ESTIMATING PROJECTION NOISE FOR COMPUTED TOMOGRAPHY IMAGE RECONSTRUCTION
Publication Date: 2015-Jan-19
The IP.com Prior Art Database
The invention proposes a technique for parameterized modeling of projection noise for computed tomography. The technique includes an experimental scheme to compute coefficients by using polynomial fitting techniques. The proposed technique to approximate prep noise is efficient and gives an error rate between exact and approximation within 5%.
The present invention relates generally to computed tomography and more particularly to estimation of projection noise in computed tomography image reconstruction.
Fast variance prediction for reconstructed CT images is useful for analysis of resulting images and potentially for dynamic dose adjustment during a scan. However, though conventional techniques are able to approximate variance quickly, most of these techniques are computationally expensive and are applicable only for specific CT geometries.
According to a conventional technique, as shown in below derivation, if higher terms are ignored, the projection variance can be linearized by a factor Ec. However, a real time estimation of projection noise remains a challenge.
It would be desirable to have a technique for real time estimation of projection noise that includes effect of polychromaticity in computed tomography image reconstruction.
BRIEF DESCRIPTION OF THE FIGURES
Figure 1 depicts preparation of a water phantom with incremental thickness.
Figure 2 is a graphical representation for experiment 1 and experiment 2 of computed and fitted plots for computing CiS`.
Figure 3 depicts a validation of the technique described herein, with a good match in magnitude and shape between exact and the approximate modeling method.
Figure 4 depicts a good match in variable kVp cases.
Figure 5 depicts the error rate in perfusion computed tomography for sinusoidal kVp modulation.
Figure 6 depicts a flow diagram for the experimental scheme to calculate projection variance at 1st order approximation.
The invention proposes a technique for parameterized modeling of projection noise for computed tomography image reconstruction. The technique includes a corresponding experimental scheme to compute coefficients by using polynomial fitting techniques. Similar to the beam hardening calibration process, these coefficients are energy and channel dependent. The error between results obtained using the proposed technique and ground truth is small.
According to an embodiment of the proposed technique, expression of projection variance is divided into two parts as expressed in equations below. A first part is a ratio of exponential of prep to air scan, which scales with inverse of milliampere seconds (mAs). A second part is expressed as a 3rd polynomial of prep, which is invariant to mAs and is a surrogate of effective mean energy of attenuated spectrum.
Assuming data acquisition system (DAS) gain is constant,
Following simulation scheme is used to compute Cis`. As shown in Figure 1, a water phantom with incremental thickne...