Encrypted login | home

Program Information

A New Version of Multi Material Decomposition (MMD) Algorithm for DECT

no image available
F Mirzaei

R Faghihi1 , F Mirzaei2*, (1) School of mechanical engineering, Shiraz University, Shiraz,Iran (2) Radiation Research center, Shiraz University, Shiraz, Iran


MO-RAM-GePD-I-1 (Monday, July 31, 2017) 9:30 AM - 10:00 AM Room: Imaging ePoster Lounge

Purpose: Two different energy spectra are used for scanning objects in Dual-Energy Computed Tomography, DECT. Using the two energies, two materials, basis material, can be separated and density image pairs can be created. However, in computed tomography, decomposing more than two materials are necessary. To meet these need, different algorithms has been proposed in recent years.

Methods: In this investigation a new version of Multi Material Decomposition (MMD) algorithm is proposed. Barycentric coordinates were chosen an innovative local clustering method in our new implementation. Local clustering method, increases the precision in the Barycentric coordinates assignment by minifying search domain. These coordinates form a triangle in attenuation coefficient domain based on our material library. For optimizing these triangles, a fast bi-directional Hausdorff distance measurement is used. Noise is an important issue in the reconstruction of images. This problem addressed by means of a pixel-wise local adaptive filter which explore statistical property of each pixel neighborhood and degrade additive noise power. For more detailed assessment, the proposed method was applied on computed tomography of patients and QA phantom obtained by 80, and 140kVp x-rays.

Results: Considering anatomical expectation, decomposed images of the patients have an excellent agreement with anatomical expectation. Different portions of CT phantom were also decomposed without any significant noise.

Conclusion: Although available algorithms suppress the noise both before and after decomposition process, they don’t fully consider anatomical expectation, decomposed images have an excellent agreement. The pixel-wise local adaptive filter used in this study can degrade well the noise power.

Contact Email: