CN105741241A - Tumor region image enhancement method and system based on enhanced composite image - Google Patents

Abstract

The invention discloses a tumor region image enhancement method and system based on an enhanced composite image, and relates to the field of medical image processing. The invention puts forwards a concept "enhanced composite image" for the first time, a ROI (Region Of Interest) which covers all tumor regions is independently subjected to noise reduction processing and enhancement processing in an original CT (Computed Tomography) or MR (Magnetic Resonance) image to obtain a noise reduction image and an enhancement image of the RIO, and the noise reduction image and the enhancement image are subjected to weight fusion to obtain the enhanced composite image. The RIO in the original CT or MR image can be enhanced, and different weighting factors in the enhanced composite image are selected to cause the surface and the fuzzy boundary of the tumor in the CT or MR image to become clear so as to bring convenience for doctors to observe the surface and the boundary of the tumor. A threshold value segmentation method is adopted on the enhanced composite image, and therefore, the tumor region can be precisely segmented.

Description

Tumor region image enchancing method and the system of image is strengthened based on synthesis

Technical field

The present invention relates to field of medical image processing, be specifically related to a kind of tumor region image enchancing method and system strengthening image based on synthesis.

Background technology

Doctor is in the process of actual diagnosing tumour relevant disease, by to CT (ComputedTomography, CT scan) or MR (MagneticResonance, magnetic resonance) etc. the border of tumor, area of section and volume measurement and analyze in medical image, the clinical definite state of an illness can be helped, at this moment need the tumor boundaries in the medical images such as CT or MR is split.The Accurate Segmentation of tumor boundaries is extremely important to treatment plan, and this part work at present depends on hand drawing, and degree of accuracy is not high.Due to the noise in the medical images such as CT or MR and ambiguity, cause that the border of many tumors is fuzzyyer, take traditional image partition method, for instance on the medical images such as original CT or MR, adopt thresholding method, it is difficult to the border of tumor is carried out Accurate Segmentation.

Summary of the invention

The invention aims to overcome the deficiency of above-mentioned background technology, a kind of tumor region image enchancing method and system strengthening image based on synthesis is provided, can strengthen original CT or MR image cover whole tumor region ROI region, by selecting synthesis to strengthen weighter factors different in image, the surface and the smeared out boundary that make tumor in original CT or MR image become apparent from, and facilitate doctor to observe surface and the border of tumor；Strengthening in synthesis and adopt threshold segmentation method on image, it is possible to Accurate Segmentation goes out tumor region, degree of accuracy is apparently higher than the degree of accuracy being made directly Threshold segmentation in original CT or MR image.

The present invention provides a kind of tumor region image enchancing method strengthening image based on synthesis, comprises the following steps:

A, in the frame original CT image comprising tumor or MR image, choose one cover whole tumor regions oval region of interest ROI；Pixel beyond ROI region is all set to zero, the pixel value within ROI region is carried out the adjustment of window width and window level, so as to the requirement that the compound mankind observe；

B, employing anisotropy parameter method, carry out noise reduction process to the ROI region in the image after regulating through window width and window level, obtain covering the noise-reduced image of the ROI region of whole tumor region；

C, employing multi-scale enhancement method, carry out enhancement process to the ROI region in the image after regulating through window width and window level, obtain covering the enhancing image of the ROI region of whole tumor region；

The enhancing image of the ROI region covering whole tumor regions that D, the noise-reduced image of ROI region covering whole tumor regions that step B is obtained and step C obtain is weighted merging, and the synthesis obtaining covering the ROI region of whole tumor region strengthens image.

On the basis of technique scheme, the process of the adjustment that the pixel value within ROI region carries out in step A window width and window level is: centered by oval barycenter, selects a rectangular area, the length r of this rectangle_lengthLess than oval major axis, the width r of this rectangle_widthLess than oval short axle；Choose the maximum C that the maximum gradation value in this rectangle is window width in window technique_max, choosing the minimum gradation value in this rectangle is the minima C of window width in window technique_min；Then arbitrarily select a formula in following three formula, the pixel value in ellipse mapped:

$I=\mathrm{\α}\×\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}---\left(1\right)$

$I=\mathrm{\α}\×log(1+\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }})---\left(2\right)$

$I=\mathrm{\α}\×{\left(\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}\right)}^{{\mathrm{\γ}}_1}---\left(3\right)$

Wherein, I_oriFor original CT or MR image, α is the coefficient controlling image overall brightness, γ₁For adjustable coefficient, by changing γ₁Value obtain the mapping curve of different modes, I is the image after window width and window level regulates.

On the basis of technique scheme, in step A, γ₁Value less than 1 time, input value narrower for scope is transformed into the output valve an of wider range by formula (3).

On the basis of technique scheme, the anisotropy parameter method in step B is: using Anisotropic Diffusion Model to remove noise, the formula of Anisotropic Diffusion Model is:

$\frac{\∂I}{\∂t}=div\left(c\right(x,y,t)\▿I)=\▿c\·\▿I+c(x,y,t)\mathrm{\Δ}I---\left(4\right)$

Wherein, div () is divergence operator, c (x, y, t) be diffusion coefficient, and diffusion coefficient is selected as the function of image gradient, and t is the time,Represent gradient operator,It is then the gradient of diffusion coefficient c,Being the gradient of image I, Δ represents Laplace operator, and Δ I is the Laplace operator of image I；

The Solving Partial Differential Equations that formula (4) is represented, the solution of the equation is expressed as I_rn, I_rnNoise-reduced image for the step B ROI region covering whole tumor regions obtained.

On the basis of technique scheme, in step B, diffusion coefficient c in the formula of described Anisotropic Diffusion Model (x, y, function t) has two kinds of expression waies:

$c\left(\right||\▿I|\left|\right)=e^{-{\left(\right||\▿I||/K)}^2}---\left(5\right)$

$c\left(\right||\▿I|\left|\right)=\frac{1}{1+{\left(\frac{\left|\right|\▿I\left|\right|}{K}\right)}^2}---\left(6\right)$

Wherein,Representing the mould of gradient of image I, K is constant, and the two diffusion coefficient function is using the mould of the gradient of image I as the foundation of diffusion speed, and the position diffusion coefficient big in gradient is little, thus playing the purpose of Protect edge information.

On the basis of technique scheme, the multi-scale enhancement method in step C is:

The ROI region covering whole tumor regions is carried out the enhancement process under different scale:

First, by formula (7), the ROI region covering whole tumor regions is carried out down-sampling, obtains the image I after down-sampling_re；

I_re=REDUCE (I) (7)

Wherein, REDUCE () represents down-sampling operation；

Then, by formula (8), to the image I after down-sampling_reZoom in and out enhancement process；

I_u=(I_re-I_re*G)^γ2(8)

Wherein, I_uIt is the down-sampling enhancing image of size reduction, I_reBeing the image after down-sampling, G is gaussian kernel, I_re-I_re* the radio-frequency component of G representative image, γ₂For controlling the coefficient of tumor region and the difference of non-tumor region, I_uSize and down-sampling after image I_reEquivalently-sized, be respectively less than through window width and window level regulate after image I；

Then, by formula (9), the down-sampling of the size reduction that formula (8) obtains is strengthened image I_uCarry out up-sampling, obtain I_ex: I_ex=EXPAND (I_u)(9)

Wherein, I_exFor the enhancing image of the step C ROI region covering whole tumor regions obtained, I_exSize with regulate through window width and window level after image I equivalently-sized；EXPAND () represents up-sampling operation.

On the basis of technique scheme, EXPAND () in step C is obtained by extension interpolation operator, if extending the size of a times, then strengthen the pixel of a times in both the horizontal and vertical directions, namely needing between each row any two pixel to insert a value, every two in the ranks need to insert a line；Interpolation operator adopts bi-cubic interpolation method.

On the basis of technique scheme, adopting the weighted average fusion method of Pixel-level to be weighted merging in step D, formula is as follows:

I_en=w I_rn+(1-w)·I_ex(10)

Wherein, I_rnFor the noise-reduced image of the step B ROI region covering whole tumor regions obtained, I_exFor the enhancing image of the step C ROI region covering whole tumor regions obtained, w is weighter factor, I_enSynthesis for covering the ROI region of whole tumor region strengthens image；By selecting weighter factor w different in formula (10), doctor is facilitated to observe surface and the border of tumor.

The present invention also provides for a kind of tumor region Image Intensified System strengthening image based on synthesis, and this system includes ROI region and chooses unit, window width and window level regulon, noise reduction unit, enhancement unit, Weighted Fusion unit, wherein:

Described ROI region choose unit for: in the frame original CT image comprising tumor or MR image, choose one cover whole tumor regions oval region of interest ROI；

Described window width and window level regulon is used for: the pixel beyond ROI region is all set to zero, and the pixel value within ROI region carries out the adjustment of window width and window level, so as to the requirement that the compound mankind observe；

Described noise reduction unit is used for: adopt anisotropy parameter method, the ROI region in the image after regulating through window width and window level is carried out noise reduction process, obtains covering the noise-reduced image of the ROI region of whole tumor region；

Described enhancement unit is used for: adopt multi-scale enhancement method, the ROI region in the image after regulating through window width and window level is carried out enhancement process, obtains covering the enhancing image of the ROI region of whole tumor region；

Described Weighted Fusion unit is used for: being weighted merging with the enhancing image of the ROI region covering whole tumor regions to the noise-reduced image of ROI region covering whole tumor regions, the synthesis obtaining covering the ROI region of whole tumor region strengthens image.

On the basis of technique scheme, the process of the adjustment that the pixel value within ROI region is carried out window width and window level by described window width and window level regulon is: centered by oval barycenter, selects a rectangular area, the length r of this rectangle_lengthLess than oval major axis, the width r of this rectangle_widthLess than oval short axle；Choose the maximum C that the maximum gradation value in this rectangle is window width in window technique_max, choosing the minimum gradation value in this rectangle is the minima C of window width in window technique_min；Arbitrarily select a formula in following three formula, the pixel value in ellipse mapped:

$I=\mathrm{\α}\×\frac{I_{ori}-C_{min}}{C_{max}-C_{min}}---\left(1\right)$

$I=\mathrm{\α}\×log(1+\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }})---\left(2\right)$

$I=\mathrm{\α}\×{\left(\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}\right)}^{{\mathrm{\γ}}_1}---\left(3\right)$

Wherein, I_oriFor original CT or MR image, α is the coefficient controlling image overall brightness, γ₁For adjustable coefficient, by changing γ₁Value obtain the mapping curve of different modes, I is the image after window width and window level regulates.

On the basis of technique scheme, γ₁Value less than 1 time, input value narrower for scope is transformed into the output valve an of wider range by formula (3).

On the basis of technique scheme, the anisotropy parameter method that described noise reduction unit adopts is: using Anisotropic Diffusion Model to remove noise, the formula of Anisotropic Diffusion Model is:

$\frac{\∂I}{\∂t}=div\left(c\right(x,y,t)\▿I)=\▿c\·\▿I+c(x,y,t)\mathrm{\Δ}I---\left(4\right)$

Wherein, div () is divergence operator, c (x, y, t) be diffusion coefficient, and diffusion coefficient is selected as the function of image gradient, and t is the time,Represent gradient operator,It is then the gradient of diffusion coefficient c,Being the gradient of image I, Δ represents Laplace operator, and Δ I is the Laplace operator of image I；

The Solving Partial Differential Equations that formula (4) is represented, the solution of the equation is expressed as I_rn, I_rnFor covering the noise-reduced image of the ROI region of whole tumor region.

On the basis of technique scheme, diffusion coefficient c in the formula of described Anisotropic Diffusion Model (x, y, function t) has two kinds of expression waies:

$c\left(\right||\▿I|\left|\right)=e^{-{\left(\right||\▿I||/K)}^2}---\left(5\right)$

$c\left(\right||\▿I|\left|\right)=\frac{1}{1+{\left(\frac{\left|\right|\▿I\left|\right|}{K}\right)}^2}---\left(6\right)$

Wherein,Representing the mould of gradient of image I, K is constant, and the two diffusion coefficient function is using the mould of the gradient of image I as the foundation of diffusion speed, and the position diffusion coefficient big in gradient is little, thus playing the purpose of Protect edge information.

On the basis of technique scheme, the multi-scale enhancement method that described enhancement unit adopts is:

The ROI region covering whole tumor regions is carried out the enhancement process under different scale:

First, by formula (7), the ROI region covering whole tumor regions is carried out down-sampling, obtains the image I after down-sampling_re；

I_re=REDUCE (I) (7)

Wherein, REDUCE () represents down-sampling operation；

Then, by formula (8), to the image I after down-sampling_reZoom in and out enhancement process；

I_u=(I_re-I_re*G)^γ2(8)

Wherein, I_uIt is the down-sampling enhancing image of size reduction, I_reBeing the image after down-sampling, G is gaussian kernel, I_re-I_re* the radio-frequency component of G representative image, γ₂For controlling the coefficient of tumor region and the difference of non-tumor region, I_uSize and down-sampling after image I_reEquivalently-sized, be respectively less than through window width and window level regulate after image I；

Then, by formula (9), the down-sampling of the size reduction that formula (8) obtains is strengthened image I_uCarry out up-sampling, obtain I_ex: I_ex=EXPAND (I_u)(9)

Wherein, I_exFor the enhancing image of the step C ROI region covering whole tumor regions obtained, I_exSize with regulate through window width and window level after image I equivalently-sized；EXPAND () represents up-sampling operation.

On the basis of technique scheme, described EXPAND () is obtained by extension interpolation operator, if to extend the size of a times, then strengthens the pixel of a times in both the horizontal and vertical directions, namely needing between each row any two pixel to insert a value, every two in the ranks need to insert a line；Interpolation operator adopts bi-cubic interpolation method.

On the basis of technique scheme, described Weighted Fusion unit adopts the weighted average fusion method of Pixel-level to be weighted merging, and formula is as follows:

I_en=w I_rn+(1-w)·I_ex(10)

Wherein, I_rnFor covering the noise-reduced image of the ROI region of whole tumor region, I_exFor covering the enhancing image of the ROI region of whole tumor region, w is weighter factor, I_enSynthesis for covering the ROI region of whole tumor region strengthens image；By selecting weighter factor w different in formula (10), doctor is facilitated to observe surface and the border of tumor.

Compared with prior art, advantages of the present invention is as follows:

Present invention firstly provides " synthesis strengthens image " this concept, in original CT or MR image, the ROI region covering whole tumor regions is carried out noise reduction process and enhancement process respectively, obtain the noise-reduced image of ROI region and strengthen image, it is weighted merging to noise-reduced image and enhancing image, obtains synthesis and strengthen image.Compared with original CT or MR image, it is apparent that synthesis strengthens the border of tumor in image, and the information loss of tumor surface is also few simultaneously.The present invention can strengthen original CT or MR image cover whole tumor region ROI region, by selecting synthesis to strengthen weighter factors different in image, the surface and the smeared out boundary that make tumor in original CT or MR image become apparent from, and facilitate doctor to observe surface and the border of tumor；Strengthening in synthesis and adopt threshold segmentation method on image, it is possible to Accurate Segmentation goes out tumor region, degree of accuracy is apparently higher than the degree of accuracy being made directly Threshold segmentation in original CT or MR image.

Accompanying drawing explanation

Fig. 1 is the flow chart of the tumor region image enchancing method strengthening image in the embodiment of the present invention based on synthesis.

Fig. 2 is the image in the embodiment of the present invention after window width and window level adjustment.

Fig. 3 be in the embodiment of the present invention window width and window level regulate after image in ROI region.

Fig. 4 is the noise-reduced image of ROI region in the embodiment of the present invention.

Fig. 5 is the enhancing image of ROI region in the embodiment of the present invention.

The synthesis that Fig. 6 is ROI region in the embodiment of the present invention strengthens image.

Detailed description of the invention

Below in conjunction with drawings and the specific embodiments, the present invention is described in further detail.

Being difficult to, for the obscurity boundary of tumor in original CT or MR image, the problem split, the embodiment of the present invention provides a kind of tumor region image enchancing method strengthening image based on synthesis, and shown in Figure 1, the method comprises the following steps:

A, in the frame original CT image comprising tumor or MR image, choose one cover whole tumor regions oval ROI (RegionOfInterest, area-of-interest)；Pixel beyond ROI region is all set to zero, the pixel value within ROI region is carried out the adjustment of window width and window level, so as to the requirement that the compound mankind observe.

In actual applications, first select ROI region, regulate window width and window level again, but, owing to the pixel value of original medical image is at-1000 to+1000HU (HounsfieldUnit, Korea Spro's Sen Feierde unit) between, it is impossible to show in regular display, so image after a vision-control window width and window level in the embodiment of the present invention.Image after window width and window level adjustment is shown in Figure 2；ROI region in the image that window width and window level regulates is shown in Figure 3.

B, employing anisotropy parameter method (Anisotropicdiffusion), carry out noise reduction process to the ROI region in the image after regulating through window width and window level, obtain covering the noise-reduced image of the ROI region of whole tumor region, shown in Figure 4；

C, employing multi-scale enhancement method, carry out enhancement process to the ROI region in the image after regulating through window width and window level, obtain covering the enhancing image of the ROI region of whole tumor region, shown in Figure 5；

The enhancing image of the ROI region covering whole tumor regions that D, the noise-reduced image of ROI region covering whole tumor regions that step B is obtained and step C obtain is weighted merging, the synthesis obtaining covering the ROI region of whole tumor region strengthens image, shown in Figure 6.

Compared with original CT or MR image, it is apparent that synthesis strengthens the border of tumor in image, and the information loss of tumor surface is also few simultaneously.This synthesis strengthens image and as the input of threshold segmentation method, can strengthen in this synthesis and adopt threshold segmentation method on image, it is possible to Accurate Segmentation goes out tumor region, and degree of accuracy is apparently higher than the degree of accuracy being made directly Threshold segmentation in original CT or MR image.Simultaneously, it is possible to by selecting synthesis to strengthen weighter factors different in image, make the surface of tumor in original CT or MR image and smeared out boundary become apparent from, facilitate doctor to observe surface and the border of tumor.

The process of the adjustment that the pixel value within ROI region carries out in step A window width and window level is: centered by oval barycenter, selects a rectangular area, the length r of this rectangle_lengthLess than oval major axis, the width r of this rectangle_widthLess than oval short axle；Choose the maximum C that the maximum gradation value in this rectangle is window width in window technique_max, choosing the minimum gradation value in this rectangle is the minima C of window width in window technique_min；Then arbitrarily select a formula in following three formula, the pixel value in ellipse mapped:

$I=\mathrm{\α}\×\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}---\left(1\right)$

$I=\mathrm{\α}\×log(1+\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }})---\left(2\right)$

$I=\mathrm{\α}\×{\left(\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}\right)}^{{\mathrm{\γ}}_1}---\left(3\right)$

Wherein, I_oriFor original CT or MR image, α is the coefficient controlling image overall brightness, γ₁For adjustable coefficient, by changing γ₁Value can obtain the mapping curve of different modes, γ₁Value less than 1 time, input value narrower for scope can be transformed into the output valve an of wider range by formula (3), and I is the image after window width and window level regulates.

Anisotropy parameter method in step B is: using Anisotropic Diffusion Model to remove noise, the formula of Anisotropic Diffusion Model is:

$\frac{\∂I}{\∂t}=div\left(c\right(x,y,t)\▿I)=\▿c\·\▿I+c(x,y,t)\mathrm{\Δ}I---\left(4\right)$

Wherein, div () is divergence operator, c (x, y, t) be diffusion coefficient, and diffusion coefficient is typically selected to be the function of image gradient, and t is the time,Represent gradient operator,It is then the gradient of diffusion coefficient c,Being the gradient of image I, Δ represents Laplace operator, and Δ I is the Laplace operator of image I.

The formula of above-mentioned Anisotropic Diffusion Model is proposed by two scholars of Perona and Malik (two names), and Perona and Malik also proposes: diffusion coefficient c (x, y, function t) has two kinds of expression waies:

$c\left(\right||\▿I|\left|\right)=e^{-{\left(\right||\▿I||/K)}^2}---\left(5\right)$

$c\left(\right||\▿I|\left|\right)=\frac{1}{1+{\left(\frac{\left|\right|\▿I\left|\right|}{K}\right)}^2}---\left(6\right)$

Representing the mould of the gradient of image I, K is constant.The two diffusion coefficient function foundation using the mould of the gradient of image I as diffusion speed, the position diffusion coefficient big in gradient is little, thus playing the purpose of Protect edge information.

The Solving Partial Differential Equations that formula (4) is represented, the solution of the equation is expressed as I_rn, I_rnNoise-reduced image for the step B ROI region covering whole tumor regions obtained.

Multi-scale enhancement method in step C is:

In order to obtain tumor boundaries clearly, the ROI region covering whole tumor regions is carried out the enhancement process under different scale:

First, by formula (7), the ROI region covering whole tumor regions is carried out down-sampling, obtains the image I after down-sampling_re；

I_re=REDUCE (I) (7)

Wherein, REDUCE () represents down-sampling operation；

Then, by formula (8), to the image I after down-sampling_reZoom in and out enhancement process；

I_u=(I_re-I_re*G)^γ2(8)

Wherein, I_uIt is the down-sampling enhancing image of size reduction, I_reBeing the image after down-sampling, G is gaussian kernel, and the line number of gaussian kernel and columns can regulate, and are typically between 20 to 90.

I_re-I_re* the radio-frequency component of G representative image, tumor boundaries therein and unchanged clearly, in order to expand the difference of tumor region and non-tumor region further, the γ of this radio-frequency component to be calculated₂Power, γ₂For controlling the coefficient of tumor region and the difference of non-tumor region, due to I_reBeing the image after down-sampling, it is smaller in size than the image I after window width and window level regulates, and through formula (8), the down-sampling obtaining size reduction strengthens image I_u, I_uSize and down-sampling after image I_reEquivalently-sized, be respectively less than through window width and window level regulate after image I.

Then, by formula (9), the down-sampling of the size reduction that formula (8) obtains is strengthened image I_uCarry out up-sampling, obtain I_ex: I_ex=EXPAND (I_u)(9)

Wherein, I_exFor the enhancing image of the step C ROI region covering whole tumor regions obtained, I_exSize with regulate through window width and window level after image I equivalently-sized；EXPAND () represents up-sampling operation.

EXPAND () is obtained by extension interpolation operator, if to extend the size of a times, then strengthens the pixel of a times in both the horizontal and vertical directions, namely needs between each row any two pixel to insert a value, and every two in the ranks need to insert a line；Interpolation operator adopts bi-cubic interpolation method (Bicubicinterpolation).

The weighted average fusion method adopting Pixel-level in step D is weighted merging, and formula is as follows:

I_en=w I_rn+(1-w)·I_ex(10)

Wherein, I_rnFor the noise-reduced image of the step B ROI region covering whole tumor regions obtained, I_exFor the enhancing image of the step C ROI region covering whole tumor regions obtained, w is weighter factor, I_enSynthesis for covering the ROI region of whole tumor region strengthens image.Can pass through to select weighter factor w different in formula (10), facilitate doctor to observe surface and the border of tumor.

The reason setting up synthesis enhancing image in step D is: the noise in the noise-reduced image of step B has certain reduction relative to artwork, but the smeared out boundary of its tumor is not apparent from, and is unfavorable for segmentation.Although and the enhancing image border in step C is apparent from, but due to change of scale, the information loss of tumor surface is more, is unfavorable for doctor's inspections and examinations.In order to take into account the inspections and examinations of tumor image segmentation and doctor simultaneously, the enhancing image of the ROI region covering whole tumor regions that the noise-reduced image of the ROI region covering whole tumor regions that step B is obtained and step C obtain is weighted merging, the synthesis obtaining covering the ROI region of whole tumor region strengthens image, synthesis strengthens the tumor boundaries of image and is apparent from, and tumor surface information loss is few.

Next for CT image, demonstrate and process, by the method, the result obtained.

Owing to the span of the pixel value in original CT image is between-1000 and+1000HU, the effect in regular display is bad.Therefore, the image after window width regulates outside window is provided only below.

The image after window width and window level adjustment in step A is shown in Figure 2, and the ROI region in the image that window width and window level regulates is shown in Figure 3, it can be seen that the wherein obscurity boundary of tumor.

The noise-reduced image of the ROI region in step B is shown in Figure 4, compared with Fig. 3, and the noise decrease to some degree in Fig. 4, but tumor boundaries is not obviously improved.

The enhancing image of the ROI region in step C is shown in Figure 5, and compared with Fig. 3, in Fig. 5, the border of tumor is apparent from, but the information loss of tumor surface is more.

It is shown in Figure 6 that the synthesis of the ROI region in step D strengthens image, and compared with Fig. 4 and Fig. 5, Fig. 6 is a kind of half-way house of Fig. 4 and Fig. 5, and tumor boundaries therein is apparent from, and tumor surface information loss is few.

The embodiment of the present invention also provides for a kind of tumor region Image Intensified System strengthening image based on synthesis, and this system includes ROI region and chooses unit, window width and window level regulon, noise reduction unit, enhancement unit, Weighted Fusion unit, wherein:

ROI region choose unit for: in the frame original CT image comprising tumor or MR image, choose one cover whole tumor regions oval ROI (RegionOfInterest, area-of-interest)；

Window width and window level regulon is used for: the pixel beyond ROI region is all set to zero, and the pixel value within ROI region carries out the adjustment of window width and window level, so as to the requirement that the compound mankind observe.

In actual applications, first select ROI region, then regulate window width and window level, but, owing to the pixel value of original medical image is between-1000 to+1000HU, it is impossible to show in regular display, so image after a vision-control window width and window level in the embodiment of the present invention.Image after window width and window level adjustment is shown in Figure 2；ROI region in the image that window width and window level regulates is shown in Figure 3.

Noise reduction unit is used for: adopt anisotropy parameter method (Anisotropicdiffusion), ROI region in image after regulating through window width and window level is carried out noise reduction process, obtain covering the noise-reduced image of the ROI region of whole tumor region, shown in Figure 4；

Enhancement unit is used for: adopt multi-scale enhancement method, the ROI region in the image after regulating through window width and window level is carried out enhancement process, obtains covering the enhancing image of the ROI region of whole tumor region, shown in Figure 5；

Weighted Fusion unit is used for: be weighted merging with the enhancing image of the ROI region covering whole tumor regions to the noise-reduced image of ROI region covering whole tumor regions, the synthesis obtaining covering the ROI region of whole tumor region strengthens image, shown in Figure 6.

The process of the adjustment that the pixel value within ROI region is carried out window width and window level by window width and window level regulon is: centered by oval barycenter, selects a rectangular area, the length r of this rectangle_lengthLess than oval major axis, the width r of this rectangle_widthLess than oval short axle；Choose the maximum C that the maximum gradation value in this rectangle is window width in window technique_max, choosing the minimum gradation value in this rectangle is the minima C of window width in window technique_min；Then arbitrarily select a formula in following three formula, the pixel value in ellipse mapped:

$I=\mathrm{\α}\×\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}---\left(1\right)$

$I=\mathrm{\α}\×log(1+\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }})---\left(2\right)$

$I=\mathrm{\α}\×{\left(\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}\right)}^{{\mathrm{\γ}}_1}---\left(3\right)$

Wherein, I_oriFor original CT or MR image, α is the coefficient controlling image overall brightness, γ₁For adjustable coefficient, by changing γ₁Value can obtain the mapping curve of different modes, γ₁Value less than 1 time, input value narrower for scope can be transformed into the output valve an of wider range by formula (3), and I is the image after window width and window level regulates.

The anisotropy parameter method that noise reduction unit adopts is: using Anisotropic Diffusion Model to remove noise, the formula of Anisotropic Diffusion Model is:

$\frac{\∂I}{\∂t}=div\left(c\right(x,y,t)\▿I)=\▿c\·\▿I+c(x,y,t)\mathrm{\Δ}I---\left(4\right)$

Wherein, div () is divergence operator, c (x, y, t) be diffusion coefficient, and diffusion coefficient is typically selected to be the function of image gradient, and t is the time,Represent gradient operator,It is then the gradient of diffusion coefficient c,Being the gradient of image I, Δ represents Laplace operator, and Δ I is the Laplace operator of image I.

The formula of above-mentioned Anisotropic Diffusion Model is proposed by two scholars of Perona and Malik (two names), and Perona and Malik also proposes: diffusion coefficient c (x, y, function t) has two kinds of expression waies:

$c\left(\right||\▿I|\left|\right)=e^{-{\left(\right||\▿I||/K)}^2}---\left(5\right)$

$c\left(\right||\▿I|\left|\right)=\frac{1}{1+{\left(\frac{\left|\right|\▿I\left|\right|}{K}\right)}^2}---\left(6\right)$

Wherein,Representing the mould of the gradient of image I, K is constant.The two diffusion coefficient function foundation using the mould of the gradient of image I as diffusion speed, the position diffusion coefficient big in gradient is little, thus playing the purpose of Protect edge information.

The Solving Partial Differential Equations that formula (4) is represented, the solution of the equation is expressed as I_rn, I_rnFor covering the noise-reduced image of the ROI region of whole tumor region.

The multi-scale enhancement method that enhancement unit adopts is:

In order to obtain tumor boundaries clearly, the ROI region covering whole tumor regions is carried out the enhancement process under different scale:

First, by formula (7), the ROI region covering whole tumor regions is carried out down-sampling, obtains the image I after down-sampling_re；

I_re=REDUCE (I) (7)

Wherein, REDUCE () represents down-sampling operation；

Then, by formula (8), to the image I after down-sampling_reZoom in and out enhancement process；

I_u=(I_re-I_re*G)^γ2(8)

Wherein, I_uIt is the down-sampling enhancing image of size reduction, I_reBeing the image after down-sampling, G is gaussian kernel, and the line number of gaussian kernel and columns can regulate, and are typically between 20 to 90.

I_re-I_re* the radio-frequency component of G representative image, tumor boundaries therein and unchanged clearly, in order to expand the difference of tumor region and non-tumor region further, the γ of this radio-frequency component to be calculated₂Power, γ₂For controlling the coefficient of tumor region and the difference of non-tumor region, due to I_reBeing the image after down-sampling, it is smaller in size than the image I after window width and window level regulates, and through formula (8), the down-sampling obtaining size reduction strengthens image I_u, I_uSize and down-sampling after image I_reEquivalently-sized, be respectively less than through window width and window level regulate after image I.

Then, by formula (9), the down-sampling of the size reduction that formula (8) obtains is strengthened image I_uCarry out up-sampling, obtain I_ex: I_ex=EXPAND (I_u)(9)

Wherein, I_exFor covering the enhancing image of the ROI region of whole tumor region, I_exSize with regulate through window width and window level after image I equivalently-sized；EXPAND () represents up-sampling operation.

EXPAND () is obtained by extension interpolation operator, if to extend the size of a times, then strengthens the pixel of a times in both the horizontal and vertical directions, namely needs between each row any two pixel to insert a value, and every two in the ranks need to insert a line；Interpolation operator adopts bi-cubic interpolation method (Bicubicinterpolation).

Weighted Fusion unit adopts the weighted average fusion method of Pixel-level to be weighted merging, and formula is as follows:

I_en=w I_rn+(1-w)·I_ex(10)

Wherein, I_rnFor covering the noise-reduced image of the ROI region of whole tumor region, I_exFor covering the enhancing image of the ROI region of whole tumor region, w is weighter factor, I_enSynthesis for covering the ROI region of whole tumor region strengthens image.Can pass through to select weighter factor w different in formula (10), facilitate doctor to observe surface and the border of tumor.

From the foregoing, it will be observed that the tumor region image enchancing method and the system that strengthen image based on synthesis that the present invention proposes, it is possible to it is used for making the surface of tumor in original CT or MR image and smeared out boundary become apparent from, facilitates doctor to observe surface and the border of tumor.Strengthen in synthesis and image adopts threshold segmentation method, tumor region can be gone out by Accurate Segmentation.

The embodiment of the present invention can be carried out various modifications and variations by those skilled in the art, if these amendments and modification are within the scope of the claims in the present invention and equivalent technologies thereof, then these amendments and modification are also within protection scope of the present invention.

The prior art that the content not being described in detail in description is known to the skilled person.

Claims (16)

1. the tumor region image enchancing method strengthening image based on synthesis, it is characterised in that comprise the following steps:

A, in the frame original CT image comprising tumor or MR image, choose one cover whole tumor regions oval region of interest ROI；Pixel beyond ROI region is all set to zero, the pixel value within ROI region is carried out the adjustment of window width and window level, so as to the requirement that the compound mankind observe；

B, employing anisotropy parameter method, carry out noise reduction process to the ROI region in the image after regulating through window width and window level, obtain covering the noise-reduced image of the ROI region of whole tumor region；

C, employing multi-scale enhancement method, carry out enhancement process to the ROI region in the image after regulating through window width and window level, obtain covering the enhancing image of the ROI region of whole tumor region；

The enhancing image of the ROI region covering whole tumor regions that D, the noise-reduced image of ROI region covering whole tumor regions that step B is obtained and step C obtain is weighted merging, and the synthesis obtaining covering the ROI region of whole tumor region strengthens image.

2. the tumor region image enchancing method strengthening image based on synthesis as claimed in claim 1, it is characterized in that: the process of the adjustment that the pixel value within ROI region carries out in step A window width and window level is: centered by oval barycenter, select a rectangular area, the length r of this rectangle_lengthLess than oval major axis, the width r of this rectangle_widthLess than oval short axle；Choose the maximum C that the maximum gradation value in this rectangle is window width in window technique_max, choosing the minimum gradation value in this rectangle is the minima C of window width in window technique_min；Then arbitrarily select a formula in following three formula, the pixel value in ellipse mapped:

$I=\mathrm{\α}\×\frac{I_{ori}-C_{min}}{C_{max}-C_{min}}---\left(1\right)$

$I=\mathrm{\α}\×log(1+\frac{I_{ori}-C_{\min }}{C_{\max }-C_{\min }})---\left(2\right)$

$I=\mathrm{\α}\×{\left(\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}\right)}^{{\mathrm{\γ}}_1}---\left(3\right)$

Wherein, I_oriFor original CT or MR image, α is the coefficient controlling image overall brightness, γ₁For adjustable coefficient, by changing γ₁Value obtain the mapping curve of different modes, I is the image after window width and window level regulates.

3. the tumor region image enchancing method strengthening image based on synthesis as claimed in claim 2, it is characterised in that: in step A, γ₁Value less than 1 time, input value narrower for scope is transformed into the output valve an of wider range by formula (3).

4. the tumor region image enchancing method strengthening image based on synthesis as claimed in claim 2, it is characterised in that: the anisotropy parameter method in step B is: using Anisotropic Diffusion Model to remove noise, the formula of Anisotropic Diffusion Model is:

$\frac{\∂I}{\∂t}=div\left(c\right(x,y,t)\▿I)=\▿c\·\▿I+c(x,y,t)\mathrm{\Δ}I---\left(4\right)$

Wherein, div () is divergence operator, c (x, y, t) it is diffusion coefficient, diffusion coefficient is selected as the function of image gradient, and t is the time, represents gradient operator, c is then the gradient of diffusion coefficient c, I is the gradient of image I, and Δ represents Laplace operator, and Δ I is the Laplace operator of image I；

The Solving Partial Differential Equations that formula (4) is represented, the solution of the equation is expressed as I_rn, I_rnNoise-reduced image for the step B ROI region covering whole tumor regions obtained.

5. the as claimed in claim 4 tumor region image enchancing method strengthening image based on synthesis, it is characterised in that: in step B, diffusion coefficient c in the formula of described Anisotropic Diffusion Model (x, y, function t) has two kinds of expression waies:

$c\left(\right||\▿I|\left|\right)=e^{-{\left(\right||\▿I||/K)}^2}---\left(5\right)$

$c\left(\right||\▿I|\left|\right)=\frac{1}{1+{\left(\frac{\left|\right|\▿I\left|\right|}{K}\right)}^2}---\left(6\right)$

Wherein, | | I | | represents the mould of gradient of image I, and K is constant, and the two diffusion coefficient function is using the mould of the gradient of image I as the foundation of diffusion speed, and the position diffusion coefficient big in gradient is little, thus playing the purpose of Protect edge information.

6. the tumor region image enchancing method strengthening image based on synthesis as claimed in claim 4, it is characterised in that: the multi-scale enhancement method in step C is:

The ROI region covering whole tumor regions is carried out the enhancement process under different scale:

First, by formula (7), the ROI region covering whole tumor regions is carried out down-sampling, obtains the image I after down-sampling_re；

I_re=REDUCE (I) (7)

Wherein, REDUCE () represents down-sampling operation；

Then, by formula (8), to the image I after down-sampling_reZoom in and out enhancement process；

$I_u={(I_{\mathrm{re}}-I_{\mathrm{re}}*G)}^{{\mathrm{\γ}}_2}---\left(8\right)$

Wherein, I_uIt is the down-sampling enhancing image of size reduction, I_reBeing the image after down-sampling, G is gaussian kernel, I_re-I_re* the radio-frequency component of G representative image, γ₂For controlling the coefficient of tumor region and the difference of non-tumor region, I_uSize and down-sampling after image I_reEquivalently-sized, be respectively less than through window width and window level regulate after image I；

Then, by formula (9), the down-sampling of the size reduction that formula (8) obtains is strengthened image I_uCarry out up-sampling, obtain I_ex: I_ex=EXPAND (I_u)(9)

Wherein, I_exFor the enhancing image of the step C ROI region covering whole tumor regions obtained, I_exSize with regulate through window width and window level after image I equivalently-sized；EXPAND () represents up-sampling operation.

7. the tumor region image enchancing method strengthening image based on synthesis as claimed in claim 6, it is characterized in that: the EXPAND () in step C is obtained by extension interpolation operator, if extending the size of a times, then strengthen the pixel of a times in both the horizontal and vertical directions, namely needing between each row any two pixel to insert a value, every two in the ranks need to insert a line；Interpolation operator adopts bi-cubic interpolation method.

8. the tumor region image enchancing method strengthening image based on synthesis as claimed in claim 6, it is characterised in that: adopting the weighted average fusion method of Pixel-level to be weighted merging in step D, formula is as follows:

I_en=w I_rn+(1-w)·I_ex(10)

Wherein, I_rnFor the noise-reduced image of the step B ROI region covering whole tumor regions obtained, I_exFor the enhancing image of the step C ROI region covering whole tumor regions obtained, w is weighter factor, I_enSynthesis for covering the ROI region of whole tumor region strengthens image；By selecting weighter factor w different in formula (10), doctor is facilitated to observe surface and the border of tumor.

9. the tumor region Image Intensified System strengthening image based on synthesis, it is characterised in that: this system includes ROI region and chooses unit, window width and window level regulon, noise reduction unit, enhancement unit, Weighted Fusion unit, wherein:

Described ROI region choose unit for: in the frame original CT image comprising tumor or MR image, choose one cover whole tumor regions oval region of interest ROI；

Described window width and window level regulon is used for: the pixel beyond ROI region is all set to zero, and the pixel value within ROI region carries out the adjustment of window width and window level, so as to the requirement that the compound mankind observe；

Described noise reduction unit is used for: adopt anisotropy parameter method, the ROI region in the image after regulating through window width and window level is carried out noise reduction process, obtains covering the noise-reduced image of the ROI region of whole tumor region；

Described enhancement unit is used for: adopt multi-scale enhancement method, the ROI region in the image after regulating through window width and window level is carried out enhancement process, obtains covering the enhancing image of the ROI region of whole tumor region；

Described Weighted Fusion unit is used for: being weighted merging with the enhancing image of the ROI region covering whole tumor regions to the noise-reduced image of ROI region covering whole tumor regions, the synthesis obtaining covering the ROI region of whole tumor region strengthens image.

10. the tumor region Image Intensified System strengthening image based on synthesis as claimed in claim 9, it is characterized in that: the process of the adjustment that the pixel value within ROI region is carried out window width and window level by described window width and window level regulon is: centered by oval barycenter, select a rectangular area, the length r of this rectangle_lengthLess than oval major axis, the width r of this rectangle_widthLess than oval short axle；Choose the maximum C that the maximum gradation value in this rectangle is window width in window technique_max, choosing the minimum gradation value in this rectangle is the minima C of window width in window technique_min；Arbitrarily select a formula in following three formula, the pixel value in ellipse mapped:

$I=\mathrm{\α}\×\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }}---\left(1\right)$

$I=\mathrm{\α}\×log(1+\frac{I_{ori}-C_{\min }}{C_{max}-C_{\min }})---\left(2\right)$

$I=\mathrm{\α}\×{\left(\frac{I_{ori}-C_{min}}{C_{max}-C_{min}}\right)}^{{\mathrm{\γ}}_1}---\left(3\right)$

Wherein, I_oriFor original CT or MR image, α is the coefficient controlling image overall brightness, γ₁For adjustable coefficient, by changing γ₁Value obtain the mapping curve of different modes, I is the image after window width and window level regulates.

11. the tumor region Image Intensified System strengthening image based on synthesis as claimed in claim 10, it is characterised in that: γ₁Value less than 1 time, input value narrower for scope is transformed into the output valve an of wider range by formula (3).

12. the tumor region Image Intensified System strengthening image based on synthesis as claimed in claim 10, it is characterized in that: the anisotropy parameter method that described noise reduction unit adopts is: using Anisotropic Diffusion Model to remove noise, the formula of Anisotropic Diffusion Model is:

$\frac{\∂I}{\∂t}=div\left(c\right(x,y,t)\▿I)=\▿c\·\▿I+c(x,y,t)\mathrm{\Δ}I---\left(4\right)$

Wherein, div () is divergence operator, c (x, y, t) it is diffusion coefficient, diffusion coefficient is selected as the function of image gradient, and t is the time, represents gradient operator, c is then the gradient of diffusion coefficient c, I is the gradient of image I, and Δ represents Laplace operator, and Δ I is the Laplace operator of image I；

The Solving Partial Differential Equations that formula (4) is represented, the solution of the equation is expressed as I_rn, I_rnFor covering the noise-reduced image of the ROI region of whole tumor region.

13. the as claimed in claim 12 tumor region Image Intensified System strengthening image based on synthesis, it is characterised in that: diffusion coefficient c in the formula of described Anisotropic Diffusion Model (x, y, function t) has two kinds of expression waies:

$c\left(\right||\▿I|\left|\right)=e^{-{\left(\right||\▿I||/K)}^2}---\left(5\right)$

$c\left(\right||\▿I|\left|\right)=\frac{1}{1+{\left(\frac{\left|\right|\▿I\left|\right|}{K}\right)}^2}---\left(6\right)$

Wherein, | | I | | represents the mould of gradient of image I, and K is constant, and the two diffusion coefficient function is using the mould of the gradient of image I as the foundation of diffusion speed, and the position diffusion coefficient big in gradient is little, thus playing the purpose of Protect edge information.

14. the tumor region Image Intensified System strengthening image based on synthesis as claimed in claim 12, it is characterised in that: the multi-scale enhancement method that described enhancement unit adopts is:

The ROI region covering whole tumor regions is carried out the enhancement process under different scale:

First, by formula (7), the ROI region covering whole tumor regions is carried out down-sampling, obtains the image I after down-sampling_re；

I_re=REDUCE (I) (7)

Wherein, REDUCE () represents down-sampling operation；

Then, by formula (8), to the image I after down-sampling_reZoom in and out enhancement process；

$I_u={(I_{\mathrm{re}}-I_{\mathrm{re}}*G)}^{{\mathrm{\γ}}_2}---\left(8\right)$

Wherein, I_uIt is the down-sampling enhancing image of size reduction, I_reBeing the image after down-sampling, G is gaussian kernel, I_re-I_re* the radio-frequency component of G representative image, γ₂For controlling the coefficient of tumor region and the difference of non-tumor region, I_uSize and down-sampling after image I_reEquivalently-sized, be respectively less than through window width and window level regulate after image I；

Then, by formula (9), the down-sampling of the size reduction that formula (8) obtains is strengthened image I_uCarry out up-sampling, obtain I_ex: I_ex=EXPAND (I_u)(9)

Wherein, I_exFor the enhancing image of the step C ROI region covering whole tumor regions obtained, I_exSize with regulate through window width and window level after image I equivalently-sized；EXPAND () represents up-sampling operation.

15. the tumor region Image Intensified System strengthening image based on synthesis as claimed in claim 14, it is characterized in that: described EXPAND () is obtained by extension interpolation operator, if extending the size of a times, then strengthen the pixel of a times in both the horizontal and vertical directions, namely needing between each row any two pixel to insert a value, every two in the ranks need to insert a line；Interpolation operator adopts bi-cubic interpolation method.

16. the tumor region Image Intensified System strengthening image based on synthesis as claimed in claim 14, it is characterised in that: described Weighted Fusion unit adopts the weighted average fusion method of Pixel-level to be weighted merging, and formula is as follows:

I_en=w I_rn+(1-w)·I_ex(10)

Wherein, I_rnFor covering the noise-reduced image of the ROI region of whole tumor region, I_exFor covering the enhancing image of the ROI region of whole tumor region, w is weighter factor, I_enSynthesis for covering the ROI region of whole tumor region strengthens image；By selecting weighter factor w different in formula (10), doctor is facilitated to observe surface and the border of tumor.