CN106296592A - A kind of axially cone-beam scan resolves CT and rebuilds image DC shift correcting algorithm - Google Patents

Abstract

The present invention proposes a parsing CT being applicable to axial cone-beam scan and rebuilds the DC shift correcting algorithm of (CB-FBP/BPF) image.This method image area correction suppresses to resolve CT and rebuilds the DC shift of image, and the approximation feature that this drift is resolved CT reconstruction by axial cone-beam scan causes.By rectangular histogram cluster and spline interpolation, DC shift is effectively corrected.Result shows, the inventive method can efficiently reduce the DC shift of the parsing CT reconstruction image of axial cone-beam scan.

Description

A kind of axially cone-beam scan resolves CT and rebuilds image DC shift correcting algorithm

Technical field

The present invention relates to engineering in medicine technical field, be specifically related to a kind of axially cone-beam scan and resolve CT reconstruction image DC shift correcting algorithm.

Background technology

At present, spiral scan trajectory or circle add the parsing CT algorithm for reconstructing of extra scanning track have been occurred, and meets data sufficiency condition.But, cone-beam (CB) CT imaging is also frequently with a circular scan track (i.e. axially cone-beam scan), and circular scan track be unsatisfactory for data sufficiency condition.Therefore, resolving CT in axial cone-beam scan and rebuild in image, cone beam artefacts inevitably there will be, and this makes CT reconstructed image quality be deteriorated.

But on the other hand, convenient, the feature of stable operation that axial cone-beam scan has, this makes it in clinical imaging, still has the status do not replaced.Axially cone-beam scan resolves CT image reconstruction approximate data, is still largely used in X ray CT imaging.Above-mentioned situation is ordered about CT reconstruction field and is started to explore a kind of method, to reduce the cone beam artefacts in axial cone-beam scan parsing CT reconstruction image.Especially, when detector row increases, and cone angle becomes big.

In filtered back projection (FBP) types of image of spiral cone-beam tomography gained data for projection is rebuild, Katsevich algorithm is a kind of theoretic exact algorithm.Another kind of important algorithm, is so-called backprojection-filtration (BPF) algorithm；By it, can get the accurate reconstruction on π line segment.

Summary of the invention:

Not enough for prior art, the present invention proposes a kind of axially cone-beam scan and resolves CT reconstruction image DC shift correcting algorithm, including:

(1) by outside input image sequence；

(2) reference picture is selected by image sequence or other image sources；

(3) image sequence and reference picture are carried out region segmentation respectively；

(4) image sequence region segmentation and reference picture region segmentation are carried out parameter extraction respectively；

(5) parameter drift is estimated by image sequence region parameter with reference picture region parameter；

(6) parameter drift is carried out interpolation；

(7) according to drift interpolation, image sequence is carried out DC shift correction；

(8) image sequence output after correcting.

Described image sequence may be from variety classes image acquisition process, as long as the image reconstruction algorithm used is similar to；

Described reference picture can be selected from image sequence or other image sources；

Described image sequence region segmentation and reference picture Region Segmentation Algorithm can be with also can be different；

Described Region Segmentation Algorithm can be threshold method, region growing, region merging technique, rectangular histogram cluster or other algorithms；

Described region parameter can be average in region, interregional threshold value or other parameters；

Described parameter drift is estimated to use subtraction to may be used without additive method；

Described drift interpolation can use linearly, multinomial, batten or other interpolation algorithms.

The most parallel cone-beam imaging geometry

In natural cone-beam imaging geometry, including flat board cone-beam and two kinds of imaging geometries of arc cone-beam.In natural cone-beam imaging geometry, represent X source focus with S, P (x, y, z) be expressed as interior of articles a bit, then rise at S and will (x, y, ray z) will uniquely depend on visual angle η, segment angle γ and cone angle by P.

In natural cone-beam imaging geometry, circular trace is expressed as:

$(1-1)---\mathrm{ST}\left(\mathrm{\η}\right)=(R\sin \mathrm{\η},R\cos \mathrm{\η},0),\mathrm{\η}\⊆({\mathrm{\η}}_{\min },{\mathrm{\η}}_{\max }),$

Wherein η_minAnd η_maxInitial and the termination of expression source track.

It is believed that carry out in parallel cone-beam imaging geometry cone beam reconstruction can make in CT image produce noise evenly, therefore, inventive algorithm is applied to parallel cone-beam imaging geometry.Being reset by a kind of data for projection, raw cone beam imaging geometry can be changed into parallel cone-beam imaging geometry.Then, send radiation and by the ray of P point from focus S, uniquely depend on visual angle β, to distance t and the cone angle of rotary shaft.

Axial CB-FBP algorithm in the most parallel cone-beam imaging geometry

Definition central plane is the plane that in axial cone-beam scan, circular source track determines, re-defines some virtual planes in central plane above and below.If using constant Z₀Represent central plane or virtual plane (coordinate axes along the Z direction).Z₀=0 represents central plane, Z₀≠ 0 represents virtual plane.

With s (α, β, t；Z₀) represent Z₀The projection of planar imaging target.The axial CB-FBP algorithm of parallel cone-beam imaging geometry is expressed as:

$f(x,y,Z_0)\≈\frac{\mathrm{\π}}{{\mathrm{\β}}_{\max }-{\mathrm{\β}}_{\min }}\underset{{\mathrm{\β}}_{\min }}{\overset{{\mathrm{\β}}_{\max }}{\∫}}\cos \mathrm{\α}{s}_h(\mathrm{\α},\mathrm{\β},t;Z_0)d\mathrm{\β},$

$s_h(\mathrm{\α},\mathrm{\β},t;Z_0)=s(\mathrm{\α},\mathrm{\β},t;Z_0)\⊗h\left(t\right),$

H (t) is the filtering core function for parallel cone-beam imaging geometry.Object function f (x, y, Z₀) accurate reconstruction central plane obtain, and approximate reconstruction virtual plane obtain.

Axial CB-BPF algorithm in the most parallel cone-beam imaging geometry:

If making deeper definition:

$s_t(\mathrm{\α},\mathrm{\β},t;Z_0)=\frac{\∂}{\∂t}s(\mathrm{\α},\mathrm{\β},t;Z_0),$

So have

$-2\mathrm{\π}{H}_{\mathrm{\θ}}(x,y,Z_0)\≅b_{\mathrm{\θ}}(x,y,Z_0)=\underset{{\mathrm{\β}}_{\min }}{\overset{{\mathrm{\β}}_{\mathrm{mxa}}}{\∫}}\cos \mathrm{\αsgn}\left(\sin (\mathrm{\β}-\mathrm{\θ})\right)s_t(\mathrm{\α},\mathrm{\β},t;Z_0)\mathrm{d\β},$

H_θRepresent object function f (x, y, Z₀) along the Hilbert transform of direction (-sin θ, cos θ, 0).

In Performance Evaluation and proof of algorithm, coronalplane is as demonstration plane.Fig. 2,3 being the image rebuilding out with axial CB-FBP algorithm and corresponding transversal, these data for projection are obtained by full angle cone-beam scan.DC shift correction is accordingly for Fig. 2 a ' and a, and output image is b ' and b.It may be clearly seen that, axial CB-FBP algorithm present in the image rebuild, DC shift artifact can be eliminated by " axial CB-FBP corrects unified algorithm with DC shift " of the present invention.For prominent inventive algorithm at DC shift calibration result, the differential disply of Fig. 3 a and b is real and c ' in the difference scheming c, a and a ', all shows the correction to DC shift and really decreases cone beam artefacts.

Accompanying drawing explanation

Figure 1Axially cone-beam scan resolves CT and rebuilds image DC shift correcting algorithm flow processFigure

Figure 2Being rebuild image by axial CB-FBP under full angle cone-beam scan, (a) is the most calibrated with (a ')；B () is corrected with (b ')；C () and (c ') is difference between (a) and (b) and (a ') and (b ') respectively

Figure 3Image transversal (green line represents and do not corrects, and blue line represents and corrected), (a) noiseless is rebuild by axial CB-FBP under full angle cone-beam scan；B () has noise

Figure 4By axial CB-BPF and carry out Hilbert transform in the horizontal direction and the image rebuild under full angle cone-beam scan, (a) is the most calibrated with (a ')；B () is corrected with (b ')；C () and (c ') is the difference between (a) and (b) and (a ') and (b ') respectively

Figure 5By axial CB-BPF and carry out Hilbert transform in the horizontal direction and rebuild image transversal under full angle cone-beam scan, (a) noiseless；B () has noise

Detailed description of the invention

Below in conjunction with the accompanying drawings and a kind of embodiment of this method is described in detail:

Refer to shown in Fig. 1, a kind of axially cone-beam scan resolves CT and rebuilds image DC shift correcting algorithm, including: inputted by outside and do not correct image sequence 1, reference picture 3 is selected from image sequence 1 or other image sources 2 such as physics die body uncalibrated image, image sequence 1 and reference picture 3 are carried out region segmentation 4 respectively, 5, image sequence region segmentation 4 and reference picture region segmentation 5 are carried out parameter extraction 6 respectively, 7, parameter drift 8 is estimated by image sequence region parameter 6 and reference picture region parameter 7, parameter drift 8 is carried out interpolation 9, according to drift interpolation 9, image sequence 1 is carried out gray-scale shift correction 10, and image sequence output 11 after correcting.

Image sequence 1 is not corrected such as X-ray cone-beam CT reconstruction image human body vertical direction Slice Sequence described in this method.In the two ends section of deviation middle slice, gradation of image is susceptible to DC shift, mainly shows as soft tissue gray scale slightly biased low.

Reference picture 3 described in this method can never correct image sequence 1 or other image sources 2 are selected.As never corrected image sequence 1, then may select cone-beam CT reconstruction image middle slice, this section deteriorates to fan-beam algorithm for reconstructing due to cone beam reconstruction algorithm, rebuilds gradation of image and can be considered that numerical value is accurate；And if from other image sources 2, then optional carry out physics die body with fan-beam algorithm for reconstructing demarcating rebuilding, it is thus achieved that physics die body uncalibrated image.

Described in this method, image sequence 1 is carried out by region segmentation 4,5 respectively with reference picture 3.Region segmentation can be selected for based on histogrammic clustering algorithm, has same material and is in the pixel of different connected domain and is seen as belonging to the same area.

Described in this method, image sequence region segmentation 4 is carried out by parameter extraction 6,7 respectively with reference picture region segmentation 5.Average and interregional threshold value in parameters selection region.

Image sequence region parameter 6 can be subtracted each other with reference picture region parameter 7 and be obtained by parameter drift estimation 7 described in this method.

Interpolation 9 of drifting about described in this method obtains by parameter drift 8 is carried out spline interpolation.

Image sequence drift correction 10 described in this method is added by not correcting image sequence 1 obtains with drift interpolation 9.

The foregoing is only a kind of embodiment of this method, be not all of or unique embodiment, the conversion of any equivalence that this method technical scheme is taked by those of ordinary skill in the art by reading this method description, the claim being this method is contained.

Claims (7)

1. an axial cone-beam scan resolves CT and rebuilds image DC shift correcting algorithm, it is characterised in that: Inputted by outside and do not correct image sequence (1), select from image sequence (1) or other image sources (2) Reference picture (3), carries out region segmentation (4), (5) respectively to image sequence (1) and reference picture (3), Image sequence region segmentation (4) and reference picture region segmentation (5) are carried out parameter extraction (6), (7) respectively, Parameter drift (8) is estimated with reference picture region parameter (7), to ginseng by image sequence region parameter (6) Number drift (8) carries out interpolation (9), according to drift interpolation (9), image sequence (1) is carried out DC shift Correction (10), and by image sequence output (11) after correction.

Axial cone-beam scan the most according to claim 1 resolves CT and rebuilds the correction calculation of image DC shift Method, it is characterised in that: do not correct image sequence (1) and may be from various types of image acquisition process.

Axial cone-beam scan the most according to claim 1 resolves CT and rebuilds the correction calculation of image DC shift Method, it is characterised in that: described reference picture (3) can be selected from image sequence (1) or other image sources (2).

Axial cone-beam scan the most according to claim 1 resolves CT and rebuilds the correction calculation of image DC shift Method, it is characterised in that: described region segmentation (4), (5) algorithm may be the same or different.

Axial cone-beam scan the most according to claim 1 resolves CT and rebuilds the correction calculation of image DC shift Method, it is characterised in that: described region parameter (6), (7) can be average in region, interregional threshold value or its His parameter.

Axial cone-beam scan the most according to claim 1 resolves CT and rebuilds the correction calculation of image DC shift Method, it is characterised in that: described parameter drift estimates that (8) can use subtraction to may be used without additive method.

Axial cone-beam scan the most according to claim 1 resolves CT and rebuilds the correction calculation of image DC shift Method, it is characterised in that: described drift interpolation (9) can use linearly, multinomial, batten or other interpolation are calculated Method.