Predict CT image of MRI data through feature matching

Predict CT image of MRI data through feature matching

Mohammad esmaili1 Nima Farajian2

1) Graduate Student of Artificial Intelligence, Ivanki University
2) Dr. Nima Farajian Assistant Professor, Ivanki University

IOI Code : XBKF-EBZDD
Publication : Third National Conference on Electrical and Computer Engineering(iscelec.ir)
Abstract :
    Abstract The robustness correction for hybrid PET / MR imaging systems and dose planning for MR-based radiation are challenging due to the lack of high-energy photons. Damping information We propose a novel approach that utilizes nonlinear localized and predictive features. Pseudo-CT images from T1w and T2w MRI data Nonlinear local descriptors are obtained by linear projectionDescriptions to nonlinear nonlinear space using that map Explicit and approximate low degree observations Multiply ordered nearest local descriptors Input MR images at a search limit Locates the location of MR images within the training dataset Then pseudo-CT patches are estimated through the nearest neighbor k. The proposed method for predicting pseudo-CT is quantified in a paired brain dataset analyzing MRI and CT images of 13 subjects. Our method produces pseudo-CT images with an average absolute error of 75/25 18 18/5 Hounsfield units, a peak signal-to-noise ratio of 1.15 dB 1.8 30 30.87, a relative average absolute error of 0.50 6.6 6.6 in PET attenuation correction, and a difference. The relative structure volume of the dose is 1098  0.055 in. At D98, compared to actual CT. The experimental results also show that our method exceeds four most advanced methods.

Keywords : ct prediction nonlinear descriptor low degree of approximation knn regression