CN114862773A - Liver medical image registration and fusion method and system - Google Patents

Abstract

The invention discloses a liver medical image registration and fusion method and a system, wherein the method comprises the following steps: acquiring a liver CT image sequence map group of an arterial phase and a liver CT image sequence map group of a venous phase of the same liver, wherein the liver CT image sequence map group of the arterial phase comprises at least ten liver CT images of the arterial phase, and the liver CT image sequence map group of the venous phase comprises at least thirty liver CT images of the venous phase; sequentially registering each arterial-phase liver CT image, finding venous-phase liver CT images with the same cross section as the venous-phase liver CT images in a venous-phase liver CT image sequence map group, and acquiring at least ten venous-phase liver CT images registered with at least ten arterial-phase liver CT images; sequentially fusing each arterial-stage liver CT image and one venous-stage liver CT image registered with each arterial-stage liver CT image to obtain at least ten liver CT fused images; and performing sequence three-dimensional reconstruction on at least ten liver CT fusion images to obtain a liver three-dimensional fusion CT image.

Description

Liver medical image registration and fusion method and system

technical field

The present invention relates to the technical field of medical image processing, in particular to a method and system for processing liver medical images.

Background technique

The liver is an important organ in the human body. There are many complex piping systems in the liver, such as hepatic artery, hepatic vein, portal vein, and bile duct system. Image processing, visualization research, and biological modeling are of great significance.

The visualization study of the liver can be applied to clinical diagnosis, surgery and liver transplantation. It can help doctors to understand the distribution and connection of the piping system in the liver before surgery, and provide assistance for undiagnosed positioning and surgery.

Medical image registration refers to seeking one or a series of spatial transformations for a medical image to make it spatially consistent with the corresponding points on another medical image. This agreement means that the same anatomical point on the human body has the same spatial location on the two matched images. The result of the registration should match all anatomical points, or at least all points of diagnostic significance, on the two images.

Image fusion (Image Fusion) refers to the image data collected by multiple source channels about the same target through image processing and computer technology processing, etc., to maximize the extraction of favorable information in each channel, and finally synthesized into high-quality images. In order to improve the utilization rate of image information, improve the accuracy and reliability of computer interpretation, and improve the spatial resolution and spectral resolution of the original image, it is conducive to monitoring. The images to be fused have been registered and the pixel bit widths are consistent, and the information of two or more multi-source images is synthesized and extracted. The efficient image fusion method can comprehensively process the information of multi-source channels according to the needs, thereby effectively improving the utilization rate of image information, the reliability of the system for target detection and recognition, and the automation degree of the system. Its purpose is to integrate the multi-band information of a single sensor or the information provided by different types of sensors, to eliminate the possible redundancy and contradiction between the multi-sensor information, to enhance the transparency of the information in the image, and to improve the accuracy and reliability of interpretation. and usage rates to form a clear, complete, and accurate information description of the target.

Traditionally, enhanced liver CT scans are divided into arterial phase, portal venous phase and delayed phase. During the scan, the examiner is in a supine position and holds his breath. The operator injects a certain dose of contrast agent into the patient at a certain flow rate. After the injection, the whole hepatic arterial phase scan is performed first; The whole hepatic portal venous phase was scanned; finally, the lesion area was scanned after a certain interval. The portal venous phase and the delayed phase last for a relatively long time, and the requirements for the scanning time are relatively loose. The duration of the arterial phase is short, and the peak disappears quickly.

Images in different phases are images captured by CT scans when the contrast agent flows through different blood vessels inside the liver, and have different characteristics. The arterial phase image is the image obtained when the contrast agent flows through the hepatic artery. Therefore, the hepatic artery in the arterial phase image is displayed with high brightness and can be easily distinguished from the image. Contrast agent is present, so these ducts are numerically close to the liver parenchyma and cannot be easily distinguished from the image. Similarly, in the portal venous phase image, it is easy to distinguish the hepatic vein and the portal vein from the image.

In the diagnosis of liver disease, it is often necessary to analyze several multi-phase images at the same location, and to observe different duct structures in the liver at the same location at the same time to provide a basis for diagnosis. However, since the starting position of the CT scan may not be exactly the same, and the patient may have breathing motion or other slight body position movement during the scanning process, the scanning image sequences of different phases of the liver, their spatial positional relationship is not the same. There is not a one-to-one correspondence, and the images during different phases are not perfectly aligned. For example, the fifth image in the arterial phase does not necessarily correspond to the fifth image in the venous phase, they may represent different locations of the liver. At present, doctors can only select and match different images of the same location based on subjective experience, which brings difficulties and inconvenience to diagnostic analysis. Therefore, before fusion of liver CT multiphase images, these images need to be aligned first, that is, image registration of liver CT multiphase images is required. The quality of the registration results directly affects the quality of image fusion, which may even affect the doctor's diagnosis.

For example, Chinese Patent Application Publication No. 104835112B discloses a liver multiphase CT image fusion method. First, the multi-resolution CT image registration method based on joint histogram is used to perform rough registration on the source image sequence, and then combined with the confidence connection The region growing algorithm realizes the automatic segmentation of the liver image and the liver image segmentation based on the gradient vector flow snake model, and effectively extracts the edge information of the liver; then the liver image is extracted based on the directional region growing algorithm, and then the liver parenchyma image is processed. Based on B-spline free deformation transformation and liver non-rigid registration based on spatially weighted mutual information, the image pairs in the same spatial position can be accurately found. Finally, image fusion is performed based on wavelet transform. However, the CT image fusion method in the multiphase phase of the liver disclosed in this patent application uses wavelet transform to fuse CT images, and the details of the obtained fused images are severely damaged.

Another example is a method for heterogeneous fusion of PET images and CT images based on wavelet transform disclosed in Chinese Patent Application Publication No. 101987019A, comprising the steps of: fixing the patient in a positioning frame; collecting PET images according to the frame; The frame collects CT images; the PET/CT images are extracted and processed separately; the PET/CT images after step processing are registered according to the positioning points obtained in the positioning frame; the registered PET/CT images are subjected to wavelet decomposition and transformation. Finally, the fused PET/CT image is subjected to wavelet inverse transformation to generate a fusion image. However, the image fusion method disclosed in this patent application is mainly used for fusion between PET images and CT images, and cannot process CT images in different phases.

Therefore, to provide a liver medical image registration and fusion method and system that can first perform image registration on liver CT multiphase images and then perform fusion, which can quickly obtain clear, accurate and comprehensive liver images, has become an urgent solution in the industry. The problem.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a liver medical image registration and fusion method and system, which can register the liver CT images at the same position in different phases and then perform rapid fusion to obtain a clear, accurate and comprehensive liver image. , to help with subsequent diagnosis.

In order to achieve the above object, the first object of the present invention is to provide a liver medical image registration and fusion method, which includes the following steps: (1) Obtaining an arterial phase liver CT image sequence map group and veins for the same liver The liver CT image sequence map group in the arterial phase, the arterial phase liver CT image sequence map group includes at least ten arterial phase liver CT images taken from different sections of the same liver, and the venous phase liver CT image sequence map group includes the same liver. At least thirty venous phase liver CT images obtained from different sections; (2), sequentially preprocessing at least ten arterial phase liver CT images and at least thirty venous phase liver CT images; (3), arterial phase liver CT images Each arterial phase liver CT image in the liver CT image sequence map group is registered by the multi-resolution CT image registration method based on joint histogram in turn, and found in the venous phase liver CT image sequence map group. Obtaining at least ten venous phase liver CT images registered with at least ten arterial phase liver CT images of the venous phase liver CT images of the same cross-section; (4), sequentially analyzing each image in the arterial phase liver CT image sequence map group Perform image fusion on an arterial phase liver CT image and a registered venous phase liver CT image through frequency-domain non-subsampling contourlet transform based on graphics processor hardware acceleration, and obtain at least ten liver CT fusion images; and (5) Sequence three-dimensional reconstruction is performed on at least ten liver CT fusion images to obtain liver three-dimensional fusion CT images.

Optionally, step (2) sequentially includes: sequentially performing denoising processing, contrast enhancement processing, and target area separation processing on each liver CT image in the liver CT image sequence group.

Preferably, in the denoising processing step of the present invention, Gaussian smoothing filtering is used for denoising, and the subtle discontinuous pixels of the image are filled smoothly through the neighborhood operation, so as to reduce the leakage phenomenon during contour extraction and improve the image quality.

Preferably, the contrast enhancement processing step of the present invention may adopt grayscale transformation enhancement or spatial filtering enhancement, wherein the grayscale transformation enhancement may be direct grayscale transformation enhancement or histogram correction transformation enhancement, and the spatial filtering enhancement may be smooth filtering enhancement or Sharpening filter enhancement.

Optionally, step (3) includes: (3-1) determining an arterial phase liver CT image in the arterial phase liver CT image sequence map group as the arterial phase liver CT image to be registered, and selecting a venous phase liver CT image A venous phase liver CT image in the sequence map group is used as the venous phase liver CT image to be registered. The arterial phase liver CT image pyramid and the venous phase liver CT image pyramid, of which the top image resolution of the arterial phase liver CT image pyramid and the venous phase liver CT image pyramid is the smallest, and the bottom layer is the original image; (3-2) For the arterial phase liver CT image pyramid The specified area of the top image of the liver CT image pyramid is subjected to geometric coordinate rigid body transformation to obtain a new area; (3-3) The top image of the liver CT image pyramid in the venous phase is obtained by the cubic B-spline interpolation method. The top image obtained in step (3-2) The coordinates of the new area; (3-4) Calculate the similarity between the top-level image of the arterial phase liver CT image pyramid and the interpolation map using the similarity measure based on the joint histogram, and obtain the similarity function; (3-5) Step (3-4) The obtained similarity function is input into the particle swarm optimization algorithm for optimization calculation to obtain the optimal transformation parameters, and steps (3-2) to (3-4) are repeated until the maximum value is obtained; (3- 6) Take the optimal transformation parameter as the starting point of Powell's algorithm search, and process the last two images of the liver CT image pyramid in the arterial phase and the last two layers of the liver CT image pyramid in the venous phase respectively, and repeat steps (3-2) to (3-5). ); and (3-7) output the registration images of the venous phase liver CT images under the final transformation; after registration, obtain at least ten venous phase liver CT images registered with at least ten arterial phase liver CT images.

Among them, the joint histogram of the two images in step (3-4) is defined as: the statistical probability distribution of the gray value of the corresponding pixels in the two images. For discrete images, the probability density is:

P(m,n)=N(m,n)/M

Among them, M is the total number of pixels contained in an image; N(m, n) is the total number of pixels whose gray values are m and n respectively in the two images, and the gray value of one image is the abscissa, The gray value of the other image is the ordinate, and the drawn probability distribution map is called the joint histogram.

The joint histogram-based registration similarity measure is as follows:

The threshold h takes 10% of the pixel value range in the image to obtain better results.

时，就能够计算相似函数中

的值。During implementation, the number of points in the threshold region is calculated, and the solution is

, it is possible to calculate the similarity function in

value of .

Further, in the step (3-5), a search problem is set, the particle swarm of scale R is optimized in the D-dimensional space, and the particle i (1≤i≤R) is in the nth (1≤n≤Nmax) The velocity of generation is represented as Vi(n)=(vi1,vi2,…,viD), and the position is represented as Xi(n)=(xi1,xi2,…,xiD); in order to maintain the velocity inertia of each particle, inertia is introduced The constant ω is the quantity; when the value of ω is small, the particle can finally converge to the optimal position; when the value of ω is large, the search ability of the particle in the global scope can be improved; the improved particle velocity update formula is as follows:

v _ij (n)=ωv _ij (n-1)+α ₁ β ₁ [p _ij -x _ij (n-1)]+α ₂ β ₂ [p _uj -x _ij (n-1)]

Among them, n is the number of iterations, α ₁ and α ₂ are acceleration constants, and the values of β ₁ and β ₂ are random numbers in the range of [0, 1]. p _ij is the value of the historical best position of particle i in the j dimension, and p _uj is the value of the historical best position of all particles in a certain local area in the j dimension. If the particles contained in this area are the entire swarm, then p _u represents the best historical position of the entire particle swarm, otherwise p _u represents the local best position. The algorithm will terminate the iteration when the number of algorithm iterations reaches the maximum value or when the best historical position of the particle search is lower than the set threshold and the distance between particles in the parameter space is close to the set threshold.

和高频融合图像

(4-4)对融合结果进行FFT，即快速傅里叶变换，对整个过程进行频域非下采样轮廓波重建变换，得到该截面的肝脏CT融合图像结果；以及(4-5)依次选择动脉期肝脏CT图像序列图组中的其他的一幅动脉期肝脏CT图像及与其处于同一截面的静脉期肝脏CT图像，重复步骤(4-2)～(4-4)，直至获取至少十幅肝脏CT融合图像。Optionally, step (4) includes: (4-1) selecting an arterial phase liver CT image and a venous phase liver CT image in the same section of the arterial phase liver CT image sequence image group; (4-2) Perform FFT on the selected arterial phase liver CT images and venous phase liver CT images, that is, fast Fourier transform, to obtain image frequency domain signals, and perform frequency domain non-subsampling contourlet decomposition transformation on the image frequency domain signals to obtain the transformation results. (4-3) IFFT is performed on the transformation result, that is, inverse fast Fourier transform, and corresponding transformation is taken to the low-frequency part ^CL and the high-frequency part ^CH respectively, and the low-frequency fusion image is obtained.

and high-frequency fusion images

(4-4) Perform FFT on the fusion result, that is, fast Fourier transform, perform frequency domain non-subsampling contourlet reconstruction transformation on the whole process, and obtain the result of the liver CT fusion image of the section; and (4-5) Select in turn For another arterial phase liver CT image and the venous phase liver CT image in the same section of the arterial phase liver CT image sequence image group, repeat steps (4-2) to (4-4) until at least ten images are acquired Liver CT fusion images.

Among them, the non-subsampling contourlet transform in the frequency domain in step (4) can decompose the image information at different frequencies, which can be divided into two parts: high frequency and low frequency, wherein the high frequency part is the frequency obtained by the directional filter transformation , signals in each direction, and the low-frequency part is the low-frequency signal generated by the first pyramid decomposition. Since the previous decomposition was carried out in the frequency domain, it is necessary to perform sequential IFFT transformation on the decomposed image (inverse fast Fourier transform) Lie transform), convert the result in the frequency domain to the time domain, and then proceed to the next operation.

1) Low frequency fusion rules

表示低频融合后的图像结果，由于在图像融合之前，两幅图像已经经过了诸如去噪、配准、灰度矫正等处理，因此可以直接在低频部分对两幅图像进行操作。For convenience of representation, let the two source images to be fused are C _1(x,y) and C _2(x,y) respectively,

Represents the image result after low-frequency fusion. Since the two images have undergone processing such as denoising, registration, and grayscale correction before image fusion, the two images can be directly manipulated in the low-frequency part.

Because after multi-scale decomposition, the low-frequency information of the image mainly determines its general outline, and the high-frequency information is the key to determine the clarity of the fusion result. Therefore, the difference between the low-frequency coefficients of the two images after multi-scale decomposition is far is smaller than the difference between high frequency coefficients, so the fusion of wavelet low frequency coefficients can be expressed as:

In the formula, the superscript L represents that the image is a low frequency (Low Frequency) signal, and α and β are used as low frequency fusion coefficients. In the present invention, α=β=0.5.

2) High frequency fusion rules

High frequency parts can reflect image details. At the edge of the image, the decomposition result of the high-frequency part of the period is larger, and the non-edge part is lower. Therefore, in order to keep the edge features of the two images as much as possible, the present invention adopts a high-frequency image fusion method based on regional energy, that is, the pixel energy of the images generated by comparing the filtering results in different directions for sampling 3×3 or 5×5 regions is calculated respectively. The size E, by setting an appropriate threshold, is determined by the following formula:

In the formula, the superscript H indicates that the image is a high-frequency signal, and the energy conformity function is:

Here R(x,y) ranges between 0 and 1. And, T is the threshold of the energy compliance, and ω ₁ +ω ₂ =1. According to the above formula, the edge information of the generated fused image is well preserved, and through the selection of the threshold, the residual high-frequency noise in the image can be suppressed to a certain extent.

Optionally, step (5) includes: (5-1) setting the value of the isosurface, extracting the target contour of the liver in at least ten liver CT fusion images, calculating the area of the liver contour, and finding the difference between the liver contours of different sections. (5-2) First read the liver contours in two liver CT fusion images of at least ten liver CT fusion images, and then read a slice of adjacent liver CT fusion images each time , the adjacent four of the pixels on each slice and the four pixels of the corresponding next slice form a cube, which is called a voxel, and then processed sequentially from left to right and front to back All adjacent cubes in one layer, identify boundary voxels, extract isosurfaces, and then continue to read the next adjacent slice of liver CT fusion image in the same way after processing one layer, until all slices are processed. After the isosurface is extracted, the algorithm ends, and the three-dimensional fusion CT image of the liver is obtained.

The second object of the present invention is to provide a liver medical image registration and fusion system, which includes: an image acquisition device for obtaining an arterial phase liver CT image sequence map group and a venous phase liver CT image for the same liver Sequence diagram group, the arterial phase liver CT image sequence diagram group includes at least ten arterial phase liver CT images taken from different sections of the same liver, and the venous phase liver CT image sequence diagram group includes the same liver taken from different sections at least thirty venous phase liver CT images; an image preprocessing device, which is connected in communication with the image acquisition device, and is used to sequentially preprocess at least ten arterial phase liver CT images and at least thirty venous phase liver CT images; An image registration device, which is connected in communication with the preprocessing device, is used for sequentially using a multi-resolution CT image registration method based on joint histogram for each arterial phase liver CT image in the arterial phase liver CT image sequence map group Perform registration, find the venous phase liver CT images from the same section in the venous phase liver CT image sequence map group, and obtain at least ten venous phase liver CT images registered with at least ten arterial phase liver CT images; image The fusion device is connected in communication with the image registration device, and is used for sequentially performing graphics-based processing on each arterial phase liver CT image in the arterial phase liver CT image sequence map group and a venous phase liver CT image registered therewith hardware-accelerated frequency-domain non-subsampling contourlet transform to achieve image fusion to obtain at least ten liver CT fusion images; and a three-dimensional reconstruction device, which is connected in communication with the image fusion device and used to sequence at least ten liver CT fusion images 3D reconstruction to obtain a 3D fusion CT image of the liver.

The image acquisition device is installed on each user terminal that logs in to the liver medical image registration and fusion system with a specific registered user account through the Internet. The image acquisition device includes a log-in user to upload liver CT image data and its specific liver source information. An uploading unit, and a downloading unit connected in communication with the liver database to store the liver source information and the three-dimensional image of the liver from the three-dimensional reconstruction device together in the liver database.

Preferably, the source information of the liver at least includes: the type of disease, the gender of the patient, the age of the patient, the living area and the hospital for treatment.

More preferably, the source information of the liver may also include: the patient's living habits, biochemical examination information, the patient's typical symptoms, signs, medical images, imaging diagnosis results, treatment plans, adverse reactions, and attending physicians.

Optionally, the image preprocessing device includes: a denoising unit for performing denoising processing on each liver CT image in the liver CT image sequence group, an enhancement unit for performing contrast enhancement processing, and a target point, which are sequentially connected in communication Separation unit for zone separation processing.

Optionally, the image registration device includes: a resolution processing unit, which is configured to determine an arterial phase liver CT image in the arterial phase liver CT image sequence map group as the arterial phase liver CT image to be registered, and select the venous phase liver CT image. A venous phase liver CT image in the CT image sequence map group is used as the venous phase liver CT image to be registered, and the arterial phase liver CT image and the venous phase liver CT image to be registered are respectively subjected to multi-resolution processing. The arterial phase liver CT image pyramid and the venous phase liver CT image pyramid, in which the top image resolution of the arterial phase liver CT image pyramid and the venous phase liver CT image pyramid is the smallest, and the bottom layer is the original image; the rigid body transformation unit, which is the same as The resolution processing unit is communicatively connected, and is used for performing geometric coordinate rigid body transformation on the designated area of the top-level image of the liver CT image pyramid in the arterial phase to obtain a new area; the coordinate acquiring unit is communicatively connected with the rigid body transformation unit, and is used for interpolation through cubic B-splines The method obtains the coordinates of the top image of the venous phase liver CT image pyramid in the new area acquired by the rigid body transformation unit; the function acquisition unit, which is connected in communication with the coordinate acquisition unit, is used to calculate the arterial phase liver CT using the similarity measure based on the joint histogram. The similarity between the top-level image of the image pyramid and the interpolation map is used to obtain the similarity function; the parameter acquisition unit is connected in communication with the function acquisition unit, and is used to input the similarity function obtained by the function acquisition unit into the particle swarm optimization algorithm. Carry out the optimization calculation to obtain the optimal transformation parameters, and then start the rigid body transformation unit, the coordinate acquisition unit, and the function acquisition unit until the maximum value is obtained; the reprocessing unit, which is connected in communication with the parameter acquisition unit, is used to convert the optimal transformation parameters. As the starting point of Powell's algorithm search, the last two images of the liver CT image pyramid in the arterial phase and the last two layers of the liver CT image pyramid in the venous phase are processed, and the rigid body transformation unit, coordinate acquisition unit, function acquisition unit, and parameter acquisition unit are activated again. ; and an image output unit, which is connected in communication with the reprocessing unit and is used for outputting a registered image of the venous phase liver CT image under the final transformation.

和高频融合图像

重建变换单元，其与频域融合单元通信连接，用于对低频融合图像

和高频融合图像

进行快速傅里叶变换，对整个过程进行频域非下采样轮廓波重建变换，得到该截面的肝脏CT融合图像结果。Optionally, the image fusion device includes: an image selection unit for selecting an arterial phase liver CT image and a venous phase liver CT image in the same section of the arterial phase liver CT image sequence image group; an image frequency domain unit , which is connected in communication with the image selection unit, and is used to perform fast Fourier transform on the selected arterial phase liver CT images and venous phase liver CT images to obtain image frequency domain signals, and perform frequency domain non-subsampling on the image frequency domain signals respectively. The contourlet is decomposed and transformed to obtain the transformation result; the frequency domain fusion unit, which is connected in communication with the image frequency domain unit, is used to perform inverse fast Fourier transform on the transformation result, and take corresponding steps for the low-frequency part ^CL and the high-frequency part ^CH respectively. transformation to obtain a low-frequency fusion image

and high-frequency fusion images

The reconstruction transform unit, which is connected in communication with the frequency domain fusion unit, is used to fuse the low-frequency image

and high-frequency fusion images

Fast Fourier transform is performed, and the whole process is reconstructed by non-subsampling contourlet in frequency domain, and the result of liver CT fusion image of this section is obtained.

Optionally, the three-dimensional reconstruction device includes: a contour matching unit, which is used to set the value of the isosurface, extract the target contour of the liver in at least ten liver CT fusion images, calculate the area of the liver contour, and find the liver contours of different sections. The sequence relationship and matching between them; the MC reconstruction unit, which is connected in communication with the contour matching unit, is used to first read the liver contours in the two liver CT fusion images in the at least ten liver CT fusion images, and then read the liver contours each time. Into a slice of adjacent liver CT fusion images, the adjacent four pixels on each slice and the corresponding four pixels of the next slice form a cube, which is called a voxel, and then from the Process all adjacent cubes in one layer sequentially from left to right and from front to back, identify boundary voxels, extract isosurfaces, and then continue to read the next adjacent liver CT fusion after processing one layer in the same way The slice of the image, until all slices are processed, the isosurface is extracted, the algorithm ends, and the three-dimensional fusion CT image of the liver is obtained.

Among them, the MC reconstruction unit extracts the isosurface of the liver CT image of the specific liver according to the MC algorithm to realize the three-dimensional reconstruction of the specific liver. When the MC reconstruction unit adopts the MC algorithm, it obtains its internal isosurface from each voxel as follows: each voxel has eight vertices, and the grayscale values of these eight vertices can be obtained directly from the pixel values of the input slice, The threshold of the isosurface to be extracted also needs to be given by the user before the calculation. Among these eight vertices, we mark the vertices whose gray value is greater than the given threshold as black, and the vertices whose gray value is less than the threshold are not marked.

Obviously, if there are both "marked" and "unmarked" points in a cube, then there must be isosurfaces in the cube. Except for the cases where all marked and all marked voxels do not contain isosurfaces, all 8 vertices in a voxel cube may have two states of "marked" and "unmarked". There are a total of 256 possible distributions of isosurfaces on a voxel. Since the rotation of the cube does not affect the topology of the isosurface, the situation with rotational symmetry in the cube can be removed. In addition, all "unmarked" and "marked" can be interchanged, which also does not change the topology of the isosurface. So only 15 basic cubes are needed to represent all 256 possible cases. The 15 cases are constructed into an index table with a length of 256, which records 256 possible connection methods of isosurfaces within a voxel. After obtaining the marking of the eight vertices, according to the marking, an index value between 0 and 15 is obtained, and then directly comparing the index table according to the index value can know which edge of the voxel cube the equivalent point is on. , and the connection method of the voxel's iso-value points can also be obtained from the index table. At this time, the iso-value points can be connected to form an iso-value surface.

Optionally, the three-dimensional reconstruction device may further draw the liver three-dimensional fusion CT image entity by defining the illumination, viewing angle and focus information of the specified scene.

Optionally, two-phase liver CT image sequence groups of the liver as the original two-dimensional image information are stored in the liver database together with the three-dimensional images of the liver to facilitate user research and analysis.

In addition, the present invention can also select other feasible technologies in the process of multi-phase fusion of liver CT images, such as image fusion methods such as HIS transform fusion method and principal component analysis fusion method.

The beneficial effects of the present invention are as follows: (1), CT images in different phases at the same position of the liver can be matched and registered, which effectively improves the image quality and facilitates diagnosis; (2), in the process of image fusion, adopting The frequency-domain non-subsampling contourlet transform not only has multi-resolution characteristics and time-frequency locality, but also has more direction selectivity and translation without deformation. Therefore, the fused liver CT image of the present invention retains more The salient features of the source image and different frequency domains make the fusion image clearer and more complete; (3) Doctors can retrieve patient data, find the required data and image data from the liver database at any time, which is useful for demonstration teaching and occupation of doctors in related fields. Growth helps a lot.

Description of drawings

FIG. 1 is a schematic diagram of steps of the method for registration and fusion of liver medical images according to the present invention.

FIG. 2 is a schematic structural diagram of the liver medical image registration and fusion system of the present invention.

FIG. 3 is a schematic structural diagram of an image registration device and an image fusion device of the present invention.

Detailed ways

The following describes in detail the embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary, and are intended to explain the present invention and should not be construed as limiting the present invention.

First, please refer to FIG. 1 , as a non-limiting embodiment, in the method for registration and fusion of liver medical images according to the present invention, firstly, in step S1 , a sequence map group of liver CT images and veins in the arterial phase of the same liver are obtained. The liver CT image sequence map group in the arterial phase, the arterial phase liver CT image sequence map group includes at least ten arterial phase liver CT images taken from different sections of the same liver, and the venous phase liver CT image sequence map group includes the same liver. At least thirty venous phase liver CT images were taken from different sections. Next, in step S2, at least ten liver CT images in the arterial phase and at least thirty liver CT images in the venous phase are sequentially preprocessed. Subsequently, in step S3, each arterial phase liver CT image in the arterial phase liver CT image sequence map group is sequentially registered using the multi-resolution CT image registration method based on the joint histogram. Find the venous phase liver CT images taken from the same section in the CT image sequence map group, and acquire at least ten venous phase liver CT images registered with at least ten arterial phase liver CT images. Then, in step S4 , perform frequency domain non-linear image processing based on graphics processor hardware acceleration for each arterial phase liver CT image in the arterial phase liver CT image sequence map group and a venous phase liver CT image registered with it in turn. Downsampling contourlet transform realizes image fusion, and obtains at least ten liver CT fusion images. Finally, in step S5, serial three-dimensional reconstruction is performed on at least ten liver CT fusion images to obtain liver three-dimensional fusion CT images.

In this non-limiting embodiment, step S2 sequentially includes: sequentially performing a denoising processing step, a contrast enhancement processing step, and a target area separation processing step on each liver CT image in the liver CT image sequence group.

As a non-limiting implementation, step S3 specifically includes: (3-1) Determine an arterial phase liver CT image in the arterial phase liver CT image sequence map group as the arterial phase liver CT image to be registered, and select a vein A venous phase liver CT image in the sequence map group of liver CT images is used as the venous phase liver CT image to be registered. There are three layers of the arterial phase liver CT image pyramid and the venous phase liver CT image pyramid, of which the top image resolution of the arterial phase liver CT image pyramid and the venous phase liver CT image pyramid is the smallest, and the bottom layer is the original image; (3-2 ) Perform geometric coordinate rigid body transformation on the specified area of the top-level image of the liver CT image pyramid in the arterial phase to obtain a new area; (3-3) Obtain the top-level image of the liver CT image pyramid in the venous phase through the cubic B-spline interpolation method in step (3-3- 2) The acquired coordinates of the new region; (3-4) Using the similarity measure based on the joint histogram to calculate the similarity between the top-level image of the liver CT image pyramid in the arterial phase and the interpolation map to obtain a similarity function; (3- 5) Input the similarity function obtained in step (3-4) into the particle swarm optimization algorithm for optimization calculation to obtain the optimal transformation parameters, and repeat steps (3-2) to (3-4) until the maximum value is obtained (3-6) Take the optimal transformation parameter as the starting point of Powell’s algorithm search, and process the last two layers of the arterial phase liver CT image pyramid and the venous phase liver CT image pyramid respectively, and repeat steps (3-2)～ (3-5); and (3-7) outputting the registration image of the CT image of the liver in the venous phase under the final transformation; after the registration, obtain at least ten images of the liver in the venous phase that are registered with the at least ten CT images of the liver in the arterial phase CT images.

和高频融合图像

(4-4)对低频融合图像

和高频融合图像

(融合结果)进行快速傅里叶变换，对整个过程进行频域非下采样轮廓波重建变换，得到该截面的肝脏CT融合图像结果；以及(4-5)依次选择动脉期肝脏CT图像序列图组中的其他的一幅动脉期肝脏CT图像及与其处于同一截面的静脉期肝脏CT图像，重复步骤(4-2)～(4-4)，直至获取至少十幅肝脏CT融合图像。As another non-limiting embodiment, step S4 specifically includes: (4-1) selecting an arterial phase liver CT image and a venous phase liver CT image in the same section of the arterial phase liver CT image sequence map group; (4-2) Perform fast Fourier transform on the selected arterial phase liver CT images and venous phase liver CT images to obtain image frequency domain signals, and perform frequency domain non-subsampling contourlet decomposition transformation on the image frequency domain signals respectively to obtain Transform results; (4-3) Perform inverse fast Fourier transform on the transform results, and take corresponding transforms on the low-frequency part ^CL and the high-frequency part ^CH respectively to obtain a low-frequency fusion image

and high-frequency fusion images

(4-4) For low-frequency fusion images

and high-frequency fusion images

(Fusion result) Perform fast Fourier transform, perform frequency domain non-subsampling contourlet reconstruction transformation on the whole process, and obtain the result of the liver CT fusion image of the section; and (4-5) sequentially select the arterial phase liver CT image sequence diagram Repeat steps (4-2) to (4-4) for the other arterial phase liver CT image in the group and the venous phase liver CT image in the same section until at least ten liver CT fusion images are obtained.

Step S5 specifically includes: (5-1) setting the value of the isosurface, extracting the target contour of the liver in at least ten liver CT fusion images, calculating the area of the liver contour, finding the order relationship between the liver contours in different sections, and Matching; (5-2) First read the liver contours in two liver CT fusion images out of at least ten liver CT fusion images, and then read a slice of adjacent liver CT fusion images each time, and each slice is The adjacent four of the pixels and the corresponding four pixels of the next slice form a cube, which is called a voxel, and then process all the pixels in one layer sequentially from left to right and front to back. Adjacent cubes, discriminate the boundary voxels, extract the isosurface, and then continue to read the next adjacent slice of liver CT fusion image in the same way after processing one layer, until all the slices are processed and the isosurface is extracted , the algorithm ends, and the three-dimensional fusion CT image of the liver is obtained.

The liver medical image registration and fusion system provided by the present invention, as shown in FIG.

The image acquisition device 10 is used to obtain an arterial phase liver CT image sequence map group and a venous phase liver CT image sequence map group for the same liver, and the arterial phase liver CT image sequence map group includes at least the same liver taken from different sections. The ten liver CT images in arterial phase and the CT image series in venous phase include at least thirty liver CT images in venous phase taken from different sections of the same liver.

The image preprocessing device 20 is connected in communication with the image acquisition device 10, and is used for sequentially preprocessing at least ten liver CT images in the arterial phase and at least thirty liver CT images in the venous phase.

The image registration device 30 is connected in communication with the preprocessing device 20, and is used for sequentially using the multi-resolution CT image registration method based on the joint histogram for each arterial phase liver CT image in the arterial phase liver CT image sequence group. Perform registration, find the venous phase liver CT images from the same section in the venous phase liver CT image sequence map group, and obtain at least ten venous phase liver CT images registered with at least ten arterial phase liver CT images.

The image fusion device 40 is connected in communication with the image registration device 30, and is configured to sequentially perform graphics-based image-based image analysis on each arterial phase liver CT image in the arterial phase liver CT image sequence map group and a venous phase liver CT image registered therewith. The frequency domain non-subsampling contourlet transform accelerated by the processor implements image fusion, and obtains at least ten liver CT fusion images.

The three-dimensional reconstruction device 50 is connected in communication with the image fusion device 40, and is used for performing sequential three-dimensional reconstruction on at least ten liver CT fusion images to obtain liver three-dimensional fusion CT images.

As shown in FIG. 2 , in this non-limiting embodiment, the image preprocessing apparatus 20 includes: a denoising unit 201 , which is sequentially connected in communication, for performing denoising processing on each liver CT image in the liver CT image sequence group , an enhancement unit 202 that performs contrast enhancement processing, and a separation unit 203 that performs target area separation processing.

As another non-limiting implementation, as shown in FIG. 3 , the image registration apparatus 30 includes: a resolution processing unit 301 , a rigid body transformation unit 302 , a coordinate obtaining unit 303 , a function obtaining unit 304 , a parameter obtaining unit 305 , and a reprocessing unit 303 . unit 306, and image output unit 307.

The discrimination processing unit 301 is configured to determine an arterial phase liver CT image in the arterial phase liver CT image sequence map group as the arterial phase liver CT image to be registered, and select a venous phase liver CT image sequence map group in the venous phase liver CT image sequence map group. The liver CT image is used as the venous phase liver CT image to be registered. The arterial phase liver CT image and the venous phase liver CT image to be registered are separately processed with multi-resolution to form a three-layer arterial phase liver CT image pyramid and venous phase liver CT image. In the liver CT image pyramid, the top layer of the arterial phase liver CT image pyramid and the venous phase liver CT image pyramid have the smallest resolution, and the bottom layer is the original image.

The rigid body transformation unit 302 is connected in communication with the resolution processing unit 301, and is configured to perform geometric coordinate rigid body transformation on a designated area of the top-level image of the liver CT image pyramid in the arterial phase to obtain a new area.

The coordinate obtaining unit 303 is connected in communication with the rigid body transforming unit 302, and is used for obtaining the coordinates of the top image of the venous phase liver CT image pyramid in the new region obtained by the rigid body transforming unit through the cubic B-spline interpolation method.

The function obtaining unit 304 is connected in communication with the coordinate obtaining unit 303, and is configured to calculate the similarity between the top-level image of the liver CT image pyramid in the arterial phase and the interpolation map by using the similarity measure based on the joint histogram to obtain a similarity function.

The parameter acquisition unit 305 is connected in communication with the function acquisition unit 304, and is used to input the similarity function obtained by the function acquisition unit into the particle swarm optimization algorithm for optimization calculation to obtain the optimal transformation parameters, and then start the rigid body transformation unit 302, coordinate The acquisition unit 303 and the function acquisition unit 304 until the maximum value is obtained.

The reprocessing unit 306 is connected in communication with the parameter acquisition unit 305, and is used to use the optimal transformation parameter as the starting point of the Powell algorithm search, and respectively process the last two layers of images of the arterial phase liver CT image pyramid and the venous phase liver CT image pyramid, And start the rigid body transformation unit 302 , the coordinate acquisition unit 303 , the function acquisition unit 304 , and the parameter acquisition unit 305 again.

The image output unit 307 is connected in communication with the reprocessing unit 306, and is used for outputting the registered image of the liver CT image in the venous phase under the final transformation.

In this non-limiting embodiment, the image fusion apparatus 40 includes: an image selection unit 401 , an image frequency domain unit 402 , a frequency domain fusion unit 403 , and a reconstruction transformation unit 404 .

The image selection unit 401 is configured to select an arterial phase liver CT image and a venous phase liver CT image in the same section of the arterial phase liver CT image sequence image group.

The image frequency domain unit 402 is connected in communication with the image selection unit 401, and is used to perform fast Fourier transform on the selected arterial phase liver CT image and venous phase liver CT image to obtain image frequency domain signals, and perform frequency domain frequency analysis on the image frequency domain signals respectively. Domain non-subsampling contourlet decomposition transform to obtain the transform result.

和高频融合图像

The frequency domain fusion unit 403 is connected in communication with the image frequency domain unit 402, and is used to perform an inverse fast Fourier transform on the transformation result, and take corresponding transformations on the low-frequency part ^CL and the high-frequency part ^CH respectively to obtain a low-frequency fusion image.

and high-frequency fusion images

和高频融合图像

进行快速傅里叶变换，对整个过程进行频域非下采样轮廓波重建变换，得到该截面的肝脏CT融合图像结果。The reconstruction and transformation unit 404 is connected in communication with the frequency domain fusion unit 403, and is used to fuse the low-frequency image

and high-frequency fusion images

Fast Fourier transform is performed, and the whole process is reconstructed by non-subsampling contourlet in frequency domain, and the result of liver CT fusion image of this section is obtained.

The three-dimensional reconstruction device 50 includes: a contour matching unit 501 and an MC reconstruction unit 502 .

The contour matching unit 501 is used to set the value of the isosurface, extract the target contour of the liver in at least ten liver CT fusion images, calculate the area of the liver contour, and find and match the order relationship between the liver contours in different sections.

The MC reconstruction unit 502 is connected in communication with the contour matching unit 501, and is configured to first read the liver contour in two liver CT fusion images of the at least ten liver CT fusion images, and then read in an adjacent liver CT fusion image each time. The slices, the adjacent four of the pixels on each slice and the corresponding four pixels of the next slice form a cube, the cube is called a voxel, and then the order from left to right, front to back Process all adjacent cubes in one layer in turn, identify boundary voxels, extract isosurfaces, and then continue to read the next adjacent slice of liver CT fusion image in the same way after processing one layer, until all the processing is done. The isosurface is extracted after slicing, the algorithm ends, and the three-dimensional fusion CT image of the liver is obtained.

Although the preferred embodiments of the present invention have been described in detail herein, it is to be understood that the present invention is not limited to the specific structures and steps described and illustrated in detail herein, but can be implemented by the present invention without departing from the spirit and scope of the present invention. Other modifications and variations will occur to those skilled in the art.

Claims (10)

1. A liver medical image registration and fusion method is characterized by comprising the following steps:

(1) acquiring an arterial-phase liver CT image sequence map group and a venous-phase liver CT image sequence map group for the same liver, wherein the arterial-phase liver CT image sequence map group comprises at least ten arterial-phase liver CT images which are taken from different sections for the same liver, and the venous-phase liver CT image sequence map group comprises at least thirty venous-phase liver CT images which are taken from different sections for the same liver;

(2) sequentially preprocessing the at least ten arterial-stage liver CT images and the at least thirty venous-stage liver CT images;

(3) sequentially registering each arterial-stage liver CT image in the arterial-stage liver CT image sequence map group by using a multi-resolution CT image registration method based on a joint histogram, finding venous-stage liver CT images with the same cross section as that of the venous-stage liver CT image sequence map group in the venous-stage liver CT image sequence map group, and acquiring at least ten venous-stage liver CT images registered with the at least ten arterial-stage liver CT images;

(4) sequentially carrying out frequency domain non-subsampled contourlet transform based on hardware acceleration of a graphic processor on each arterial-phase liver CT image in the arterial-phase liver CT image sequence chart group and one venous-phase liver CT image registered with the arterial-phase liver CT image to realize image fusion, and obtaining at least ten liver CT fusion images; and

(5) and performing sequence three-dimensional reconstruction on the at least ten liver CT fusion images to obtain a liver three-dimensional fusion CT image.

2. A liver medical image registration and fusion method according to claim 1, wherein the step (2) comprises sequentially: and sequentially performing a denoising processing step, a contrast enhancement processing step and a target point region separation processing step on each liver CT image in the liver CT image sequence group.

3. The liver medical image registration and fusion method of claim 1, wherein step (3) comprises:

(3-1) determining one arterial-phase liver CT image in the arterial-phase liver CT image sequence map group as an arterial-phase liver CT image to be registered, selecting one venous-phase liver CT image in the venous-phase liver CT image sequence map group as a venous-phase liver CT image to be registered, and respectively performing multi-resolution processing on the arterial-phase liver CT image to be registered and the venous-phase liver CT image to form an arterial-phase liver CT image pyramid and a venous-phase liver CT image pyramid which comprise three layers, wherein the top image resolutions of the arterial-phase liver CT image pyramid and the venous-phase liver CT image pyramid are minimum, and the bottom layer is an original image;

(3-2) carrying out rigid body transformation on the geometric coordinates of the specified region of the top layer image of the artery-stage liver CT image pyramid to obtain a new region;

(3-3) obtaining the coordinates of the new region of the top-level image of the vein-phase liver CT image pyramid obtained in the step (3-2) by a cubic B spline interpolation method;

(3-4) calculating the similarity between the top-level image of the artery-stage liver CT image pyramid and the interpolation image by using the similarity measure based on the joint histogram to obtain a similarity function;

(3-5) inputting the similarity function obtained in the step (3-4) into a particle swarm optimization algorithm for optimization calculation to obtain an optimal transformation parameter, and repeating the steps (3-2) - (3-4) until a maximum value is obtained;

(3-6) taking the optimal transformation parameters as the initial points of the Bawell algorithm search, respectively processing the last two layers of images of the artery-stage liver CT image pyramid and the vein-stage liver CT image pyramid, and repeating the steps (3-2) - (3-5); and

(3-7) outputting a registration image of the venous-phase liver CT image under final transformation; and acquiring at least ten vein-phase liver CT images which are registered with the at least ten artery-phase liver CT images after registration.

4. A liver medical image registration and fusion method according to claim 3, wherein step (4) comprises:

(4-1) selecting one arterial-stage liver CT image and a venous-stage liver CT image which is positioned at the same section in the arterial-stage liver CT image sequence map group;

(4-2) carrying out fast Fourier transform on the selected artery-phase liver CT image and vein-phase liver CT image to obtain image frequency domain signals, and respectively carrying out frequency domain non-downsampling contour wavelength division decomposition transform on the image frequency domain signals to obtain a transform result;

(4-3) performing inverse fast Fourier transform on the transform result, and respectively performing inverse fast Fourier transform on the low-frequency part C ^L And a high frequency part C ^H Obtaining low-frequency fusion image by adopting corresponding transformation

And high frequency fused images

(4-4) fusing the images at the low frequency

And high frequency fused images

Performing fast Fourier transform, and performing frequency domain non-subsampled contourlet reconstruction transformation on the whole process to obtain a liver CT fusion image result of the section; and

and (4-5) sequentially selecting other arterial-stage liver CT images and venous-stage liver CT images which are positioned at the same section in the arterial-stage liver CT image sequence map group, and repeating the steps (4-2) - (4-4) until at least ten liver CT fusion images are obtained.

5. A liver medical image registration and fusion method as claimed in claim 3, wherein the step (5) comprises:

(5-1) setting the value of the isosurface, extracting the target contour of the liver in the at least ten liver CT fusion images, calculating the area of the liver contour, and searching the sequence relation among the liver contours with different sections and matching;

(5-2) reading the liver contour in two liver CT fusion images in the at least ten liver CT fusion images, reading in a slice of an adjacent liver CT fusion image every time, wherein four adjacent pixel points in each slice and four corresponding pixel points in a next slice form a cube, the cube is called a voxel, sequentially processing all adjacent cubes in a layer from left to right and from front to back, judging boundary voxels, extracting an isosurface, then continuously reading in the slice of the next adjacent liver CT fusion image in the same way after processing the layer, extracting the isosurface until all slices are processed, and finishing the algorithm to obtain the three-dimensional liver fusion CT image.

6. A liver medical image registration and fusion system, comprising:

the image acquisition device is used for acquiring an arterial-phase liver CT image sequence map group and a venous-phase liver CT image sequence map group of the same liver, wherein the arterial-phase liver CT image sequence map group comprises at least ten arterial-phase liver CT images of the same liver, which are taken from different sections, and the venous-phase liver CT image sequence map group comprises at least thirty venous-phase liver CT images of the same liver, which are taken from different sections;

the image preprocessing device is in communication connection with the image acquisition device and is used for sequentially preprocessing the at least ten arterial-stage liver CT images and the at least thirty venous-stage liver CT images;

the image registration device is in communication connection with the preprocessing device and is used for sequentially registering each arterial-phase liver CT image in the arterial-phase liver CT image sequence map group by using a multi-resolution CT image registration method based on a joint histogram, finding venous-phase liver CT images with the same cross section as the venous-phase liver CT images in the venous-phase liver CT image sequence map group and acquiring at least ten venous-phase liver CT images registered with the at least ten arterial-phase liver CT images;

the image fusion device is in communication connection with the image registration device and is used for sequentially carrying out frequency domain non-subsampled contourlet transform based on hardware acceleration of a graphic processor on each arterial-phase liver CT image in the arterial-phase liver CT image sequence map group and one venous-phase liver CT image registered with the same to realize image fusion and obtain at least ten liver CT fusion images; and

and the three-dimensional reconstruction device is in communication connection with the image fusion device and is used for performing sequence three-dimensional reconstruction on the at least ten liver CT fusion images to obtain liver three-dimensional fusion CT images.

7. Liver medical image registration and fusion system according to claim 6, characterized in that the image pre-processing means comprises: and the denoising unit, the enhancement unit and the separation unit are sequentially connected in a communication manner and are used for denoising each liver CT image in the liver CT image sequence group, performing contrast enhancement processing and performing target region separation processing.

8. Liver medical image registration and fusion system according to claim 7, characterized in that the image registration means comprises:

the resolution processing unit is used for determining one arterial-phase liver CT image in the arterial-phase liver CT image sequence map group as an arterial-phase liver CT image to be registered, selecting one venous-phase liver CT image in the venous-phase liver CT image sequence map group as a venous-phase liver CT image to be registered, and respectively performing multi-resolution processing on the arterial-phase liver CT image to be registered and the venous-phase liver CT image to form an arterial-phase liver CT image pyramid and a venous-phase liver CT image pyramid which comprise three layers, wherein the top-layer image resolutions of the arterial-phase liver CT image pyramid and the venous-phase liver CT image pyramid are the minimum, and the bottom layer is an original image;

the rigid body transformation unit is in communication connection with the resolution processing unit and is used for performing rigid body transformation on geometric coordinates on a specified region of the top-level image of the artery-stage liver CT image pyramid to obtain a new region;

the coordinate acquisition unit is in communication connection with the rigid body transformation unit and is used for obtaining the coordinates of the top-level image of the vein-phase liver CT image pyramid in the new region acquired by the rigid body transformation unit through a cubic B-spline interpolation method;

the function acquisition unit is in communication connection with the coordinate acquisition unit and is used for calculating the similarity between the top-level image of the artery-stage liver CT image pyramid and the interpolation image by utilizing the similarity measure based on the joint histogram to obtain a similarity function;

the parameter acquisition unit is in communication connection with the function acquisition unit and is used for inputting the similarity function obtained by the function acquisition unit into a particle swarm optimization algorithm for optimization calculation to obtain an optimal transformation parameter and then starting the rigid body transformation unit, the coordinate acquisition unit and the function acquisition unit until a maximum value is obtained;

the reprocessing unit is in communication connection with the parameter acquiring unit and is used for respectively processing the last two layers of images of the arterial-stage liver CT image pyramid and the venous-stage liver CT image pyramid by taking the optimal transformation parameter as a starting point of the Bawell's algorithm search and restarting the rigid body transformation unit, the coordinate acquiring unit, the function acquiring unit and the parameter acquiring unit; and

and the image output unit is connected with the reprocessing unit in a communication mode and is used for outputting a registration image of the venous-phase liver CT image under the final transformation.

9. Liver medical image registration and fusion system according to claim 8, characterized in that the image fusion means comprises:

the image selection unit is used for selecting one artery-phase liver CT image in the artery-phase liver CT image sequence map group and a vein-phase liver CT image which is positioned in the same section with the artery-phase liver CT image;

the image frequency domain unit is in communication connection with the image selection unit and is used for performing fast Fourier transform on the selected arterial-period liver CT image and the selected venous-period liver CT image to obtain image frequency domain signals, and performing frequency domain non-downsampling contour wavelength division decomposition transform on the image frequency domain signals to obtain transform results;

a frequency domain fusion unit, which is connected with the image frequency domain unit in communication and is used for carrying out inverse fast Fourier transform on the transform result and respectively carrying out inverse fast Fourier transform on the low-frequency part C ^L And a high frequency part C ^H Obtaining low-frequency fusion image by adopting corresponding transformation

And high frequency fused images

A reconstruction transformation unit, communicatively connected to the frequency domain fusion unit, for fusing the images at the low frequency

And high frequency fused images

And performing fast Fourier transform, and performing frequency domain non-subsampled contourlet reconstruction transformation on the whole process to obtain a liver CT fusion image result of the section.

10. The liver medical image registration and fusion system of claim 8, wherein the three-dimensional reconstruction device comprises:

the contour matching unit is used for setting the value of an isosurface, extracting the target contour of the liver in the at least ten liver CT fusion images, calculating the area of the liver contour, searching the sequence relation among the liver contours with different sections and matching;

and the MC reconstruction unit is in communication connection with the contour matching unit and is used for reading the liver contour in two liver CT fusion images in the at least ten liver CT fusion images, reading slices of one adjacent liver CT fusion image every time, forming a cube by four adjacent pixel points in each slice and four pixel points of the corresponding next slice, wherein the cube is called a voxel, sequentially processing all adjacent cubes in one layer from left to right and from front to back, distinguishing boundary voxels, extracting an isosurface, continuously reading slices of the next adjacent liver CT fusion image in the same way after processing one layer, extracting the isosurface until all slices are processed, finishing the algorithm and obtaining the three-dimensional liver fusion CT image.

Images