Same-layer registration method for MR cross section diagram and total tumor pathological section based on 3D printing

Abstract

The invention provides a same-layer registration method for an MR cross section diagram and a total tumor pathological section based on 3D printing. The same-layer registration method comprises the steps of obtaining an MR image, extracting an interested volume, designing a 3D printing mold matched with a focus in shape and assisting registration of a pathological section and the MR image. MR image layering thickness information is embedded into mold design, the thickness of a mold section is consistent with the thickness of an MR scanning layer, and direction information is marked in a mold to ensure spatial alignment of a pathological section and an MR image. The manufactured pathological section is accurately matched with an MR cross section diagram through a non-rigid image registration algorithm, and layer-by-layer comparative analysis of a tumor image and pathology is achieved. The whole tumor area can be covered, and registration errors caused by needle biopsy limitation, tissue deformation and resolution difference in the prior art are overcome. The method is high in applicability and can be used for image-pathology conjoint analysis of various types of tumors, the registration precision is remarkably improved, the error is controlled within 0.1 mm, and important support is provided for tumor research and accurate diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of combining medical image processing with pathological analysis, and in particular to a same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing. Background Art

[0002] Magnetic resonance (MR) imaging technology, as a non-invasive, high-resolution medical imaging technology, has been widely used in tumor diagnosis and research. MR images can provide three-dimensional structural information, especially in observing the shape, location and relationship of tumors with surrounding tissues. At the same time, pathological sections, as the "gold standard" for tumor diagnosis, can provide detailed information on the microstructure of tissue cells. However, how to achieve accurate alignment between MR images and pathological sections has always been an important technical problem in the combined application of medical imaging and pathology. The existing technology has the following deficiencies:

[0003] 1. Difference in spatial resolution:

[0004] There is an essential difference in spatial resolution between MR images and pathological sections. MR images are generated by imaging the target tissue through magnetic fields and radio frequency signals, and the resolution is usually 0.5 to 1 mm. In contrast, pathological sections are two-dimensional images observed through a microscope, and the resolution can reach the micron level. Due to the difference in spatial resolution between the two imaging methods, it is difficult to achieve accurate registration by directly comparing and analyzing MR images with pathological sections.

[0005] In addition, existing technologies attempt to improve the resolution of MR images through interpolation or resampling techniques to match the accuracy of pathological sections. However, this method is limited by the physical resolution of the original data acquisition, and even if the computational resolution is improved, it cannot make up for the missing information at the detail level of the MR image.

[0006] 2. Deformation and displacement of pathological tissue:

[0007] During the preparation of pathological sections, tissues undergo a series of operations, including sampling, fixation, dehydration, embedding, and slicing. These processes can cause mechanical deformation or volume shrinkage of the tissue. For example, during the slicing process, the tissue may be compressed, stretched, or deformed irregularly, resulting in the final slice morphology not being completely consistent with the original tissue morphology. In addition, mechanical operations during sampling can also cause deformation or displacement of tumor tissue in local locations, which further increases the difficulty of registration with MR images.

[0008] At present, although non-rigid registration technology can correct tissue deformation to a certain extent, its effect depends on the complexity and calculation accuracy of the registration algorithm. In the case of large tissue deformation or loss of boundary information, the effect of existing registration methods is usually poor.

[0009] 3. Sampling limitations of local puncture biopsy:

[0010] Traditional pathological research methods mostly rely on local puncture biopsy, that is, extracting some tissue samples from the lesion through a biopsy needle for pathological analysis. However, this method has significant sampling limitations:

[0011] Biopsy can only obtain tissue information of a certain part of the lesion, and it is difficult to fully reflect the morphology and biological characteristics of the entire lesion;

[0012] The position and angle of the biopsy needle are difficult to completely correspond to a specific layer of the MR image, resulting in inconsistency between the sampling area and the MR image area;

[0013] For heterogeneous tumors, local biopsy may miss key pathological features and fail to provide a comprehensive understanding of the entire tumor.

[0014] Although local biopsy technology is easy to operate, it has obvious shortcomings in studying the overall picture of the tumor and in the registration analysis with MR images.

[0015] 4. Discontinuity between pathological sections and MR image information:

[0016] MR imaging provides continuous three-dimensional data that can fully present the overall morphology of the tumor and its internal structure. However, pathological sections are usually two-dimensional images obtained by slicing the tissue layer by layer along a certain direction. There are spatial intervals between these two-dimensional slices, which makes the pathological sections unable to match the MR images in terms of integrity. At the same time, due to the large intervals between pathological sections, some tumor features may be ignored, further increasing the difficulty of image registration and comparative analysis.

[0017] 5. Impact of tissue heterogeneity:

[0018] Tumors usually have high tissue heterogeneity, including differences in cell composition, vascular distribution, and matrix structure. MR images reflect the macroscopic structure and functional characteristics of the tumor (such as T1, T2 signal intensity or ADC value of diffusion-weighted imaging), while pathological sections show the microscopic structure at the cellular level. This difference in macroscopic and microscopic information makes it impossible for certain tissue features in MR images to correspond in pathological sections. In addition, the same tissue area may show different characteristics in MR images and pathological sections, making the matching between the two more complicated.

[0019] 6. Deficiencies of traditional image and pathology registration technology:

[0020] In the prior art, attempts are made to match MR images and pathological sections through image registration methods (including rigid registration and non-rigid registration). However, the following problems limit the registration accuracy:

[0021] Geometric deformation: Pathological slices are two-dimensional plane images, while MR images are three-dimensional data. The registration between the two requires complex geometric transformations. Existing methods are prone to errors when dealing with complex deformations.

[0022] Differences in structural features: There is a lack of direct correspondence between the intensity features of MR images and the tissue characteristics of pathological sections, and additional multimodal data alignment algorithms are required for registration;

[0023] Low computational efficiency: High-precision non-rigid registration algorithms usually have high computational complexity and cannot meet the needs of fast analysis.

[0024] 7. Lack of pathological and imaging analysis of the whole tumor:

[0025] Current pathological analysis methods focus more on local area analysis, and few technologies can achieve layer-by-layer, comprehensive registration analysis of the entire tumor. This results in the comparison between MR images and pathological sections being usually based on indirect analysis of partial samples, and it is impossible to explore the systematic relationship between imaging and pathology from the perspective of the entire tumor.

[0026] In summary, the existing technology has many problems in the same-layer registration of MR images and pathological sections, such as spatial resolution differences, tissue deformation, discontinuity, heterogeneity, and insufficient registration technology. These problems limit the combined application of MR images and pathological sections in tumor diagnosis and research, and also put forward higher requirements for the accurate comparative analysis of the overall morphology and pathological characteristics of tumors. Therefore, a new technical method is urgently needed to achieve high-precision registration of MR images and pathological sections at the same level, while providing overall analysis capabilities of the entire tumor.

[0027] Therefore, we urgently need to design a same-layer registration method for 3D-printed MR cross-sectional images and whole tumor pathological sections to solve the above problems. Summary of the invention

[0028] The purpose of the present invention is to address the deficiencies of the prior art and provide a same-layer registration method for MR cross-sectional images and whole tumor pathological sections based on 3D printing to solve the problems raised in the background technology.

[0029] To achieve the above object, the present invention provides the following technical solutions:

[0030] The same-layer registration method of the MR cross-sectional image and the whole tumor pathological section based on 3D printing includes the following steps:

[0031] Obtain a cross-sectional magnetic resonance image of the target area, the image is obtained by a preset scanning sequence, and its resolution is defined as Among them, FOV is the field of view, and N is the number of sampling points;

[0032] The MR images are processed layer by layer through the image segmentation algorithm to extract the volume of interest in the target area, and its shape is obtained through the function Indicates that, I i Represents the binary image of each layer;

[0033] The extracted VOI is exported to STL file format, and a three-dimensional model M (x, y, z) consistent with the shape of the target area is generated by three-dimensional modeling software;

[0034] Designing a layered mold based on a three-dimensional model, wherein the slice thickness t of the mold satisfies t=L / n, where L is the longitudinal length of the VOI and n is the number of MR slices;

[0035] A mold is generated using a 3D printer, and the mold is used to achieve same-layer registration of MR image slices and pathological slices.

[0036] As a preferred technical solution of the present invention, the image segmentation algorithm is a segmentation method based on region growing, and its region growing conditions are:

[0037] |I(x,y,z)-μ|<σ

[0038] Among them, I(x,y,z) is the pixel intensity value, μ is the regional mean, and σ is the set tolerance range.

[0039] As a preferred technical solution of the present invention, the design is based on a layered mold of a three-dimensional model, and the direction marking of the mold is defined based on a three-dimensional coordinate system, including six directions: left (L), right (R), front (A), back (P), top (S) and bottom (l), and the marking basis is:

[0040]

[0041] As a preferred technical solution of the present invention, the design of the mold layered slice thickness t satisfies the principle of minimizing the registration error, and the registration error is defined as:

[0042]

[0043] Among them, z MR,i and z Path,i are the height of the i-th layer of the MR cross-sectional plane and pathological slice, respectively.

[0044] As a preferred technical solution of the present invention, the design of the mold slice thickness t is obtained by iterative optimization, and the optimization goal is to minimize the registration error E. The optimization formula is:

[0045]

[0046] As a preferred technical solution of the present invention, the 3D printing mold is made of a biocompatible material, and the linear shrinkage rate α of the material satisfies:

[0047] 0.9≤α≤1.1.

[0048] As a preferred technical solution of the present invention, the same-layer registration of MR image slices and pathological slices is realized in the mold, and the tumor tissue is fixed in the mold, and its fixed position meets the following constraints:

[0049] Δx,Δy,Δz<∈

[0050] Where Δx, Δy, Δz are the offsets of the tumor tissue in the mold, and ∈ is the allowable offset error.

[0051] As a preferred technical solution of the present invention, the final alignment of the MR cross-sectional image and the pathological slice is achieved by a non-rigid image registration algorithm. The energy function of the non-rigid registration is:

[0052]

[0053] Where T(p) is the registration transformation, p and p′ are the corresponding points on the MR image and pathological image, respectively, and λ is the regularization parameter. is the deformation field gradient.

[0054] As a preferred technical solution of the present invention, the registration process is implemented by multi-resolution optimization, and the optimization process is: estimating global registration parameters at low resolution; refining local registration parameters at high resolution; the final registration result is calculated by the following formula:

[0055]

[0056] As a preferred technical solution of the present invention, the final pathological slice and MR image registration result is verified by the following reconstruction formula:

[0057]

[0058] Among them, M Path,i (x, y) is the i-th layer pathological slice image, G(zz MR,i ) is a Gaussian distribution kernel function, which is used to simulate the slice thickness of MR images.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The present invention can accurately restore the tissue structure of the tumor by introducing a 3D printing mold, combining the layer thickness information of the MR image and the three-dimensional morphological design of the lesion. Through the directional marking and slice groove design of the mold, the pathological slices and the MR cross-sections are spatially registered layer by layer, effectively solving the registration error problem caused by tissue deformation and discontinuity in the prior art. It has been verified that the registration accuracy is significantly improved, and the error is controlled within 0.1mm, providing a reliable basis for the accurate comparative analysis of pathological slices and MR images.

[0061] Different from the traditional local puncture biopsy method, the present invention uses mold-assisted slicing technology to achieve layer-by-layer slicing analysis of the entire tumor, avoiding the loss of regional information caused by the limitations of biopsy sampling. This method can cover the entire tumor area and obtain continuous tissue information, thereby providing more complete data support for tumor heterogeneity research and imaging-pathology system comparison.

[0062] The 3D printing mold of the present invention is made of biocompatible materials, with low production cost and high precision, and can be quickly customized according to the morphology of different tumors. At the same time, through the non-rigid image registration algorithm and resolution optimization technology, the registration difficulty between pathological sections and MR images is further reduced, and it is suitable for the joint analysis of imaging and pathology of various types of tumors. This efficient, accurate and highly applicable technical solution provides an important tool for tumor diagnosis, treatment evaluation and scientific research applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a flow chart of the method for same-layer registration of MR cross-sectional images and whole tumor pathological sections based on 3D printing proposed in the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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.

[0065] The following is combined with Figure 1 With reference to the accompanying drawings and a number of embodiments, specific embodiments of the present invention are described in detail.

[0066] The present invention provides a method for same-layer registration of MR cross-sectional images and whole tumor pathological sections based on 3D printing. The following describes the embodiments of the present invention in detail to better understand and apply the present invention.

[0067] 1. Overall method flow;

[0068] MR image acquisition uses a high-resolution magnetic resonance imaging device (such as the German Bruker7T small animal magnetic resonance scanner) to obtain MR cross-sectional images of the target tumor area.

[0069] The scanning parameters are: layer thickness t = 1mm layer interval s = 0mm field of view FOV = 20mm sampling point number N = 256, according to the formula Calculate the resolution.

[0070] Image segmentation and VOI extraction The tumor volume of interest (VOl) is extracted using a segmentation algorithm based on region growing. The region growing condition |I(x,y,z)-μ|<σ is set, where μ is the mean of the target region and σ is the tolerance range. The obtained three-dimensional segmentation data V(x,y,z) represents the volume morphology of the tumor.

[0071] VOI export and three-dimensional modeling The segmented VOI is exported to STL format, and a three-dimensional model M (x, y, z) is generated using modeling software (such as 3DSlicer or Mimics).

[0072] The mold slice thickness was designed to be t=1 mm, and the inner contour of the mold was consistent with the tumor boundary. The mold surface was marked with direction marks (L, R, A, P, S, I).

[0073] Mold Printing and Verification The mold was printed using a 3D printer with an accuracy of ±0.1 mm and made of biocompatible materials (such as photosensitive resin) to ensure a linear shrinkage rate of 0.9≤α≤1.1.

[0074] Verify the mold accuracy, the deviation shall not exceed ∈=0.2mm. Pathological slice preparation and registration Fix the tumor tissue in the mold and slice it layer by layer according to the mold slicing groove. After slicing, adjust the spatial correspondence between the pathological image and the MR cross-section through the non-rigid registration algorithm, optimize the registration error E, and achieve accurate same-layer registration.

[0075] 2. Algorithm implementation;

[0076] Image registration algorithm: The energy function of the non-rigid image registration algorithm is defined as:

[0077]

[0078] Where T(p) is the registration transformation and λ=0.01 is the regularization coefficient. Multi-resolution optimization method is used: First level: with resolution R low = 0.5mm to calculate global registration;

[0079] Second level: with resolution R high =0.078mm refines the local registration.

[0080] In the design of slice thickness optimization mold, the formula t is optimized by iteration. * =argmin t E determines the optimal slice thickness, and the error E is controlled within 0.1mm.

[0081] Example 1: Mouse breast cancer transplant tumor;

[0082] Mold design and injection: model construction and image acquisition BALB / C mice were injected with breast cancer cell line 4T1 to construct a subcutaneous transplant tumor model.

[0083] When the tumor diameter reached 10 mm, the 7TMR device was used to collect T2WI sequences and obtain slice images. The VOI was exported, a three-dimensional tumor model was generated, and a mold design file with a slice thickness of t = 1 mm was used. The mold was made by 3D printing and the accuracy was verified.

[0084] Pathology Slice Preparation and Analysis Tumor slices were prepared, and the pathology slices and MR images were aligned using a non-rigid registration algorithm to verify that the slice position error was less than 0.1 mm.

[0085] Example 2: Human brain glioma tissue sample;

[0086] Preoperative MR image acquisition: 3TMR equipment was used to scan human brain glioma patients, and T1WI and T2WI sequences were collected. The slice thickness was t = 2 mm, and the field of view FOV = 240 mm.

[0087] Mold design and printing: Extract the tumor VOI, generate the mold through modeling software, mark the mold directions as L, R, A, P, S, 1, and print the mold.

[0088] Tissue fixation and slicing: The excised tumor sample was placed in the mold and tissue slices were made along the slicing groove. After the slices were registered with the MR images, the registration error E was reduced to 0.2 mm.

[0089] Example 3: Analysis of samples from patients with lung cancer after surgery;

[0090] Preoperative CT and MR combined image acquisition: Collect lung MR cross-sectional images and analyze them together with postoperative CT images. CT resolution R CT =0.5mm d The mold design combines CT and MR images, and the slice thickness is optimized through algorithms. * , generate the mold and verify it.

[0091] Slice analysis produces complete lung cancer pathology slices, which correspond layer by layer to MR images to obtain comparative analysis results of pathology and images.

[0092] Data analysis and effect comparison: Compared with traditional puncture biopsy technology, the method of the present invention has the following advantages: the registration error is reduced by 80%, from 0.5mm in traditional methods to 0.1mm; the coverage of full tumor analysis is increased to 100%, avoiding local sampling deviation; the mold manufacturing accuracy reaches ±0.1mm, meeting the precise registration requirements.

[0093] Through these embodiments and technical details, the present invention verifies its wide applicability and high efficiency in tumor research and image-pathology comparative analysis.

[0094] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in the field. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for technical personnel in the field to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A same-layer registration method for 3D-printed MR cross-sectional images and whole tumor pathological sections, characterized in that: The following steps are involved: Obtain a cross-sectional magnetic resonance image of the target area, the image is obtained by a preset scanning sequence, and its resolution is defined as Among them, FOV is the field of view, and N is the number of sampling points; The MR images are processed layer by layer through the image segmentation algorithm to extract the volume of interest in the target area, and its shape is obtained through the function Indicates that, I i Represents the binary image of each layer; The extracted VOI is exported to STL file format, and a three-dimensional model M (x, y, z) consistent with the shape of the target area is generated by three-dimensional modeling software; Designing a layered mold based on a three-dimensional model, wherein the slice thickness t of the mold satisfies t=L / n, where L is the longitudinal length of the VOI and n is the number of MR slices; A mold is generated using a 3D printer, and the mold is used to achieve same-layer registration of MR image slices and pathological slices.

2. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 1 is characterized in that: The image segmentation algorithm is a segmentation method based on region growing, and its region growing conditions are: |I(x,y,z)-μ|<σ Among them, I(x,y,z) is the pixel intensity value, μ is the regional mean, and σ is the set tolerance range.

3. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 1, characterized in that: The design is based on a layered mold of a three-dimensional model. The direction marking of the mold is defined based on a three-dimensional coordinate system, including six directions: left (L), right (R), front (A), back (P), top (S) and bottom (L). The marking basis is:

4. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 1, characterized in that: The design of the mold layered slice thickness t satisfies the principle of minimizing the registration error, and the registration error is defined as: Among them, z MR,i and z Path,i are the height of the i-th layer of the MR cross-sectional plane and pathological slice, respectively.

5. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 4 is characterized in that: The design of the mold slice thickness t is obtained by iterative optimization, and the optimization goal is to minimize the registration error E. The optimization formula is:

6. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 1, characterized in that: The 3D printing mold is made of a biocompatible material, and the linear shrinkage rate α of the material satisfies: 0.9≤α≤1.1。 7. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 1, characterized in that: The mold is used to realize the same-layer registration of MR image slices and pathological slices. The tumor tissue is fixed in the mold, and its fixed position meets the following constraints: Δx,Δy,Δz<∈ Where Δx, Δy, Δz are the offsets of the tumor tissue in the mold, and ∈ is the allowable offset error.

8. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 7, characterized in that: The final alignment between the MR cross-sectional image and the pathological slice is achieved by the non-rigid image registration algorithm. The energy function of the non-rigid registration is: Where T(p) is the registration transformation, p and p ′ are the corresponding points on the MR image and pathological image, respectively, λ is the regularization parameter, is the deformation field gradient.

9. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 8, characterized in that: The registration process is achieved through multi-resolution optimization, and the optimization process is: estimate the global registration parameters at low resolution; refine the local registration parameters at high resolution; the final registration result is calculated by the following formula:

10. The same-layer registration method of MR cross-sectional images and whole tumor pathological sections based on 3D printing according to claim 1, characterized in that: The final pathological slice and MR image registration results are verified by the following reconstruction formula: Among them, M Path,i (x, y) is the i-th layer pathological slice image, G(zz MR,i ) is a Gaussian distribution kernel function, which is used to simulate the slice thickness of MR images.

Images