacy of the proposed method has been demonstrated through the successful estimation of CT sinograms (and subsequent image reconstructions) from MWI measurements for several thousand imaging scenarios.

publication date

  • January 1, 2022

Digital Object Identifier (DOI)

start page

  • 164

end page

  • 167