Method and device for registering medical images before and after ablation

Abstract

The invention provides a medical image registration method and equipment before and after ablation. The method is applied to the technical field of medical image processing. The method comprises the following steps: acquiring a preoperative medical image I0 and a postoperative ablation medical image I1; performing coarse registration on the preoperative medical image I0 and the postoperative ablation medical image I1 to obtain a preliminary aligned medical image pair; multi-resolution registration is carried out on the preliminarily aligned medical image pair from coarse to fine through a free deformation model based on a cubic B spline, the output of the previous resolution serves as the input of the next resolution, the registration result of fine registration is obtained, a free deformation cost function serves as an optimization target, and the registration result of fine registration is obtained. And the registration process of the free deformation model based on the cubic B-spline under different resolutions is guided. According to the invention, the accuracy and efficiency of registration are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and device for registering medical images before and after ablation. Background Art

[0002] In the medical field, ablation therapy is a commonly used minimally invasive treatment method, which is widely used in the treatment of various diseases such as tumors and arrhythmias. During the ablation process, physical or chemical means are used to cause coagulative necrosis of the target tissue to achieve the treatment goal. However, the evaluation of the effect of ablation therapy and the formulation of subsequent treatment plans often rely on the accurate analysis and comparison of images before and after ablation. The registration of images before and after ablation can align the images before and after treatment in the same spatial coordinate system, so as to intuitively show the changes in the ablation area. Doctors can accurately evaluate the ablation effect by comparing the registered images, including whether the ablation range is sufficient and whether the lesions are completely inactivated. Based on the registration results of images before and after ablation, doctors can understand the patient's treatment more accurately and formulate a more scientific and reasonable follow-up treatment plan.

[0003] However, the image registration task before and after ablation also faces some challenges. First, the diversity of image sources brings certain difficulties to image registration, especially when the images before and after ablation come from different medical devices (such as CT, MRI, etc.), which have differences in imaging principles, resolution, noise level, etc. Secondly, the influence of tissue deformation increases the complexity of registration. During the ablation process, the target tissue will deform and displace, resulting in differences in morphology and position between the images before and after ablation, which requires a more complex registration algorithm to deal with. In addition, the complexity of the lesions is also a major challenge. Some lesions may show complex morphological changes before and after ablation, such as shrinkage and deformation. These changes make it more difficult to identify and locate the lesions in the image, further increasing the complexity of image registration. Although the current registration technology has made some progress in some aspects, it still has certain limitations in the face of these complex challenges. More efficient and accurate algorithms are needed to improve the evaluation accuracy and treatment effect of ablation therapy. Summary of the invention

[0004] In view of this, the present invention provides a method for pre- and post-ablation medical image registration, comprising:

[0005] Acquire a preoperative medical image I0 and a postoperative ablation medical image I1;

[0006] Performing rough registration on the pre-operative medical image I0 and the post-operative ablation medical image I1 to obtain a preliminary aligned medical image pair;

[0007] The preliminary aligned medical image pair is subjected to multi-resolution registration from coarse to fine using a free deformation model based on cubic B-splines, wherein the output of a previous resolution is used as the input of a next resolution to obtain a registration result of fine registration, wherein a free deformation cost function is used as an optimization target to guide the registration process of the free deformation model based on cubic B-splines at different resolutions.

[0008] Optionally, before roughly registering the pre-operative medical image I0 and the post-operative ablation medical image I1, the method further includes:

[0009] Selecting the center point of the pre-operative medical image I0 and the center point of the ablation area of ​​the post-operative ablation medical image I1;

[0010] Taking the center point of the preoperative medical image I0 and the center point of the ablation area of ​​the postoperative ablation medical image I1 as the sphere center, the preoperative medical image I0 and the postoperative ablation medical image I1 are cropped using a 3D spherical mask M1 with a preset radius r1 to obtain a preoperative spherical medical image and a postoperative spherical medical image.

[0011] Optionally, the pre-operative medical image I0 and the post-operative ablation medical image I1 are roughly registered to obtain a preliminary aligned medical image pair, including:

[0012] The lesion area is removed from the 3D spherical mask M2 with a preset radius of r2, and the ablation area is removed from the 3D spherical mask M2 with a preset radius of r2 to obtain a preoperative registration constraint mask and a postoperative registration constraint mask, wherein r2≤r1;

[0013] Performing rigid registration according to the preoperative medical image I0, the postoperative ablation medical image I1, the preoperative registration constraint mask, the postoperative registration constraint mask, and the center point of the preoperative medical image I0 and the center point of the ablation region of the postoperative ablation medical image I1 to obtain a rigid deformation field;

[0014] The rigid deformation field is applied to the preoperative medical image I0 to obtain a preliminary aligned medical image pair.

[0015] Optionally, the free deformation model is:

[0016]

[0017] Among them, T μ (x, y, z) represents a free deformation model based on cubic B-spline, which is used to calculate the new position point of any point (x, y, z) in the preliminary aligned medical image after deformation, μ is the deformation parameter, φ represents the control point grid space, φ i+l,j+m,k+nRepresents the control point located at (i+l,j+m,k+n) on the control point grid space, and the resolution of the control point is n x ×n y ×n z ,in, (u,ν,ω) represents the relative position of any point (x,y,z) in the control point grid space. And 0≤u,ν,ω≤1,B l represents the basis function of the lth cubic B-spline, B m represents the basis function of the mth cubic B-spline, B n represents the basis function of the nth cubic B-spline.

[0018] Optionally,

[0019] B0(u)=(1-u) 3 / 6

[0020] B1(u)=(3u 3 -6u 2 +4) / 6

[0021] B2(u)=(-3u 3 +3u 2 +3u+1) / 6

[0022] B3(u)=u 3 / 6.

[0023] Optionally, the free deformation cost function is:

[0024]

[0025] Among them, MI(T μ (I0), I1) represents the mutual information, which represents the similarity measure of the images before and after registration. Represents the rigidity coefficient, which is used to distinguish rigid tissue from non-rigid tissue. For rigid tissue, is 1, for a completely non-rigid organization Set to 0, otherwise represents the associated constraint term for the free deformation model, OC(T μ ; x) represents the orthogonal constraint term for the free deformation model, PC(T μ ; x) represents the directional constraint term for the free deformation model, λ A is the regularization weight of the associated constraint term, λ O is the regularization weight of the orthogonality constraint, λ P is the regularization weight of the direction constraint.

[0026] Optionally, the free deformation cost function guides the registration process of the free deformation model based on cubic B-spline at different resolutions by using Gaussian kernel functions with different standard deviations to perform a blur operation or a blur operation plus a downsampling operation on the preliminary aligned medical image pair, and performs registration optimization from low resolution to high resolution.

[0027] Optionally, after obtaining the registration result, the following steps are further included:

[0028] The target registration error is calculated according to the registration result.

[0029] Optionally, the target registration error is calculated as follows:

[0030]

[0031] Wherein, TRE is the distance between each corresponding target point in the preoperative medical image I0 and the postoperative ablation medical image I1 after registration, is the target point in the postoperative ablation medical image I1, is the target point in the image after registration of the preoperative medical image I0, and ‖·‖ represents the Euclidean distance.

[0032] The second aspect of the present invention provides a pre- and post-ablation medical image registration device, which includes: a processor and a memory connected to the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor performs the above-mentioned pre- and post-ablation medical image registration method.

[0033] The present invention registers medical images from different sources through a coarse registration technique, achieves preliminary alignment of preoperative medical images I0 and postoperative ablation medical images I1, and effectively reduces the large-scale differences between images. The present invention adopts a free deformation model based on cubic B-splines, which can significantly prevent excessive deformation or distortion of images when processing medical image registration. By performing coarse-to-fine multi-resolution registration on the preliminarily aligned medical image pairs, the registration range is gradually reduced and the registration accuracy is improved, thereby achieving accurate capture of subtle differences between images. Not only does it improve the accuracy of registration, but it also enhances the algorithm's adaptability to complex image changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 is a flow chart of a method for medical image registration before and after ablation in an embodiment of the present invention;

[0036] Figure 2 is the preoperative CT image I0 in the embodiment of the present invention;

[0037] Figure 3 is a postoperative CT image I1 in an embodiment of the present invention;

[0038] Figure 4 is a preoperative spherical CT image in an embodiment of the present invention;

[0039] Figure 5 is a postoperative spherical CT image in an embodiment of the present invention;

[0040] Figure 6 is a 3D annotated image of a tumor region in an embodiment of the present invention;

[0041] Figure 7 is a 3D annotation image of the ablation area in an embodiment of the present invention;

[0042] Figure 8 is a preoperative registration constraint mask in an embodiment of the present invention;

[0043] Fig. 9 is a postoperative registration constraint mask in an embodiment of the present invention;

[0044] Fig.10 This is a diagram of the registration result in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0047] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, it can also be the internal connection of two components, it can be a wireless connection, or it can be a wired connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0049] like Figure 1 As shown, an embodiment of the present invention provides a method for registration of medical images before and after ablation, which is executed by an electronic device such as a computer or a server, and specifically includes:

[0050] S1, obtaining a preoperative medical image I0 and a postoperative ablation medical image I1. The medical image is a medical image such as CT, MRI, etc., and the subsequent medical images of the present invention are described by taking CT images as an example. Figure 2 As shown, the postoperative CT image I1 is as follows Figure 3 shown.

[0051] S2, coarsely registering the preoperative medical image I0 and the postoperative ablation medical image I1 to obtain a preliminary aligned medical image pair.

[0052] S3, using the cubic B-spline based free deformation model to perform multi-resolution registration of the preliminary aligned medical image pair from coarse to fine, where the output of the previous resolution is used as the input of the next resolution to obtain the registration result of fine registration, wherein the free deformation cost function is used as the optimization target to guide the registration process of the cubic B-spline based free deformation model at different resolutions. The obtained registration results are shown in Fig.10 shown.

[0053] The registration process of the free-form deformation model based on cubic B-spline is achieved through a multi-resolution approach. On low-resolution images, the control point spacing on the corresponding B-spline control grid is larger to complete coarse deformation. On higher-resolution images, the control point spacing on the B-spline control grid becomes smaller to complete finer detail alignment.

[0054] This embodiment registers medical images from different sources through a coarse registration technique, achieves preliminary alignment of preoperative medical image I0 and postoperative ablation medical image I1, and effectively reduces the large-scale differences between images. The present invention adopts a free deformation model based on cubic B-splines, which can significantly prevent excessive deformation or distortion of images when processing medical image registration. By performing coarse-to-fine multi-resolution registration on the preliminarily aligned medical image pairs, the registration range is gradually reduced and the registration accuracy is improved, thereby achieving accurate capture of subtle differences between images. Not only does it improve the accuracy of registration, but it also enhances the algorithm's adaptability to complex image changes.

[0055] In one embodiment, before the rough registration of the preoperative medical image I0 and the postoperative ablation medical image I1 in step S2, the following steps are further included:

[0056] The center point of the preoperative medical image I0 and the center point of the ablation area of ​​the postoperative ablation medical image I1 are selected.

[0057] Taking the center point of the preoperative medical image I0 and the center point of the ablation area of ​​the postoperative ablation medical image I1 as the sphere center, the preoperative medical image I0 and the postoperative ablation medical image I1 are cropped using a 3D spherical mask M1 with a preset radius r1 to obtain a preoperative spherical medical image and a postoperative spherical medical image.

[0058] Specifically, in the process of registering the preoperative CT image I0 and the postoperative CT image I1 (rigid registration / non-rigid registration), only the anatomical structure near the ROI area is of interest. The ROI refers to the tumor area of ​​the preoperative CT image and the ablation area of ​​the postoperative CT image. This will ensure that subsequent operational tasks (such as prediction of the ablation area, whether it is completely ablated, and evaluation of the efficacy of ablation) will not be affected by distant anatomical structure features. For example, the ablation area of ​​the left lung will not be directly affected by the right lung blood vessels. Therefore, the present invention crops the CT image before performing the coarse registration operation. Specifically, the tumor area of ​​the preoperative CT image or the center of the ablation area of ​​the postoperative CT image is used as the center of the sphere, and the preoperative CT image and the postoperative CT image are cropped using a 3D spherical mask M1 with a selected radius r1 (for example, a radius of 70 mm). The cropped preoperative spherical CT image is as follows: Figure 4 As shown in the figure, the cropped postoperative spherical CT image is as follows Figure 5 This eliminates the influence of the anatomical structure far away from the ROI on the registration and other operation tasks, and the cropped image will become smaller, which can speed up the registration process and improve the registration efficiency.

[0059] In one embodiment, in step S2, the preoperative medical image I0 and the postoperative ablation medical image I1 are roughly registered to obtain a preliminary aligned medical image pair, which specifically includes:

[0060] S21, remove the lesion area from the 3D spherical mask M2 with a preset radius r2, and remove the ablation area from the 3D spherical mask M2 with a preset radius r2 to obtain a preoperative registration constraint mask and a postoperative registration constraint mask, wherein r2≤r1.

[0061] Specifically, two constraints are added during the coarse registration process.

[0062] The first constraint: the center point of the tumor area and the center point of the ablation area.

[0063] The second constraint: In order to limit the registration process to be driven only by the intensity in a smaller spherical area (e.g., radius r2 is 30-50 mm) centered on the tumor or ablation area, the present invention adopts a registration mask strategy, which can ensure that the anatomical structure around the tumor or ablation area has the greatest impact on the positioning, thereby increasing the accuracy of the positioning of the tumor or ablation area. Specifically, the tumor and ablation area are removed from the cropped preoperative and postoperative spherical masks (the 3D annotation of the tumor area is as follows Figure 6 As shown, the 3D annotation of the ablation area is as follows Figure 7 The preoperative registration constraint mask after removal is shown as Figure 8 As shown, the postoperative registration constraint mask after removal is Fig. 9 Removing the tumor and ablation area avoids the need for a conversion operation that attempts to change (eg, expand or shrink) the size and shape of the tumor to match the ablation area after surgery, thereby ensuring the accuracy of the registration process.

[0064] S22, performing rigid registration according to the preoperative medical image I0, the postoperative ablation medical image I1, the preoperative registration constraint mask, the postoperative registration constraint mask, and the center point of the preoperative medical image I0 and the center point of the ablation area of ​​the postoperative ablation medical image I1 to obtain a rigid deformation field.

[0065] S23, applying the rigid deformation field to the preoperative medical image I0 to obtain a preliminary aligned medical image pair.

[0066] This embodiment uses a simple rigid transformation to align the preoperative medical image and the postoperative medical image, ensuring that the medical images are globally aligned first and providing a good initialization for the subsequent non-rigid registration.

[0067] Specifically, the rigid registration (coarse registration) inputs the preoperative spherical medical image ( Figure 4 ) and postoperative spherical medical images ( Figure 5 ), plus the pre-operative registration constraint mask ( Figure 8 ), postoperative registration constraint mask ( Fig. 9 ) and two corresponding center point constraints (tumor center point and ablation area center point), a rigid deformation field is obtained. This deformation field acts on the preoperative spherical medical image ( Figure 4) On it, a result is obtained, for example, denoted as Figure middle.

[0068] In addition to rigid registration, affine registration can also be performed. The rigid deformation field obtained by rigid registration has 6 parameters, which respectively describe rotation and translation, specifically referring to the translation amounts along the x-axis, y-axis, and z-axis and the rotation angles along these three axes respectively. Affine registration is a combination of translation, rotation, scaling, shearing, and reflection operations and any order of them. Multiple transformations can be combined into a matrix by matrix multiplication. For example, the following formula can represent various operations mentioned above and any order combination of various operations, etc.:

[0069]

[0070] Among them, T global (x, y, z) represents the new coordinates of the point (x, y, z) after affine transformation. are translation parameters, are the combined parameters.

[0071] In one embodiment, due to the large differences in the local characteristics of the ablation regions of different patients or different tissue organs, the above rigid registration can only capture global information and cannot accurately register medical images with local differences. Therefore, the present invention uses a free-form deformation model for non-rigid registration. By manipulating a three-dimensional parallelepiped grid containing the object, the shape of the object is changed. The manipulated grid determines a deformation function, which assigns a new position to each point on the object. Since B-splines have local control, the present invention adopts the cubic B-spline tensor product as the deformation function of FFD. This characteristic enables local manipulation of the grid when a point on the flat plate moves to a specified position. Therefore, even for a large number of control points, it can be effectively calculated. To define the free-form deformation model based on cubic B-splines, the present invention describes the domain of the medical image as Ω = {(x, y, z)|0 ≤ x < X, 0 ≤ y < Y, 0 ≤ z < Z +, the control point grid space is defined as Φ, and its size is n x ×n y ×n z , the grid spacing on each axis is δ, and φ i,j,k is the control point located at the position (i, j, k) on the grid. FFD can be denoted as:

[0072]

[0073] Among them, T μ (x, y, z) represents the free-form deformation model based on cubic B-splines, which is used to calculate the new position point of any point (x, y, z) in the preliminarily aligned medical image after deformation. μ is the deformation parameter, φ represents the control point grid space, and φ i+l,j+m,k+nRepresents the control point located at (i+l,j+m,k+n) on the control point grid space, and the resolution of the control point is n x ×n y ×n z ,in, (u,ν,ω) represents the relative position of any point (x,y,z) in the control point grid space. And 0≤u,ν,ω≤1,B l represents the basis function of the lth cubic B-spline, B m represents the basis function of the mth cubic B-spline, B n represents the basis function of the nth cubic B-spline. These three basis functions measure the contribution of each control point to the transformation T.

[0074] Specifically, for B l For (u), its value can be as follows:

[0075] B0(u)=(1-u) 3 / 6

[0076] B1(u)=(3u 3 -6u 2 +4) / 6

[0077] B2(u)=(-3u 3 +3u 2 +3u+1) / 6

[0078] B3(u)=u 3 / 6.

[0079] Among them, B m (ν), B n The value of (ω) and B l The method of (u) is the same and will not be repeated here.

[0080] Furthermore, B-spline is locally controlled, which makes it efficient to calculate even with a large number of control points. According to the above free deformation model, the basis function of B-spline is controlled within the neighborhood, that is, changing the control point φ i,j,kIt only affects the transformation within the local neighborhood of the control point. The parameters of the B-spline-based free deformation model are the control points in the grid, and the degree of non-rigid deformation that can be modeled depends mainly on the resolution of the control point grid. Large-spacing control points allow modeling of global non-rigid deformations, while small-spacing control points allow modeling of local non-rigid transformations. In step S3, a free deformation model based on cubic B-splines is used to perform multi-resolution registration from coarse to fine on the preliminary aligned medical images. The present invention adopts a multi-resolution method, in which the resolution of the control grid gradually increases from coarse to fine with the resolution of the medical image. A compromise between the degree of non-rigid transformation of ablation data and the computational cost can be achieved.

[0081] Specifically, the non-rigid deformation process is to transform the middle and Figure 5 For non-rigid registration, multi-resolution is used to register the middle and Figure 5 The registration is carried out from low resolution to high resolution, and from coarse to fine.

[0082] In one embodiment, since the free deformation model based on B-spline cannot guarantee the preservation of the topological structure of the deformation tissue, it is necessary to add a constraint term to constrain the deformation of the B-spline. If the deformation of the region of interest is to be a rigid transformation, the following three conditions must be met, and if these three conditions are deviated, a penalty will be received.

[0083] The first condition: association constraint. Since the rigid transformation is definitely an affine transformation (for example, parallel lines remain parallel after deformation), the second-order derivative of the affine transformation with respect to the spatial position is 0. The deformation deviation of the affine transformation can be expressed by the bending energy:

[0084]

[0085] The double dots represent the second-order partial derivatives, ‖·‖ F is the Frobenius norm. x∈Ω M ,Ω M represents a floating image region. The subscript μ denotes the parameter of the transformation T. For an affine transformation, the expression evaluates to 0.

[0086] The second condition: orthogonality constraint, due to the non-rigid transformation T in a small neighborhood of the image voxel point x μ It can be approximated by the transformation T μ The Jacobian matrix of is orthogonal, so the orthogonality can be used to measure the deformation deviation of the rigid mapping:

[0087]

[0088] Among them, a single point represents the first-order derivative, I is the identity matrix, and if T μ Indicates rotation or reflection or Then the above terms evaluate to 0. Conceptually, OC penalizes changes in volume (e.g. scaling) and shape. However, if If the surface is reflective, the topology may change (such as folding, tearing, or the appearance or disappearance of structures), and the OC cannot detect these changes.

[0089] The third condition: direction constraint. Since the spatial Jacobian matrix of the rigid mapping is a special orthogonal matrix (i.e., determinant = 1), the deformation containing reflection and / or scaling is penalized and can be expressed by the following formula:

[0090]

[0091] where det(·) represents the determinant of the matrix. For rigid transformations and deformations involving simple and / or pure shear (volume change is 0), the expression evaluates to 0, i.e., when The volume will not change if Indicates that the local deformation is expansion. If indicates that the deformation is locally compressive. Therefore, this constraint is equivalent to maintaining volume and / or incompressibility.

[0092] Taking the rigid constraints of the above three conditions as regular terms, the free deformation cost function is:

[0093]

[0094] Among them, MI(T μ (I0), I1) represents the mutual information, which represents the similarity measure of the images before and after registration. Represents the rigidity coefficient, which is used to distinguish rigid tissue from non-rigid tissue. For rigid tissue, is 1, for a completely non-rigid organization Set to 0, otherwise represents the associated constraint term for the free deformation model, OC(T μ ; x) represents the orthogonal constraint term for the free deformation model, PC(T μ ; x) represents the directional constraint term for the free deformation model, λ A is the regularization weight of the associated constraint term, λ O is the regularization weight of the orthogonality constraint, λ P is the regularization weight of the direction constraint.

[0095] In general, the Frobenius norm of matrix A is According to the chain rule, the gradient of AC with respect to μ can be expressed as:

[0096]

[0097] Among them, dim represents the spatial dimension, i, j, k are T μ At point x it corresponds to The indices of the Hessian matrix.

[0098] Similarly, the gradient of OC with respect to μ can be expressed as:

[0099]

[0100] The gradient of PC with respect to μ can be expressed as:

[0101]

[0102] Where tr(·) represents the trace of the matrix, and adj(·) represents the adjoint matrix of the matrix.

[0103] In order to make the registration of the region of interest more accurate, in addition to the association constraints often used in existing registration tasks, the present invention also adds additional orthogonal constraints and directional constraints to ensure that the deformation of the tumor region is as close to the rigid deformation as possible, that is, only the transformation (that is, rotation or translation transformation) that does not change the size, aspect ratio and other external shapes of the tumor region is generated as much as possible, that is, after registration, while ensuring that the tumor is basically not distorted, it is aligned with the fixed image as much as possible. The region other than the tumor region can produce any non-rigid deformation, such as scaling, shearing and other transformations.

[0104] Furthermore, the free deformation cost function in step S3 guides the registration process of the free deformation model based on cubic B-spline at different resolutions by using Gaussian kernel functions with different standard deviations to perform a blur operation or a blur operation plus a downsampling operation on the preliminary aligned medical image pairs, and performs registration optimization from low resolution to high resolution. The cost function is the objective function for optimizing the free deformation model based on cubic B-spline.

[0105] Specifically, there are two multi-resolution strategies for free deformation cost function optimization. The first is to use a Gaussian pyramid without downsampling and only perform Gaussian blurring. The second is to use a downsampling Gaussian pyramid, which is to perform another downsampling operation after Gaussian blurring. The same image (the original preoperative medical image I0 and the postoperative ablation medical image I1) is blurred using a Gaussian kernel function. The standard deviation is an important parameter of the Gaussian kernel function, which determines the degree of blurring of the image. The larger the standard deviation, the more obvious the blurring effect. For the multi-resolution strategy implemented by downsampling, the number of pixels in each dimension is halved at each resolution, but the pixel size is doubled, so the size of the image is physically the same.

[0106] For example, the present invention chooses to use 4 resolutions (the number of resolutions can be set according to the task), and sequentially uses Gaussian kernel functions with standard deviations of 0.5, 1, 2, and 4 to convolve the original preoperative medical image I0 and the postoperative ablation medical image I1 (the downsampling operation can be selected according to the situation). The registration process is performed sequentially from low resolution to high resolution. At lower image resolutions, a coarse B-spline control point grid is used, and the control point spacing is larger, which only allows coarse deformation and helps to match larger global structures; at higher resolutions, a finer grid is used, and the spacing between control points is smaller, which allows local deformation and helps to match smaller local structures in the image. At the optimal resolution, for example, the control point grid spacing starts from 8 mm and increases to 11.28, 15.84, and 22.4 mm at lower resolutions.

[0107] In one embodiment, after obtaining the registration result, step S3 further includes:

[0108] The target registration error is calculated based on the registration results.

[0109] This embodiment calculates the target registration error to evaluate the geometric accuracy of the registration method, thereby achieving an intuitive evaluation of the registration accuracy of preoperative and postoperative ablation medical images. The smaller the error, the more accurate the registration, which is crucial for subsequent medical diagnosis and treatment. At the same time, the feedback of the target registration error can be used as a basis for optimizing the registration algorithm. By analyzing the source and distribution of the error, the registration strategy can be adjusted in a targeted manner to improve the efficiency and accuracy of the registration algorithm. In addition, accurate registration results help doctors more quickly identify the coverage of the ablation area on the tumor, assist doctors in formulating more accurate treatment plans, and improve patients' treatment effects and quality of life.

[0110] Furthermore, the target registration error is calculated as follows:

[0111]

[0112] Wherein, TRE is the distance between each corresponding target point in the preoperative medical image I0 and the postoperative ablation medical image I1 after registration, is the target point in the postoperative ablation medical image I1, is the target point in the image after registration of the preoperative medical image I0, and ‖·‖ represents the Euclidean distance.

[0113] The target registration error is calculated as Fig.10 Although the registration effect of the boundary part of the cropped area is not good, the registration effect of the lesion area is accurate. The present invention only cares about the lesion and its surrounding area of ​​interest.

[0114] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0115] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0116] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0117] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0118] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A method for pre- and post-ablation medical image registration, characterized in that: include: Acquire a preoperative medical image I0 and a postoperative ablation medical image I1; Performing rough registration on the pre-operative medical image I0 and the post-operative ablation medical image I1 to obtain a preliminary aligned medical image pair; The preliminary aligned medical image pair is subjected to multi-resolution registration from coarse to fine using a free deformation model based on cubic B-splines, wherein the output of a previous resolution is used as the input of a next resolution to obtain a registration result of fine registration, wherein a free deformation cost function is used as an optimization target to guide the registration process of the free deformation model based on cubic B-splines at different resolutions.

2. The method according to claim 1, characterized in that Before roughly registering the pre-operative medical image I0 and the post-operative ablation medical image I1, the method further includes: Selecting the center point of the pre-operative medical image I0 and the center point of the ablation area of ​​the post-operative ablation medical image I1; Taking the center point of the preoperative medical image I0 and the center point of the ablation area of ​​the postoperative ablation medical image I1 as the sphere center, the preoperative medical image I0 and the postoperative ablation medical image I1 are cropped using a 3D spherical mask M1 with a preset radius r1 to obtain a preoperative spherical medical image and a postoperative spherical medical image.

3. The method according to claim 2, characterized in that The preoperative medical image I0 and the postoperative ablation medical image I1 are roughly registered to obtain a preliminary aligned medical image pair, including: The lesion area is removed from the 3D spherical mask M2 with a preset radius of r2, and the ablation area is removed from the 3D spherical mask M2 with a preset radius of r2 to obtain a preoperative registration constraint mask and a postoperative registration constraint mask, wherein r2≤r1; Performing rigid registration according to the preoperative medical image I0, the postoperative ablation medical image I1, the preoperative registration constraint mask, the postoperative registration constraint mask, and the center point of the preoperative medical image I0 and the center point of the ablation region of the postoperative ablation medical image I1 to obtain a rigid deformation field; The rigid deformation field is applied to the preoperative medical image I0 to obtain a preliminary aligned medical image pair.

4. The method according to claim 1, characterized in that The free deformation model is: Among them, T μ (x, y, z) represents a free deformation model based on cubic B-spline, which is used to calculate the new position point of any point (x, y, z) in the preliminary aligned medical image after deformation, μ is the deformation parameter, φ represents the control point grid space, φ i+l,j+m,k+n Represents the control point located at (i+l,j+m,k+n) on the control point grid space, and the resolution of the control point is n x ×n y ×n z ,in, (u,ν,ω) represents the relative position of any point (x,y,z) in the control point grid space. And 0≤u,ν,ω≤1,B l represents the basis function of the lth cubic B-spline, B m represents the basis function of the mth cubic B-spline, B n represents the basis function of the nth cubic B-spline.

5. The method according to claim 4, characterized in that: B0(u)=(1-u) 3 / 6 B1(u)=(3u 3 -6u 2 +4) / 6 <h2 style=";text-align:left;direction:ltr">B2(u) = (-3u<h2 style=";text-align:left;direction:ltr"> 3 <h2 style=";text-align:left;direction:ltr"> +3u<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> +3u+1) / 6 B3(u)=u 3 / 6.

6. The method according to claim 1, characterized in that The free deformation cost function is: Among them, MI(T μ (I0), I1) represents the mutual information, which represents the similarity measure of the images before and after registration, c(x|T μ ) represents the rigidity coefficient, which is used to distinguish rigid tissue from non-rigid tissue. For rigid tissue, c(x|T μ ) is 1, and for a completely non-rigid tissue, c(x|T μ ) is set to 0, otherwise 0 <c(x|T μ )<1, AC(T μ ; x) represents the associated constraint item of the free deformation model, OC(T μ ; x) represents the orthogonal constraint term for the free deformation model, PC(T μ ; x) represents the directional constraint term for the free deformation model, λ A is the regularization weight of the associated constraint term, λ O is the regularization weight of the orthogonality constraint, λ P is the regularization weight of the direction constraint.

7. The method according to claim 1, characterized in that The free deformation cost function guides the registration process of the free deformation model based on cubic B-spline at different resolutions by using Gaussian kernel functions with different standard deviations to perform a blur operation or a blur operation plus a downsampling operation on the preliminary aligned medical image pair, and performs registration optimization from low resolution to high resolution.

8. The method according to claim 1, characterized in that After obtaining the registration result, it also includes: The target registration error is calculated according to the registration result.

9. The method according to claim 8, characterized in that The target registration error is calculated as follows: Wherein, TRE is the distance between each corresponding target point in the preoperative medical image I0 and the postoperative ablation medical image I1 after registration, is the target point in the postoperative ablation medical image I1, is the target point in the image after registration of the preoperative medical image I0, and ‖·‖ represents the Euclidean distance.

10. A pre- and post-ablation medical image registration device, characterized in that: include: A processor and a memory connected to the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor performs the pre- and post-ablation medical image registration method as described in any one of claims 1-9.

Images