Method for identifying cervical squamous cell carcinoma and cervical adenocarcinoma through synthetic MRI

Abstract

The invention relates to the technical field of synthetic MRI identification, in particular to a method for identifying cervical squamous cell carcinoma and cervical adenocarcinoma through synthetic MRI. The method comprises the following steps: scanning a cervical lesion area, obtaining local gray tracks of a structure weighted image, a lesion signal weighted image and a diffusion weighted image, fitting the structure weighted image and the lesion signal weighted image into a nonlinear track, reconstructing a virtual contrast MRI image, projecting the virtual image to the scanning area, and obtaining a non-linear track. The method comprises the following steps of: identifying a main direction expansion mode and a secondary direction distortion mode of a focus, constructing a diffusion limited curve by combining diffusion weighted images, determining a multi-distribution evolution curve of the focus, synchronously enhancing different weighted images according to the multi-distribution evolution curve, and generating global enhanced MRI (Magnetic Resonance Imaging) data so as to identify a cervical focus area. According to the invention, a clearer lesion identification reference is realized, the contrast ratio and the resolution ratio of the image are improved, and the sensitivity of lesion detection is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of synthetic MRI identification technology, and particularly relates to a method for identifying cervical squamous carcinoma and cervical adenocarcinoma by using synthetic MRI. BACKGROUND

[0002] Traditional imaging methods for identifying cervical squamous carcinoma and cervical adenocarcinoma often rely on a single MRI imaging technology, lack comprehensive analysis of multiple image information, and have certain limitations in lesion identification and classification. In particular, when identifying cervical squamous carcinoma and cervical adenocarcinoma, the existing technology cannot fully capture the morphology and expansion characteristics of the lesion, resulting in a risk of misjudgment and missed judgment. The insufficient contrast and resolution of the image make it difficult to accurately identify subtle lesions, affecting the accuracy and effectiveness of the judgment. The existing technology has defects in the fusion analysis of multiple weighted images, lacks real-time monitoring of the dynamic evolution of the lesion, and cannot effectively integrate the information of structural weighted images, lesion signal weighted images, and diffusion weighted images, resulting in a lack of comprehensiveness in the identification process of the lesion. In particular, in complex lesion conditions, the traditional method fails to timely reflect the main direction expansion mode and the secondary direction distortion mode of the lesion, limiting the in-depth understanding and analysis of the characteristics of the lesion. SUMMARY

[0003] Therefore, it is necessary to provide a method for identifying cervical squamous carcinoma and cervical adenocarcinoma by using synthetic MRI to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a method for identifying cervical squamous carcinoma and cervical adenocarcinoma by using synthetic MRI comprises the following steps:

[0005] Step S1: simultaneously acquiring local gray scale trajectories of structural weighted images, lesion signal weighted images, and diffusion weighted images by scanning the cervical lesion area;

[0006] Step S2: fitting the structural weighted images and the lesion signal weighted images into a nonlinear trajectory, and performing morphological inversion through the nonlinear trajectory to reconstruct a virtual control MRI image;

[0007] Step S3: projecting the virtual control MRI image to the scanned cervical area to identify the main direction expansion mode and the secondary direction distortion mode of the lesion; constructing a diffusion restriction curve by combining the local gray scale trajectory of the diffusion weighted image and the secondary direction distortion mode, and determining a multi-distribution evolution curve of the lesion through the diffusion restriction curve and the main direction expansion mode;

[0008] Step S4: synchronously enhancing the structural weighted images, the lesion signal weighted images, and the diffusion weighted images according to the main peak and the secondary peak of the multi-distribution evolution curve to obtain global enhancement MRI data;

[0009] Step S5: Based on the preset known cervical squamous carcinoma and cervical adenocarcinoma data set and the global enhancement MRI data, the cervical lesion area is identified.

[0010] The beneficial effects of the present application are: on the one hand, through the nonlinear trajectory fitting and morphological inversion of the structure weighted image and the lesion signal weighted image, the cervical lesion area is accurately reconstructed, the generated virtual contrast MRI image provides a clearer lesion identification benchmark, greatly improves the contrast and resolution of the image, effectively enhances the sensitivity of lesion detection, combined with the local gray trajectory of the diffusion weighted image, the subtle features of the lesion can be fully captured, and the identification accuracy of the lesion is improved, providing a more reliable data basis for subsequent image analysis, innovatively combining multiple image information to form a more comprehensive analysis perspective, improving the depth and breadth of image analysis, and promoting the application of MRI image in lesion identification.

[0011] On the other hand, the generation of multi-distribution evolution curve combines the synchronous enhancement of main peak and secondary peak, realizes the global enhancement of structure weighted image, lesion signal weighted image and diffusion weighted image, and the enhanced MRI data shows higher detail resolution and contrast, fully displays the morphological features of the lesion area, reduces the misjudgment risk caused by image noise, optimizes the image quality of the cervical lesion area, enhances the discrimination ability of different types of lesions, and further improves the overall efficiency, combined with the preset known cervical squamous carcinoma and cervical adenocarcinoma data set, forms a high-efficiency classification mechanism, making the identification of lesion area more accurate and reliable.

[0012] On the other hand, the identification method based on global enhancement MRI data has significantly improved in processing speed and accuracy, using advanced image processing algorithms to quickly analyze and classify cervical lesion types, reducing the need for manual intervention, enhancing the automation level, reducing the influence of human factors on the results, promoting the deep integration of imageology and artificial intelligence technology, providing new ideas and methods for future imageology research and application, and overall improving the scientificity and practicality of cervical lesion identification. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 It is a step flowchart of a method for synthesizing MRI to identify cervical squamous carcinoma and cervical adenocarcinoma.

[0014] Figure 2 It is a detailed implementation step flowchart of step S2. Figure 1

[0015] Figure 3 It is a multi-distribution evolution curve diagram.

[0016] Figure 4 It is a schematic diagram of the collected cervical lesion area.​

[0017] The objectives, functional characteristics and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0018] The technical method of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0019] In addition, the accompanying drawings are only schematic illustrations of the present application, and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated description thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities, which do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0020] It should be understood that although the terms "first", "second" and the like can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the exemplary embodiments, a first element can be referred to as a second element, and similarly a second element can be referred to as a first element. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] To achieve the above-mentioned object, please refer to Figures 1 to 4 A method for synthesizing MRI to distinguish cervical squamous cell carcinoma from cervical adenocarcinoma, comprising the following steps:

[0022] Step S1: By scanning the cervical lesion area, local gray scale trajectories of structural weighted images, lesion signal weighted images and diffusion weighted images are obtained simultaneously;

[0023] Please refer to Figure 4 In an embodiment of the present application, the cervical lesion area is collected by a full sequence of MRI scanning equipment, and structural weighted images, lesion signal weighted images and diffusion weighted images are obtained synchronously in the same coordinate system. The images are pixel by pixel gray scale sampling to record local gray scale trajectory data.

[0024] In another embodiment of the present application, a method of adopting a block window in pixel-by-pixel gray scale sampling is adopted, i.e., the cervical region is divided into a plurality of regular grid blocks, and the average gray scale trajectory of each grid block is collected respectively to reduce random noise and improve the stability of the gray scale trajectory.

[0025] Step S2: fitting the structure weighted image and the lesion signal weighted image into a nonlinear trajectory, and performing morphological inversion through the nonlinear trajectory to reconstruct a virtual control MRI image;

[0026] In an embodiment of the present application, a nonlinear fitting is performed on the corresponding pixel point gray scale data in the structure weighted image and the lesion signal weighted image to obtain a nonlinear trajectory curve, and morphological inversion is performed based on the nonlinear trajectory to reconstruct a virtual control MRI image on the basis of the original data.

[0027] It should be noted that when performing nonlinear fitting, a polynomial fitting algorithm is selected and combined with a residual threshold to screen the fitting accuracy, and points deviating too much are removed to ensure that the fitted nonlinear trajectory is smoother and can better reflect the morphological change rule of the real tissue.

[0028] Step S3: projecting the virtual control MRI image to the scanned cervical region to identify the main direction expansion mode and the secondary direction distortion mode of the lesion; constructing a diffusion restriction curve based on the local gray scale trajectory of the diffusion weighted image and the secondary direction distortion mode, and determining the multi-distribution evolution curve of the lesion through the diffusion restriction curve and the main direction expansion mode;

[0029] Please refer to Figure 3 In the embodiment of the present application, the reconstructed virtual control MRI image is projected to the cervical scanning region through synchronous coordinate indexing, and the main direction expansion mode and the secondary direction distortion mode of the lesion are identified according to the image gradient field after projection, a diffusion restriction curve is constructed based on the local gray scale trajectory of the diffusion weighted image and the secondary direction distortion mode, and the multi-distribution evolution curve of the lesion is obtained through joint analysis of the curve and the main direction expansion mode.

[0030] It should be noted that when identifying the main direction expansion mode of the lesion, a sliding window method is adopted to extract a local main direction vector, and then a global main direction is synthesized through weighted averaging to avoid deviation caused by a single local abnormality to the overall direction recognition.

[0031] Step S4: synchronously enhancing the structure weighted image, the lesion signal weighted image and the diffusion weighted image according to the main peak and the secondary peak of the multi-distribution evolution curve to obtain global enhanced MRI data;

[0032] In the embodiment, the main peak and the secondary peak of the multi-distribution evolution curve obtained in the preamble are used to synchronously enhance the corresponding gray intensity of the structure-weighted image, the lesion signal-weighted image and the diffusion-weighted image, thereby generating the globally enhanced MRI data containing the main direction and the secondary direction characteristics.

[0033] It should be noted that the intensity of the main peak part of the curve is linearly increased in the enhancement process, and the contrast stretching method is used to enhance the local response of the secondary peak part, so as to ensure that the main direction expansion mode is prominent and the secondary direction distortion mode is clear.

[0034] Step S5: Based on the preset known cervical squamous carcinoma and cervical adenocarcinoma data set and the globally enhanced MRI data, the cervical lesion region is identified.

[0035] In the embodiment, the preset known cervical squamous carcinoma and cervical adenocarcinoma image data set is used as a control sample library, the globally enhanced MRI data is input into a classification analysis module, and the cervical lesion region is identified through feature comparison and pattern matching.

[0036] It should be noted that the globally enhanced MRI data is standardized to make it consistent with the feature range of the data set before feature comparison, and then a method based on similarity measurement is used to compare the differences between the main direction expansion mode and the secondary direction distortion mode of the enhanced image and the known sample.

[0037] Preferably, step S2 comprises the following steps:

[0038] Step S21: Projecting the structure-weighted image and the lesion signal-weighted image into the same voxel coordinate system and recording the synchronous coordinate index;

[0039] Step S22: Identifying the non-linear law of signal change with space based on the synchronous coordinate index, and identifying a plurality of non-linear change points;

[0040] Step S23: Connecting all the non-linear change points to fit a non-linear trajectory; deriving the spatial deformation parameters of the cervical lesion region through the non-linear trajectory;

[0041] Step S24: Restoring the structure-weighted image and the lesion signal-weighted image to the pre-deformation state according to the spatial deformation parameters to reconstruct a virtual control MRI image.

[0042] In one implementation manner of the embodiment, the structure-weighted image and the lesion signal-weighted image are projected into the same voxel coordinate system, and the synchronous coordinate index is recorded at each voxel position. During the projection process, a three-dimensional interpolation method is used to fill in the pixel values of images with different resolutions, and rigid registration is performed on the images before projection, that is, the key anatomical structures of the two images are made to coincide through rotation and translation operations, and then voxel projection is performed.

[0043] In the embodiment, the regularity of the signal value after projection changing with the spatial position is analyzed based on the synchronous coordinate index, the non-linear change feature points existing therein are detected, and the points are marked as non-linear change points. In the detection of the non-linear change points, a sliding window scanning mode is adopted, that is, the voxel data is moved in a fixed size cubic window block by block, the first derivative and the second derivative of the signal curve in each window are calculated, and when the derivative mutation exceeds a threshold, the point is marked as a non-linear change point. Before identifying the non-linear change points, the signal curve is smoothed, and the mean filtering or Gaussian filtering method is used to make the curve more stable, and then the change point detection is performed.

[0044] In the embodiment, all the non-linear change points are connected in the order of the synchronous coordinate index, a complete non-linear trajectory is fitted, and the spatial deformation parameters of the cervical lesion region are extracted based on the trajectory. In the connection of the change points, a polynomial fitting method is adopted, and the fitting curve is calculated by the least square method. The offset and the curvature of the non-linear trajectory in each section are calculated, these indexes are taken as the quantization parameters of the deformation, and thus the spatial deformation parameter set is obtained. The parameter results of different sections are weighted and averaged, and the weighting coefficients are determined according to the density of the non-linear change points in the section. The more dense the points are, the higher the weight is.

[0045] In the embodiment, the spatial deformation parameters are used to perform inverse recovery operation on the structure weighted image and the lesion signal weighted image, so as to eliminate the offset caused by the tissue morphological change, obtain the image before deformation, and reconstruct the virtual control MRI image. A voxel-by-voxel coordinate backtracking mode is adopted, that is, the original position of each voxel is calculated according to the spatial deformation parameter thereof, and the gray value is mapped back to the original position. The gray value of the voxel not covered is filled by using the neighborhood interpolation method.

[0046] Preferably, the step S23 comprises:

[0047] The non-linear change points are read, all the non-linear change points are sorted according to the synchronous coordinate index, and the non-linear change points are connected in sequence according to the sorting result, so as to establish an initial trajectory segment set;

[0048] The continuity and smoothness of the initial trajectory segment set are corrected to obtain a complete non-linear trajectory;

[0049] The local curvature of the trajectory is calculated based on the complete non-linear trajectory, and the trajectory direction vector in the complete non-linear trajectory is identified;

[0050] The trajectory spatial offset feature is determined through the local curvature of the trajectory and the trajectory direction vector;

[0051] The trajectory spatial offset feature is converted into the spatial deformation parameter of the cervical lesion region.

[0052] In one implementation manner of the embodiment of the present application, the non-linear change points are obtained by calculating the signal intensity mutation positions, the points are sequentially connected to obtain a plurality of initial trajectory segments after being sorted in numerical order by using the synchronous coordinate index, the broken parts of the trajectory segments are corrected by using the interpolation method to ensure the continuity and smoothness of the curve, so that the complete non-linear trajectory is obtained, and finally the offset features of the trajectory space are extracted according to the local bending degree and direction change of the curve, and the offset features are converted into spatial deformation parameters.

[0053] For example, it is assumed that there are 10 non-linear change points scanned in the cervical lesion area, the points are sequentially connected to form a trajectory segment set after being sorted by using the synchronous coordinate index, the initial trajectory is broken between the 3rd point and the 4th point, the gap is filled by using the cubic spline interpolation, and finally the complete trajectory curve is obtained. The local bending degree of the curve at the 6th point is the largest, the corresponding direction vector offset is X axis + 2, Y axis - 1 and Z axis + 3, and the offset is converted into a spatial deformation parameter matrix, which can be used for subsequent virtual contrast MRI image recovery.

[0054] It should be noted that the reading process can identify the jump points exceeding the preset threshold by scanning the gray scale change in the image matrix row by row, and record the positions and indexes of the non-linear change points.

[0055] In the embodiment, the cubic spline interpolation method is used to adjust the connection part of the trajectory segment, avoid the abrupt angle, and smooth the overall trajectory, so that the trajectory is more consistent with the real deformation trend in the spatial distribution.

[0056] It should be noted that the local bending degree is estimated by calculating the tangent slope difference of adjacent points at each trajectory point, and the curvature of the region is obtained by combining the average results of a plurality of points. The greater the bending degree is, the more significant the deformation is.

[0057] In the embodiment, the direction vector set of the whole curve is continuously calculated by selecting two adjacent points on the trajectory curve and taking the vector between the two points as the local direction.

[0058] In the embodiment, the components of the offset features are set, for example, the X axis direction offset, the Y axis direction offset and the Z axis direction offset, the components are combined into a three-dimensional vector as a spatial deformation parameter for image inverse recovery.

[0059] Especially important is that the trajectory space offset features are converted into the spatial deformation parameters of the cervical lesion area, including:

[0060] calculating the local deformation descriptor based on the spatial offset features of the trajectory;

[0061] According to the spatial offset characteristics of the nonlinear trajectory, local displacement vectors, local bending intensity and direction change distribution are calculated along the sampling points of each trajectory segment;

[0062] The local displacement vectors, local bending intensity and direction change distribution are aggregated at each corresponding synchronous coordinate index to form a local deformation descriptor dataset;

[0063] According to the spatial topological relationship, the local descriptors are fused and aggregated in scale to extract the local strain trend and the primary and secondary deformation amplitude ratio;

[0064] The spatial deformation parameters of the cervical lesion region are determined through the regional level displacement field pattern and the primary and secondary deformation amplitude ratio.

[0065] In one implementation manner of the embodiment of the present application, each point in the nonlinear trajectory is taken out in sequence, and the displacement of these points in the three-dimensional coordinate system is calculated, and then the bending intensity is obtained through the included angle between three points, and the direction change angle of adjacent segments is calculated, and all these data are aggregated in a table to form a local deformation descriptor dataset, and then the descriptors are merged according to the adjacent relationship between the trajectory points, and at the same time, a multi-scale method is used for aggregation to obtain the overall strain trend, and the spatial deformation parameters of the cervical lesion are derived according to the primary and secondary direction displacement ratio values obtained by statistics.

[0066] It should be noted that the three-dimensional coordinate data of the trajectory points are imported into tools such as Excel or Matlab, the coordinate difference between adjacent points is calculated, the result is saved as a local displacement vector, and this operation is repeated for all sampling points.

[0067] In this embodiment, any three consecutive trajectory points are selected, the included angle of the vectors of the two segments before and after the three points is calculated, the angle size is taken as the bending intensity value of the point, and the operation is repeated point by point on the entire trajectory to obtain a complete bending intensity sequence.

[0068] In this embodiment, the descriptor of each point is weighted and averaged with the descriptor of the adjacent point, which can eliminate the noise of a single point, and the same operation is repeated in different scale ranges, for example, the points are averaged once in a range of 3 points, and the points are averaged once again in a range of 5 points, to obtain an aggregated result that has both local details and overall trend.

[0069] In this embodiment, the components of all local displacement vectors in the primary direction and the secondary direction are counted and averaged respectively, and finally the average value of the primary direction is divided by the average value of the secondary direction to obtain a ratio value, which is combined with the displacement field pattern of the region as a whole to finally determine the spatial deformation parameters of the cervical lesion region.

[0070] For example, assuming that the trajectory has 15 sampling points, the three-dimensional coordinates of each point are obtained by MRI data, the displacement difference between adjacent points is calculated to obtain a series of local displacement vectors, the bending strength is calculated by the three-point included angle, and the direction change angle of adjacent vectors is recorded. Then, these data are sorted into a table according to the point number in Excel to form a local deformation descriptor data set. Then, the data is weighted and averaged in the neighborhood of 3 and 5 points, respectively, to obtain a smoothed strain trend curve. Statistics show that the average displacement in the main direction is 2 mm, and the average displacement in the secondary direction is 0.8 mm. The main-to-secondary ratio is 2.5 to 1. Combined with the displacement field pattern, the spatial deformation parameters of the cervical lesion region are determined as a main direction displacement of 2 mm and a secondary direction displacement of 0.8 mm.

[0071] Preferably, the projection of the virtual control MRI image to the scanned cervical region in step S3 to identify the main direction expansion pattern and the secondary direction distortion pattern of the lesion includes:

[0072] Projecting the virtual control MRI image back to the scanned cervical region based on the synchronous coordinate index, and calculating the local structure gradient of the lesion after projection;

[0073] Estimating the main direction vector of each local window based on the local structure gradient of the lesion;

[0074] Determining the main direction expansion pattern according to the main direction vector;

[0075] Analyzing the deformation difference based on the main direction vector and the non-linear deformation in the non-linear trajectory;

[0076] Determining the secondary direction deformation based on the deformation difference;

[0077] Identifying the secondary direction distortion cluster based on the secondary direction deformation;

[0078] Determining the secondary direction distortion pattern based on the secondary direction distortion cluster.

[0079] In one implementation of an embodiment of the present application, the virtual control MRI image and the actual scanned image are loaded in computer software and aligned by synchronous coordinate index. The lesion area is extracted in the aligned image, the local pixel gray difference is calculated to obtain the local structure gradient, the direction distribution of the gradient is counted in each small window area to estimate the main direction vector, and the main direction expansion pattern is determined. The non-linear trajectory is read in the same image, the difference between the main direction and the trajectory deformation is compared to obtain the deformation difference, the secondary direction displacement change is derived from the deformation difference result to form the secondary direction deformation, and similar distortion distribution areas are clustered in the secondary direction result to obtain the secondary direction distortion cluster. The secondary direction distortion pattern is determined by analyzing the direction consistency of the clusters.

[0080] In this embodiment, the Sobel operator or Scharr operator is used in Matlab or Python to calculate the gray gradient of the lesion area in the projection image pixel by pixel, to obtain the gradient components in the horizontal and vertical directions, and to combine them into a local structure gradient map.

[0081] In this embodiment, the gradient direction of all pixels in each local window is averaged to obtain the main direction vector of the window, and the results of all windows are summarized to form the main direction distribution of the whole lesion.

[0082] It should be noted that the displacement in the secondary direction deformation is normalized, and then the K-means clustering method is used to divide similar distortion points into several categories, and each category is a secondary direction distortion cluster.

[0083] In this embodiment, the direction consistency index of each distortion cluster is calculated, for example, the average value of the direction vector in each cluster is counted, and the final secondary direction distortion mode is determined according to the distribution law between clusters.

[0084] For example, assuming that the cervical region has a 200x200 pixel matrix, the virtual reference MRI image and the actual scanning image are first aligned in the software, then the lesion area is extracted and the local structure gradient is calculated using the Sobel operator, each 16x16 window area obtains an average direction vector, the main direction extension mode is summarized, then the average difference between the main direction and the actual deformation is detected in the nonlinear trajectory, and the displacement of the secondary direction is obtained on this basis, the displacement points are divided into 3 clusters, and finally the whole is judged as "multi-cluster secondary direction distortion mode".

[0085] Especially important is that the deformation difference is analyzed based on the main direction vector and the nonlinear deformation in the nonlinear trajectory, including:

[0086] The rate of change of the direction of the main direction vector and the nonlinear deformation in the nonlinear trajectory are matched point by point, and the difference between the local curvature in the main direction and the rate of change of the gray intensity in the trajectory is calculated;

[0087] The difference is tensorized and combined to obtain a deformation tensor containing curvature, stretch rate and local offset;

[0088] The eigenvalues of the deformation tensor are decomposed to identify the first deformation feature in combination with the nonlinear trajectory;

[0089] The main direction vector is used to simulate the deformation, and the second deformation feature is identified;

[0090] The first deformation feature and the second deformation feature are compared to obtain the deformation difference.

[0091] In one implementation form of the embodiment of the present application, the method of equidistant sampling points is adopted, the direction angle of the main direction vector on each point is compared with the tangent direction angle of the nonlinear trajectory, the direction difference is calculated as the direction change rate, and the change rate of the gray intensity of the point is combined to ensure that the quantifiable bending degree and the difference of the change rate of the gray intensity can be obtained.

[0092] In the embodiment, the local bending degree is taken as the first dimension, the stretch rate is taken as the second dimension, and the local offset is taken as the third dimension by establishing a three-dimensional tensor matrix, and the three-dimensional information of each point is uniformly stored by using the matrix superposition method.

[0093] In the embodiment, the Jacobi iteration method is adopted to decompose the deformation tensor to obtain the main eigenvalue and the secondary eigenvalue, the main eigenvalue represents the main deformation trend, the secondary eigenvalue represents the local disturbance trend, and the cross verification is performed in combination with the nonlinear offset of the whole trajectory to determine the first deformation feature.

[0094] In the embodiment, the affine transformation matrix is established, the main direction vector is introduced into the main axis direction of the transformation matrix to simulate the local stretching and compression effect of the region under different loads, and then the deformation feature caused by the affine transformation is extracted as the source of the second deformation feature.

[0095] For example, when the direction angle of a local trajectory in the scanned cervical region changes by 15 degrees between the continuous sampling points, the local gray intensity change rate is 0.3, and the deformation tensor of the point is decomposed into a main eigenvalue 2.5 and a secondary eigenvalue 0.8 after tensor combination, by comparing with the second deformation feature simulated by the affine transformation, it is found that the deformation difference value of the point is 1.7, and finally it is classified as the deformation mode of significant main direction stretching accompanied by slight secondary direction offset.

[0096] Preferably, the step S3 combines the local gray trajectory of the diffusion weighted image and the secondary direction distortion mode to construct a diffusion limited curve, and determines the multi-distribution evolution curve of the lesion through the diffusion limited curve and the main direction expansion mode, including:

[0097] extracting the texture feature vector in the main direction expansion mode;

[0098] determining the main diffusion direction of the lesion according to the texture feature vector and the local gray trajectory of the diffusion weighted image;

[0099] identifying the secondary diffusion direction of the lesion in the secondary direction distortion mode;

[0100] projecting the main diffusion direction of the lesion and the secondary diffusion direction of the lesion into the same coordinate system to judge the main and secondary diffusion directions of the lesion against data;

[0101] The diffusion limited curve is constructed by using the main and secondary diffusion direction of the lesion to oppose the data;

[0102] Based on the diffusion limited curve, the evolution mode of each edge point in the lesion is predicted by using the main direction expansion mode of the lesion;

[0103] The multi-distribution evolution curve is constructed by using the evolution mode of each edge point in the lesion.

[0104] In one implementation manner of the embodiment of the present application, the method of the gray level co-occurrence matrix is adopted, the energy, contrast, entropy and other parameters are calculated in each direction window, and the parameters are combined into a vector, so that the texture feature can reflect the directionality and local difference of the lesion tissue.

[0105] In the embodiment, the main diffusion direction and the secondary diffusion direction are unified into a synchronous coordinate index system based on the cervical region scanning, and then the angle relationship and the vector difference of the main diffusion direction and the secondary diffusion direction are calculated to obtain the main and secondary diffusion direction opposing data.

[0106] It should be noted that the diffusion limited condition of the lesion in different directions is represented as a nonlinear curve by using the angle measurement and the vector difference in the main and secondary direction opposing data, the low valley area of the curve represents the most limited direction, and the high peak area represents the relatively free direction of diffusion.

[0107] It should be noted that the lesion edge is divided into equidistant sampling points, the direction fitting and intensity correction are performed on each sampling point under the guidance of the diffusion limited curve to obtain the local evolution trend of the point, and then the trends of all the sampling points are combined into the multi-distribution evolution curve of the whole lesion.

[0108] For example, it is assumed that the texture vector extracted from the main direction expansion mode of a lesion indicates that the main direction is the horizontal direction, the local gray trajectory calculation result of the diffusion weighted image is consistent with the main diffusion direction, the main diffusion direction is identified as the horizontal right extension, the secondary direction distortion mode indicates that there is a slight diffusion in the vertical direction, the two are projected into the unified coordinate system to obtain the opposing data with an angle close to 90 degrees, the diffusion limited curve constructed by using the opposing data has an "L" type trend, the curve shows that the horizontal diffusion is free and the vertical diffusion is limited, and the finally predicted edge point evolution mode shows that the lesion rapidly expands outward in the horizontal boundary and remains stable in the vertical boundary, so that the multi-distribution evolution curve accurately describes the diffusion characteristics of the lesion.

[0109] Preferably, the main diffusion direction of the lesion and the secondary diffusion direction of the lesion are projected into the same coordinate system, and the main and secondary diffusion direction opposing data of the lesion comprises:

[0110] The main diffusion direction of the lesion and the secondary diffusion direction of the lesion are projected into the same coordinate system to obtain the main diffusion direction of the lesion and the plurality of secondary diffusion directions of the lesion in the same plane.

[0111] Calculate the angle relationship between the main diffusion direction of the lesion and all secondary diffusion directions of the lesion in the same plane;

[0112] Calculate the interaction parameter of the main diffusion direction of the lesion and the secondary diffusion direction of the lesion corresponding to the angle through the angle relationship;

[0113] Determine the main and secondary diffusion direction of the lesion according to the interaction parameter.

[0114] In an implementation manner of the embodiment of the present application, the main diffusion direction data and the plurality of secondary diffusion direction data of the cervical lesion area are read, all direction data are uniformly mapped into the same spatial coordinate system, and are in the same plane to facilitate subsequent calculation, the angle between the main diffusion direction and each secondary diffusion direction is calculated one by one, and then the corresponding interaction parameter is calculated according to the size of the angle and the direction intensity, the interaction strength is judged by setting a threshold value, and thus the confrontation data of the main and secondary diffusion directions of the lesion is obtained, wherein a unified synchronous coordinate index based on the cervical region scanning is established, all direction data are standardized to the index system, and the main direction and the secondary direction are ensured to be in the same reference plane for vectorization comparison.

[0115] In the embodiment, the angle cosine is obtained by the ratio of the vector inner product and the module length, and then the angle is obtained by the inverse cosine function, so that the angle value between the main diffusion direction and each secondary diffusion direction is accurately obtained.

[0116] In the embodiment, the angle size and the direction intensity are combined, for example, when the angle is close to 90 degrees, a higher weight is given to the interaction, and when the angle is close to 0 degrees or 180 degrees, a lower weight is given, so as to form a group of parameter values capable of representing the confrontation intensity of the main and secondary diffusion directions.

[0117] It should be noted that when the confrontation data is determined according to the interaction parameter, an confrontation threshold value can be set, when the interaction parameter is greater than the threshold value, it is determined as a strong confrontation area, and when the parameter is less than the threshold value, it is determined as a weak confrontation area, so as to form the distribution result of the main and secondary diffusion direction confrontation data.

[0118] For example, it is assumed that the main diffusion direction of a lesion is the horizontal direction, and the secondary diffusion direction has two directions, one is the 45-degree direction, and the other is the vertical direction. After being uniformly mapped into the same coordinate system, the angle between the main direction and the 45-degree secondary direction is calculated to be 45 degrees, and the angle between the main direction and the vertical direction is calculated to be 90 degrees. Through the interaction parameter calculation, the confrontation intensity of the vertical direction is obviously higher than that of the 45-degree direction, and therefore the final confrontation data result shows that the strong confrontation is formed between the horizontal direction and the vertical direction, and the weak confrontation is formed between the horizontal direction and the 45-degree direction. Such result can directly reflect the main and secondary direction relationship of the lesion during diffusion.

[0119] Preferably, constructing the diffusion restriction curve by the lesion primary-secondary diffusion direction confrontation data comprises:

[0120] According to the lesion primary-secondary diffusion direction confrontation data, calculating the confrontation parameter of each lesion secondary diffusion direction to the lesion primary diffusion direction;

[0121] Identifying the primary direction diffusion restriction condition through the confrontation parameter;

[0122] Integrating all the primary direction diffusion restriction conditions and constructing the diffusion restriction curve.

[0123] In one implementation form of the embodiment, the lesion primary-secondary diffusion direction confrontation data is read, each lesion secondary diffusion direction is compared with the lesion primary diffusion direction in the same coordinate system, the confrontation parameter of each secondary diffusion direction to the primary diffusion direction is calculated in turn, the diffusion restriction degree of the primary direction under the action of the secondary diffusion direction is judged according to the calculated confrontation parameter, the restriction conditions of all the secondary diffusion directions are summarized, and the diffusion restriction curve is integrated into a continuous diffusion restriction curve through the method of smooth interpolation or curve fitting, which is used to represent the restriction distribution of the lesion primary direction under the constraint of each direction.

[0124] It should be noted that in the calculation of the confrontation parameter, the angle size and the direction intensity in the confrontation data are combined, and the restriction coefficient of each secondary diffusion direction is obtained by integrating the two according to the weight, so as to quantify the constraint influence of each direction on the primary diffusion direction.

[0125] It should be noted that when judging the primary direction diffusion restriction condition, a threshold rule is set, such as the confrontation parameter ≥ 0.7: it is determined that the diffusion of the direction is severely restricted, which indicates that the constraint of the secondary direction on the primary diffusion direction is strong; the confrontation parameter 0.4-0.7: it is determined that the diffusion of the direction is moderately restricted, which indicates that the secondary direction exists certain constraint on the primary diffusion direction; and the confrontation parameter < 0.4: it is determined that the diffusion of the direction is weakly restricted, which indicates that the constraint of the secondary direction on the primary diffusion direction is small.

[0126] In this embodiment, the restriction degree of each direction is mapped to the amplitude value on the curve, and then a continuous curve is formed through interpolation smoothing, so as to more intuitively represent the restriction distribution.

[0127] For example, if the lesion primary diffusion direction is horizontally right, the secondary diffusion directions include vertically upward, 45 degrees right upward, and horizontally left, and the corresponding calculated restriction coefficients are 0.9, 0.6 and 0.8 respectively, then according to the restriction coefficients, a curve is drawn, the amplitude of the curve in the vertical direction is the largest, which indicates that the primary diffusion direction is most severely restricted in the vertical direction, and the amplitude in the 45 degree direction is slightly lower, and the amplitude in the horizontal left direction is the smallest, so that the diffusion constraint distribution of the lesion in each direction can be intuitively reflected through the curve.

[0128] Preferably, the evolution mode of each edge point in the lesion is predicted by the main direction expansion mode based on the diffusion limited curve, which comprises:

[0129] labeling each edge point in the lesion;

[0130] projecting the global expansion vector field of the main direction expansion mode into each edge point and constructing an edge point local coordinate system;

[0131] calculating the main direction expansion intensity and expansion direction angle of each edge point based on the edge point local coordinate system and the global expansion vector field;

[0132] calculating the diffusion constraint condition of each edge point by the diffusion limited curve;

[0133] reducing the corresponding main direction expansion intensity according to the diffusion constraint condition and the expansion direction angle of each edge point, and predicting the evolution mode of each edge point in the lesion.

[0134] In one implementation of the embodiment of the present application, edge detection is performed on the lesion area, each edge point in the lesion is labeled, the global expansion vector field generated by the main direction expansion mode of the lesion is mapped to the position of each edge point, and a local coordinate system is established around each edge point. The main direction expansion intensity and the corresponding expansion direction angle of each edge point are calculated in the edge point local coordinate system combined with the global expansion vector field. The diffusion constraint condition of each edge point is calculated according to the constructed diffusion limited curve, and the main direction expansion intensity of each edge point is reduced combined with the diffusion constraint condition. The evolution mode of each edge point in the lesion is predicted according to the reduced main direction expansion intensity. For example, the expansion intensity is 0.8 and the diffusion constraint is 0.6 for an edge point in the lesion with the main direction vector pointing outward. The reduced expansion intensity is 0.32, and the evolution trend of the edge point is predicted to be locally extended.

[0135] It should be noted that the local gray scale gradient around the edge point can be extracted using high-resolution MRI images to correct the local deviation of the global expansion vector field, thereby enhancing the accuracy of the edge point evolution prediction.

[0136] In this embodiment, the calculation of the diffusion constraint condition can consider the interaction of the edge point with the secondary direction distortion mode, and the reduction amount of the edge point is adjusted by the influence coefficient of the secondary direction on the main direction expansion intensity.

[0137] It should be noted that the time sequence prediction of the edge point evolution mode can be performed combined with the MRI scan data of multiple time points, for example, the main direction expansion intensity of the same edge point is 0.32, 0.35 and 0.37 respectively in three consecutive scans, and the future evolution trend is predicted to be continuously expanded outward.

[0138] Preferably, the main direction expansion intensity of each edge point is reduced according to the diffusion constraint condition of each edge point and the expansion direction angle, and the evolution mode of each edge point in the lesion is predicted, including:

[0139] The angle deviation of the expansion direction angle of each edge point and the main diffusion direction of the lesion is calculated;

[0140] The reduced diffusion intensity is calculated based on the angle deviation, wherein the reduced intensity = original intensity × (1-constraint coefficient) × (1-deviation coefficient);

[0141] The main direction expansion intensity of each edge point is reduced according to the diffusion constraint condition of each edge point and the reduced diffusion intensity, to obtain the reduced main direction expansion intensity;

[0142] The position coordinates and expansion state of each edge point at a future time point are calculated according to the reduced main direction expansion intensity, to obtain a single point evolution track;

[0143] The evolution mode of each edge point in the lesion is predicted through the single point evolution track.

[0144] In one implementation manner of the embodiment of the present application, a direction statistical matrix is established for each edge point in a local neighborhood to capture the spatial distribution characteristics of the main direction and the secondary direction expansion, a local constraint index is generated according to the matrix, the constraint index is used to adjust the original expansion intensity of the edge point to form a modified expansion intensity field, and the modified intensity field is mapped to a future spatial position through time step iteration, so as to obtain a local evolution track of each edge point. For example, in a certain local edge region, the expansion of points with a high constraint index is weakened, and the expansion amplitude of points with a low constraint index is large, so that an observable evolution mode is formed.

[0145] In the embodiment, the local constraint index is smoothed by using a neighborhood weighted average method, so that the evolution intensity change between edge points is continuous.

[0146] It should be noted that a dynamic constraint coefficient is introduced, and the modified intensity is adjusted according to the relative position of the edge point and the adjacent distortion mode, to reflect the influence of spatial heterogeneity on evolution.

[0147] In the embodiment, the evolution tracks of each edge point are superimposed to form a local evolution curve, so as to facilitate the observation of the evolution trend of the edge point group.

[0148] It should be noted that the local evolution track is verified in combination with actual MRI scanning sequences, for example, the continuous scanning results of the same region are compared and analyzed to verify the consistency of the predicted track and the actual expansion situation, so as to further adjust the constraint index and the modified intensity parameter.

[0149] Preferably, the position coordinates and the expansion state of each edge point at a future time point are calculated according to the reduced main direction expansion intensity, and a single-point evolution track is obtained, including:

[0150] An initial position coordinate and a current expansion state are established for each edge point according to the reduced main direction expansion intensity;

[0151] A future displacement increment of each edge point is calculated based on the reduced main direction expansion intensity and the main diffusion direction vector, and a predicted position coordinate of the edge point at a next time point is updated;

[0152] The updated edge point position coordinate is combined with the corresponding reduced main direction expansion intensity to determine the expansion state of each edge point;

[0153] A single-point evolution track is determined by the expansion state of each edge point.

[0154] In an implementation manner of the embodiment of the present application, a record table containing the initial coordinates and the current expansion state of each edge point is generated, a displacement increment is calculated by using the reduced main direction expansion intensity and the local direction vector in which the edge point is located, the displacement increment is added to the initial coordinates to generate a predicted position, and the expansion state of each point is determined according to the updated coordinates and the corresponding expansion intensity, so as to form a single-point evolution track. For example, in a dense edge point area, the points with greater intensity move farther along the main diffusion direction, and the points with smaller intensity move with a smaller amplitude, thereby forming an observable local evolution mode.

[0155] In the embodiment, a neighborhood smoothing operation is added to the predicted position of each edge point, and the moving trend of the surrounding adjacent edge points is weighted and averaged.

[0156] It should be noted that a dynamic time step is introduced, and the time step advancing speed of the edge point is adjusted according to the reduced expansion intensity, so that the time step of the point with faster expansion is smaller to improve the precision, and the time step of the point with slower expansion is slightly larger to improve the efficiency.

[0157] It should be noted that a local evolution vector field can be generated at each time step in combination with the expansion state of the edge point, and is used for visualizing the expansion direction and amplitude of the edge point group.

[0158] For example, it is assumed that 100 edge points are selected in a certain local area of a cervical lesion, the initial coordinates are recorded in millimeters, and the reduced main direction expansion intensity of each point ranges from 0.2 to 0.8. The points are moved along the main diffusion direction, the predicted positions of the points at three future frames are calculated, and the single-point evolution track is obtained after the neighborhood smoothing.

[0159] Therefore, the embodiments should be regarded, at any point, as being exemplary and not limiting, the scope of the application being defined by the appended claims and not by the above description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[0160] The foregoing is considered as illustrative only of the principles of the application. Numerous modifications and changes will readily occur to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of synthesizing MRI to differentiate squamous carcinoma of cervix from adenocarcinoma of cervix, characterized in that, The method comprises the following steps: Step S1: obtaining local gray scale trajectories of structure weighted images, lesion signal weighted images and diffusion weighted images by scanning the cervical lesion area; Step S2: fitting the structure weighted images and the lesion signal weighted images into a nonlinear trajectory, and reconstructing a virtual control MRI image through morphological inversion of the nonlinear trajectory; Step S3: projecting the virtual control MRI image to the scanned cervical area to identify the main direction expansion mode and the secondary direction distortion mode of the lesion; constructing a diffusion restriction curve based on the local gray scale trajectory of the diffusion weighted image and the secondary direction distortion mode, and determining the multi-distribution evolution curve of the lesion through the diffusion restriction curve and the main direction expansion mode; Step S4: synchronously enhancing the structure weighted images, the lesion signal weighted images and the diffusion weighted images according to the main peak and the secondary peak of the multi-distribution evolution curve to obtain global enhanced MRI data; Step S5: identifying the cervical lesion area based on a preset known cervical squamous carcinoma and cervical adenocarcinoma data set and the global enhanced MRI data.

2. The synthetic MRI method of differentiating squamous carcinoma of cervix from adenocarcinoma of cervix according to claim 1, wherein, Step S2 comprises the following steps: Step S21: projecting the structure weighted images and the lesion signal weighted images into the same voxel coordinate system and recording the synchronous coordinate index; Step S22: identifying the nonlinear law of signal change with space based on the synchronous coordinate index, and identifying a plurality of nonlinear change points; Step S23: connecting all the nonlinear change points to fit into a nonlinear trajectory; deriving the spatial deformation parameters of the cervical lesion area through the nonlinear trajectory; Step S24: reversely restoring the structure weighted images and the lesion signal weighted images to the pre-deformation state according to the spatial deformation parameters to reconstruct the virtual control MRI image.

3. The synthetic MRI method of differentiating squamous carcinoma of cervix from adenocarcinoma of cervix according to claim 2, wherein, Step S23 comprises: reading the nonlinear change points, sorting all the nonlinear change points according to the synchronous coordinate index, and sequentially connecting the nonlinear change points according to the sorting results to establish an initial trajectory segment set; correcting the continuity and smoothness of the initial trajectory segment set to obtain a complete nonlinear trajectory; calculating the local bending degree of the trajectory based on the complete nonlinear trajectory; identifying the trajectory direction vector in the complete nonlinear trajectory; determining the trajectory space offset feature through the local bending degree of the trajectory and the trajectory direction vector; converting the trajectory space offset feature into the spatial deformation parameters of the cervical lesion area.

4. The synthetic MRI method of differentiating squamous carcinoma of cervix from adenocarcinoma of cervix according to claim 1, wherein, In step S3, the virtual control MRI image is projected to the scanned cervical area to identify the main direction expansion mode and the secondary direction distortion mode of the lesion, which comprises: projecting the virtual control MRI image back to the scanned cervical area based on the synchronous coordinate index, and calculating the local structure gradient of the lesion after projection; estimating the main direction vector of each local window through the local structure gradient of the lesion; judging the main direction expansion mode according to the main direction vector; analyzing the deformation difference based on the main direction vector and the nonlinear deformation in the nonlinear trajectory; determining the secondary direction deformation condition through the deformation difference condition; identifying the secondary direction distortion cluster according to the secondary direction deformation condition; judging the secondary direction distortion mode based on the secondary direction distortion cluster.

5. The synthetic MRI method of differentiating squamous carcinoma of cervix from adenocarcinoma of cervix according to claim 1, wherein, The local gray track of the diffusion weighted image and the secondary direction distortion mode are combined in step S3 to construct a diffusion limited curve, and the multi-distribution evolution curve of the lesion is determined by the diffusion limited curve and the main direction extension mode, which comprises: extracting a texture feature vector in the main direction extension mode; determining the main diffusion direction of the lesion according to the texture feature vector and the local gray track of the diffusion weighted image; identifying the secondary diffusion direction of the lesion in the secondary direction distortion mode; projecting the main diffusion direction of the lesion and the secondary diffusion direction of the lesion into the same coordinate system to determine the main and secondary diffusion direction of the lesion; constructing a diffusion limited curve by the main and secondary diffusion direction of the lesion; predicting the evolution mode of each edge point in the lesion based on the diffusion limited curve and the main direction extension mode of the lesion; constructing a multi-distribution evolution curve by the evolution mode of each edge point in the lesion.

6. The synthetic MRI method of differentiating squamous carcinoma of cervix from adenocarcinoma of cervix according to claim 5, wherein, projecting the main diffusion direction of the lesion and the secondary diffusion direction of the lesion into the same coordinate system to determine the main and secondary diffusion direction of the lesion, which comprises: projecting the main diffusion direction of the lesion and the secondary diffusion direction of the lesion into the same coordinate system to obtain the main diffusion direction of the lesion and the plurality of secondary diffusion directions of the lesion in the same plane; calculating the angle relationship between the main diffusion direction of the lesion and all the secondary diffusion directions of the lesion in the same plane; calculating the interaction parameter of the main diffusion direction of the lesion and the corresponding secondary diffusion direction of the lesion according to the angle relationship; determining the main and secondary diffusion direction of the lesion according to the interaction parameter.

7. The synthetic MRI method of differentiating squamous carcinoma of cervix from adenocarcinoma of cervix according to claim 5, wherein, constructing a diffusion limited curve by the main and secondary diffusion direction of the lesion, which comprises: calculating the resistance parameter of each secondary diffusion direction of the lesion to the main diffusion direction of the lesion according to the main and secondary diffusion direction of the lesion; identifying the diffusion limited situation of the main direction by the resistance parameter; integrating all the diffusion limited situations of the main direction and constructing a diffusion limited curve.

8. The synthetic MRI method of differentiating squamous carcinoma of cervix from adenocarcinoma of cervix according to claim 5, wherein, predicting the evolution mode of each edge point in the lesion based on the diffusion limited curve and the main direction extension mode of the lesion, which comprises: labeling each edge point in the lesion; projecting the global extension vector field of the main direction extension mode into each edge point and constructing an edge point local coordinate system; calculating the main direction extension intensity and extension direction angle of each edge point based on the edge point local coordinate system and the global extension vector field; calculating the diffusion constraint condition of each edge point by the diffusion limited curve; reducing the corresponding main direction extension intensity according to the diffusion constraint condition and the extension direction angle of each edge point, and predicting the evolution mode of each edge point in the lesion.

9. The synthetic MRI method of differentiating squamous carcinoma of the cervix from adenocarcinoma of the cervix according to claim 8, characterized in that, reducing the corresponding main direction extension intensity according to the diffusion constraint condition and the extension direction angle of each edge point, and predicting the evolution mode of each edge point in the lesion, which comprises: calculating the angle deviation of the extension direction angle of each edge point and the main diffusion direction of the lesion; calculating the reduced diffusion intensity based on the angle deviation, wherein the reduced intensity = original intensity × (1-constraint coefficient) × (1-deviation coefficient); reducing the corresponding main direction extension intensity according to the diffusion constraint condition and the reduced diffusion intensity of each edge point to obtain the reduced main direction extension intensity; calculating the position coordinates and extension state of each edge point at the future time point according to the reduced main direction extension intensity to obtain a single point evolution track; The evolution mode of each edge point in the lesion is predicted through a single-point evolution trajectory.

10. The synthetic MRI method of differentiating squamous carcinoma of the cervix from adenocarcinoma of the cervix according to claim 9, characterized in that, The position coordinates and expansion state of each edge point at a future time point are calculated according to the expansion intensity of the main direction of reduction, and a single-point evolution trajectory is obtained, including: The initial position coordinates and the current expansion state are established according to the expansion intensity of the main direction of reduction, with each edge point as a unit; The future displacement increment of each edge point is calculated based on the expansion intensity of the main direction of reduction and the main diffusion direction vector, and the predicted position coordinates of the edge point at the next time point are updated; The updated edge point position coordinates are combined with the corresponding expansion intensity of the main direction of reduction to determine the expansion state of each edge point; The single-point evolution trajectory is determined through the expansion state of each edge point.

Images