Female pelvic floor multi-target curve medical image collaborative registration and alignment method

Abstract

The invention discloses a female pelvic floor multi-target curve medical image collaborative registration and alignment method, which comprises the steps of S1, marking a target curve, an SCIPP line and a segmentation mark point in each piece of sample data; s2, performing dimension reduction on the three-dimensional medical data to two-dimensional medical data by adopting a dimension reduction method combining Y-Z plane projection with X-coordinate dynamic compensation; s3, the curve is smoothed through cubic spline interpolation before registration, and the influence of original data noise on registration precision is avoided; s4, calculating a translation matrix based on the reference point symphysis pubis inferior margin of the SCIPP line, and moving the first reference point symphysis pubis inferior margin to the original point; s5, taking the SCIPP line as a reference vector, and calculating a rotation matrix by using a difference value between the reference vector and the target angle; s6, applying corresponding translation and rotation matrixes to the corresponding sample curve data, and establishing a unified coordinate system; s7, calculating average values of the curves in a segmented manner according to the mark points, and combining the average values calculated in the segmented manner into an average value of the whole curve; and S8, calculating the center of mass, the local center of mass and the global center of mass of each curve, and translating the center of mass of the corresponding curve to coincide with the local center of mass and the global center of mass of the corresponding curve to realize alignment. According to the method, standardized registration, two-dimensional alignment and dissection segmentation average analysis of female pelvic floor organ profile curve data are realized, and an accurate quantitative basis is provided for urinary surgery disease diagnosis and clinical research.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of medical image data processing, and particularly relates to a female pelvic floor multi-target curve medical image collaborative registration and alignment method, which is particularly suitable for synchronous morphological comparative analysis of multiple associated curves under multiple patients and multiple physiological states. BACKGROUND

[0002] In medical image analysis, the morphological changes of female pelvic floor organs are important basis for evaluating physiological function or disease state. The morphology of such organs is usually described by multiple associated curves, such as the contour curves of bladder, urethra and rectum. Clinically, the evaluation of pelvic floor organ morphology mainly relies on the naked eye observation and subjective comparison of doctors on MRI or CT images, lacking quantitative and standardized methods.

[0003] In the prior art, the curve registration of pelvic floor organ morphology has the following main problems: Single curve independent registration: multiple associated curves are registered separately, without considering the inherent relative spatial relationship (such as distance, angle) between curves, which may lead to distortion of overall morphological analysis; Single registration dimension: only translation or rotation is used to eliminate part of the difference, without combining centroid alignment to optimize overall morphological consistency; Complex three-dimensional data processing: the contours of pelvic floor organs are usually three-dimensional data, which are directly analyzed and easily affected by noise, acquisition error and posture difference, lacking effective dimension reduction and standardization processing methods; Insufficient anatomical segmentation analysis: existing methods usually only focus on overall contour or single index, lacking segmented analysis based on anatomical landmarks, which cannot finely quantify different regions of organs, limiting the accuracy of disease diagnosis and function evaluation.

[0004] Therefore, a collaborative registration method capable of synchronously processing multiple associated curves is needed, which realizes the standardization of each curve in a unified coordinate system while preserving the inherent relative spatial relationship between curves, and combines two-dimensional dimension reduction, centroid alignment and segmented average analysis, to provide reliable data for quantitative analysis of female pelvic floor organ multi-structure morphology, clinical diagnosis and scientific research. SUMMARY

[0005] The present application aims to overcome the defects of relative relationship distortion, single registration dimension and lack of collaborative verification caused by multiple curve independent registration in the prior art, and provides a collaborative method combining rigid registration and centroid alignment to realize high-precision synchronous standardization of multiple associated curves under multiple patients and multiple physiological states.

[0006] To solve the above technical problems, the present application provides a female pelvic floor multi-target curve medical image collaborative registration and alignment method, comprising the following steps: S1: In each sample data, mark the target curve, SCIPP line and segment marker; S2: Use Y-Z plane projection combined with X coordinate dynamic compensation to reduce the dimension of three-dimensional medical data to two-dimensional; S3: Before registration, smooth the curve by cubic spline interpolation to avoid the influence of original data noise on registration accuracy; S4: Calculate the translation matrix based on the reference point of the SCIPP line and the lower edge of the pubic symphysis, and move the first reference point, the lower edge of the pubic symphysis, to the origin; S5: Take the SCIPP line as the reference vector, and calculate the rotation matrix using the difference between the reference vector and the target angle; S6: Apply the corresponding translation and rotation matrix to the corresponding sample curve data to establish a unified coordinate system; S7: Calculate the average value of the curve according to the marker points, and then combine the average values of the segmented calculation into the average value of the whole curve; S8: Calculate the self-centroid, local centroid and global centroid of each curve, and translate the self-centroid of the corresponding curve to coincide with the local and global centroids of the corresponding curve, respectively, to achieve alignment.

[0007] Preferably, the specific steps of S1 in each sample data are: Step 1: In three-dimensional medical images (such as MRI, CT), manually or automatically label two types of data, including target curve data and SCIPP line data, and place marker points at key positions in the target curve data, and each curve can have multiple marker points; Step 2: After labeling, save according to the format, name the target curve as "sample ID - curve name - state", and name the SCIPP line file as "sample ID-SCIPP - state", and ensure that the curves and corresponding SCIPP lines of the same sample and physiological state (such as resting state) are associated.

[0008] Preferably, the specific steps of S2 using Y-Z plane projection combined with X coordinate dynamic compensation to reduce the dimension of three-dimensional medical data to two-dimensional are: Step 1: Extract the three-dimensional coordinates of all target curves and SCIPP lines; Step 2: Analyze the X coordinate variation, and get the variation ratio by comparing the X coordinate standard deviation with the Y coordinate standard deviation, and calculate the average variation ratio of all curves to evaluate the influence of X on the shape; Step 3: Determine the value of compensation factor k according to the calculated average variation ratio; Step 4: Convert the three-dimensional coordinates of the target curve and the SCIPP line to two-dimensional coordinates.

[0009] Preferably, the specific steps for avoiding the influence of original data noise on the registration accuracy by smoothing the curve before S3 registration through cubic spline interpolation are as follows: Step 1: Sort the output two-dimensional discrete point coordinates according to the point index to ensure the continuity of the curve; Step 2: Calculate the distance between adjacent points for the sorted point set, and accumulate the chord length L = [0, d_1, d_1+d_2,..., Σd_i] to be normalized to [0,1] as the interpolation parameter t; Step 3: Construct a cubic spline function for Y and Z coordinates to generate 300 evenly distributed interpolation points, and then calculate the smoothed curve coordinates, and output the smoothed two-dimensional curve Preferably, the specific steps for calculating the translation matrix based on the reference point of the SCIPP line lower pubic symphysis to move the first reference point lower pubic symphysis to the origin are as follows: Step 1: Extract the coordinates of the first reference point P1 (lower pubic symphysis) from the smoothed SCIPP line two-dimensional data; Step 2: Construct a translation matrix T to move P1 to the origin (0,0) of the coordinate system; Step 3: Bring the coordinates of the first reference point P1 (lower pubic symphysis) of each sample and each state into the translation matrix T respectively to obtain the corresponding translation matrix.

[0010] Preferably, the specific steps for using the SCIPP line as a reference vector and calculating the rotation matrix using the difference between the reference vector and the target angle are as follows: Step 1: Extract the coordinates of the second reference point P2 (such as the end of the pubic ramus) from the SCIPP line two-dimensional data; Step 2: Calculate the SCIPP reference vector a, and use this vector as the initial vector Step 3: Calculate the angle α between the initial vector and the positive direction of the x-axis: Step 4: Calculate the rotation angle θ and the rotation matrix R: Rotate the vector clockwise to the horizontal position as the Y-axis of the coordinate system, which generally needs to be rotated by 32°~36°, and the corresponding angle for clockwise rotation is negative according to the actual sample effect.

[0011] Preferably, the specific steps for applying the corresponding translation and rotation matrices to the corresponding sample curve data to establish a unified coordinate system are as follows: Step 1: First, calculate the combined transformation matrix M, M = T × R; Step 2: Perform coordinate conversion by multiplying each target curve data of the same sample and the same physiological state by the corresponding combined transformation matrix, to obtain the registered coordinates.

[0012] Preferably, the S7 calculates the average value of the curve according to the landmark point segmentation, and the specific steps of combining the average values of the segmented calculation into the average value of the whole curve are as follows: Step 1: Determine the segmentation boundary, which is determined by the index of the anatomical landmark point on the interpolated curve; Step 2: Then determine the number of segments, which is one more than the number of landmark points; Step 3: Calculate the average value of each segment of the curve; Step 4: Finally, splice the average point sequence of each segment into a complete curve, and remove the repeated end points of adjacent segments during splicing.

[0013] Preferably, the S8 calculates the self-center, local center and global center of each curve, and the specific steps of translating the self-center of the corresponding curve to coincide with the local and global centers of the corresponding curve are as follows: Step 1: Calculate the center of a single curve; Step 2: Calculate the global center; Step 3: Calculate the local center; Step 4: Move the self-center to coincide with the local center to align all samples of the same physiological state of all curves; move the self-center to coincide with the global center to align all samples (regardless of physiological state) of all curves.

[0014] The beneficial effects of the present application are as follows: Compared with the prior art, the present application has the following beneficial effects: Synchronize multiple curve standardization: Through the method of "cooperative rigid registration + individual center alignment", eliminate the position and direction differences of each curve in the unified coordinate system, while preserving the relative spatial relationship between curves; Improve the reliability of multi-structure analysis: avoid the distortion of the relationship between curves caused by independent registration, suitable for the analysis of the overall morphology composed of multiple related curves, such as the evaluation of the cooperative change of multiple structures of the pelvic floor organs; Higher registration accuracy: combined with rigid transformation (correct global offset) and center alignment (optimize local morphology), the method is more suitable for curve morphology characteristics than single registration method; With cooperative verification mechanism: through the center distance deviation verification, ensure that the inherent relationship between the registered curves is not destroyed, and the result is traceable; Adapt to multi-state scene: for different physiological state of different patients curve group, can realize the standardization comparison across patients across state, to provide support for dynamic function analysis. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is a flow chart of a female pelvic floor multi-target curve medical image cooperative registration and alignment method provided by the embodiment of the application; Figure 2 is a SCIPP line and target curve contour schematic diagram of the first patient in the Valsalva state manually labeled in mimics provided in the embodiment of the application; Figure 3 is a respective alignment schematic diagram of all samples of all curves in the same physiological state; Figure 4 is a respective alignment schematic diagram of all samples of all curves (regardless of physiological state). DETAILED DESCRIPTION

[0016] In order to clearly and completely describe the technical solutions and technical effects of the application, the following embodiments are used for detailed description.

[0017] Embodiment 1 Referring to the accompanying Figure 1 , a female pelvic floor multi-target curve medical image cooperative registration and alignment method includes the following steps: S1, in each sample data, mark the target curve, SCIPP line and segmented marker point; In this embodiment, the Mimics software is used to mark the bladder urethra, levator plate, rectum, uterine vaginal contour curve, SCIPP line and segmented marker point in the median sagittal plane, and each state of the resting state, the anal sphincter state and the Valsalva state needs to be marked once, as shown in the accompanying Figure 2 . Among them, the bladder urethral contour curve selects two marker points, which are the uppermost triangle area and the bladder neck; the levator plate contour curve selects the midpoint of the puborectalis muscle as the marker point; the rectal contour curve selects the bladder rectal junction as the marker point; the uterine vaginal contour curve selects the internal and external cervical orifice of the uterus as the marker point. The target curve is named “sample ID - curve name - state”, and the SCIPP line file is named “sample ID-SCIPP -state”, so as to ensure that the curves of the same sample and the same physiological state are associated with the SCIPP line under the same sample and the same physiological state, and then exported in csv format after marking; S2, adopt Y-Z plane projection combined with X coordinate dynamic compensation dimension reduction method to reduce three-dimensional medical data to two-dimensional; In this embodiment, first need to analyze the X coordinate variation, calculate the X coordinate standard deviation and Y coordinate standard deviation of each target curve, get the variation ratio by X coordinate standard deviation divided by Y coordinate standard deviation, then calculate the average variation ratio of all curves, evaluate the influence of X coordinate on morphology, and then determine the value of compensation factor k according to the calculated average variation ratio; If the average variation ratio < 0.05 (X has little influence): k = 0; If 0.05 ≤ average variation ratio < 0.1 (X has little influence): k = 0.05; If 0.1 ≤ average variation ratio < 0.2 (X has moderate influence): k = 0.1; If the average variation ratio ≥ 0.2 (X has greater influence): k = 0.15; According to the formula: Y' = Y + (X norm × k × ystd) Where Xnorm = (X-Xmin) / (Xmax-Xmin), ystd is the standard deviation of the Y coordinate of the curve, and the three-dimensional coordinates (X, Y, Z) of the target curve and the SCIPP line are converted into two-dimensional coordinates (Y', Z).

[0018] S3, before registration, smooth the curve by cubic spline interpolation to avoid the influence of original data noise on registration accuracy; In this embodiment, the output two-dimensional discrete point coordinates (Y', Z) are sorted according to point index (to ensure curve continuity), the distance between adjacent points is calculated, the cumulative chord length L = [0, d_1, d_1+d_2,..., Σd_i] is normalized to [0,1] as the interpolation parameter t, the cubic spline function is constructed for Y' and Z coordinates, 300 uniformly distributed interpolation points are generated, and then the smooth curve coordinates are calculated and the smoothed two-dimensional curve is output.

[0019] S4, calculate the translation matrix based on the reference point of the lower edge of the pubic symphysis of the SCIPP line, and move the first reference point of the lower edge of the pubic symphysis to the origin. In this embodiment, the specific method of coinciding the lower edge of the pubic symphysis with the origin point includes the following steps: Extract the coordinates (y1, z1) of the first reference point P1 (the lower edge of the pubic symphysis) from the smoothed SCIPP line two-dimensional data; Construct a translation matrix T to move P1 to the origin (0, 0) of the coordinate system, formula: T= The coordinates of the first reference point P1 (the lower edge of the pubic symphysis) of the SCIPP line of each sample and each state are respectively brought into the translation matrix T to obtain the corresponding translation matrix.

[0020] S5, taking the SCIPP line as a reference vector, and using the difference between the reference vector and the target angle to calculate a rotation matrix; In this embodiment, the specific method for calculating the rotation matrix includes the following steps: From the two-dimensional data of the SCIPP line, the coordinates (y2, z2) of the second reference point P2 (such as the end of the pubic ramus) are extracted; The SCIPP reference vector a is calculated, and the vector is taken as an initial vector a = [y2-y1, z2-z1] The angle a between the initial vector and the positive direction of the x axis is calculated: a = arctan2 (z2-z1, y2-y1) The rotation angle theta and the rotation matrix are calculated: The vector is rotated clockwise to the horizontal position as the Y axis of the coordinate system (generally needs to be rotated by 32°~36°, depending on the actual sample effect), and in this embodiment, the vector is rotated clockwise by 34°, so the rotation angle theta = a-34° (the corresponding angle of clockwise rotation is negative) The rotation matrix R is constructed, and the formula is: R= S6, applying the corresponding translation and rotation matrix to the corresponding sample curve data to establish a unified coordinate system; In this embodiment, the combined transformation matrix M is first calculated, and the combined transformation matrix M = T x R That is M= Then, each target curve data (smoothed two-dimensional point set) of the same sample and the same physiological state is multiplied by the corresponding combined transformation matrix to perform coordinate conversion to obtain the registered coordinates.

[0021] S7, calculating the average value of the curve by segmenting according to the marker points, and then combining the average values calculated by segmenting into the average value of the whole curve. In this embodiment, the specific method for calculating the average value of the curve by segmenting includes the following steps: Step 1: determination of the segmentation boundary; the segmentation boundary is determined by the index of the anatomical marker point on the interpolated curve. Then the number of segments is determined, and the formula is: M = n + 1 Wherein, M is the number of segments, and n is the number of anatomical marker points (such as 2 marker points corresponding to 3 segments).

[0022] Step 2: Single segment average point sequence calculation formula; for the pth segment (p = 1, 2,..., M), the coordinate calculation formula of its average point sequence is: wherein, : the tth average point coordinate of the pth segment (t = 0, 1,..., L p -1, L p is the number of points in the pth segment); K: the number of curves in the same group (such as the resting state bladder urethral curve group has 12 curves, K = 12); : the aligned y coordinate of the jth curve at the tth point in the pth segment; : the aligned z coordinate of the jth curve at the tth point in the pth segment; Step 3: Splicing the average point sequence of each segment into a complete curve; when splicing, the repeated end points of adjacent segments need to be removed (only the starting end point of the first segment and the terminal end point of the last segment are retained).

[0023] S8, calculate the self-center, local center and global center of each curve, and translate the self-center of the corresponding curve to coincide with the local and global centers of the corresponding curve to achieve alignment; In this embodiment, the specific method of calculating the center of the curve and translating the center of the corresponding curve to coincide with the local and global centers of the corresponding curve includes the following steps: Step 1: Calculate the center of a single curve; for a curve with a length of n, the y axis and z axis self-center calculation formula is: wherein, : the self-center coordinate of the jth curve; : the y, z coordinates of the tth point after interpolation of the jth curve; n: the uniform length of the curve after interpolation.

[0024] Step 2: Calculation of global center; for K curves of the same curve type, the global center calculation formula is: C wherein, C: global center coordinate of curve type dimension; : y, z coordinates of the self-center of the jth curve; K: total number of curves of the same curve type.

[0025] Step 3: Calculation of local center; for K g curves of a certain physiological state group of a certain curve type, the local center calculation formula is: C l wherein C l : local centroid coordinate of the group; K g : number of curves in the group.

[0026] Step 4: Centroid alignment; refer to Fig. 4, move the self-centroid to coincide with the local centroid, and achieve alignment of all samples of all curves in the same physiological state; refer to Fig. 5, move the self-centroid to coincide with the global centroid, and achieve alignment of all samples of all curves. Figure 3 Figure 4

[0027] Obviously, the above-mentioned embodiments of the present application are only examples for clearly illustrating the present application, and are not intended to limit the embodiments of the present application. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, it is not necessary and impossible to enumerate all the embodiments. Any modification, equivalent replacement and improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the claims of the present application.​​

Claims

1. A method for female pelvic floor multi-objective curve medical image co-registration and alignment, characterized in that, The method comprises the following steps: S1: In each sample data, mark the target curve, SCIPP line and segment marker; S2: Use Y-Z plane projection combined with X coordinate dynamic compensation to reduce three-dimensional medical data to two-dimensional; S3: Before registration, smooth the curve by cubic spline interpolation to avoid the influence of original data noise on registration accuracy; S4: Calculate the translation matrix based on the reference point of the SCIPP line, and move the first reference point of the lower pubic symphysis to the origin; S5: Take the SCIPP line as the reference vector, and calculate the rotation matrix using the difference between the reference vector and the target angle; S6: Apply the corresponding translation and rotation matrix to the corresponding sample curve data to establish a unified coordinate system; S7: Calculate the average value of the curve according to the marker points, and then combine the average values of the segmented calculation into the average value of the whole curve; S8: Calculate the self-centroid, local centroid and global centroid of each curve, and respectively translate the self-centroid of the corresponding curve to coincide with the local and global centroids of the corresponding curve to achieve alignment.

2. The female pelvic floor multi-objective curve medical image co-registration and alignment method according to claim 1, characterized in that, The specific steps of S1 are as follows: Step 1: In three-dimensional medical images (such as MRI, CT), manually or automatically label two types of data, including target curve data and SCIPP line data, and place marker points at key positions in the target curve data, and each curve can have multiple marker points; Step 2: After labeling, save according to the format, name the target curve as "sample ID - curve name - state", and name the SCIPP line file as "sample ID-SCIPP-state", and ensure that the curves and corresponding SCIPP lines of the same sample and the same physiological state (such as resting state) are associated.

3. The female pelvic floor multi-objective curve medical image co-registration and alignment method according to claim 1, characterized in that, The specific steps of three-dimensional to two-dimensional in S2 are as follows: Step 1: Extract the three-dimensional coordinates of all target curves and SCIPP lines; Step 2: Analyze the X coordinate variation, get the variation ratio by comparing the X coordinate standard deviation with the Y coordinate standard deviation, and calculate the average variation ratio of all curves to evaluate the influence of X on the shape; Step 3: Determine the value of compensation factor k according to the average variation ratio calculated in step 2; Step 4: Convert the three-dimensional coordinates of the target curve and the SCIPP line to two-dimensional coordinates.

4. The female pelvic floor multi-objective curve medical image co-registration and alignment method according to claim 1, characterized in that, The specific steps of interpolation for target curve data in S3 are as follows: Step 1: Sort the output two-dimensional discrete point coordinates by point index to ensure curve continuity; Step 2: Calculate the distance between adjacent points for the sorted point set, accumulate the chord length L = [0, d_1, d_1+d_2,...,Σd_i], and normalize it to [0,1] as the interpolation parameter t; Step 3: Construct a cubic spline function for Y and Z coordinates to generate 300 evenly distributed interpolation points, and then calculate the smoothed curve coordinates, output the smoothed two-dimensional curve.

5. The female pelvic floor multi-objective curve medical image co-registration and alignment method according to claim 1, characterized in that, The specific steps of calculating the translation matrix in S4 are as follows: Step 1: Extract the coordinates of the first reference point P1 (lower pubic symphysis) from the smoothed SCIPP line two-dimensional data; Step 2: Construct a translation matrix T to move P1 to the origin (0,0) of the coordinate system; Step 3: The coordinates of the first reference point P1 (the lower edge of the pubic symphysis) of the SCIPP line of each sample and each state are respectively brought into the translation matrix T to obtain the corresponding translation matrix.

6. The female pelvic floor multi-objective curve medical image co-registration and alignment method according to claim 1, characterized in that, The specific steps of calculating the rotation matrix in S5 are as follows: Step 1: Extract the coordinates of the second reference point P2 (such as the end of the pubic ramus) from the two-dimensional data of the SCIPP line; Step 2: Calculate the SCIPP reference vector a and take the vector as the initial vector Step 3: Calculate the angle a between the initial vector and the positive direction of the x-axis: Step 4: Calculate the rotation angle θ and the rotation matrix R: Rotate the vector clockwise to the horizontal position as the Y-axis of the coordinate system, which generally needs to be rotated by 32°~36°, and the corresponding angle of clockwise rotation is negative according to the actual sample effect.

7. The female pelvic floor multi-objective curve medical image co-registration and alignment method according to claim 1, characterized in that, The specific steps of unifying the coordinate system in S6 are as follows: Step 1: Calculate the combined transformation matrix M, M=T×R; Step 2: Multiply each target curve data of the same sample and the same physiological state by the corresponding combined transformation matrix to perform coordinate conversion to obtain the registered coordinates.

8. The female pelvic floor multi-objective curve medical image co-registration and alignment method of claim 1, wherein, The specific steps of calculating the average value of the curve data in S7 are as follows: Step 1: Determine the segmentation boundary, which is determined by the index of the anatomical marker point on the interpolated curve; Step 2: Then determine the number of segments, which is one more than the number of marker points; Step 3: Calculate the average value of each segment of each curve; Step 4: Finally, splice the average point sequence of each segment into a complete curve, and remove the duplicate end points of adjacent segments when splicing.

9. The female pelvic floor multi-objective curve medical image co-registration and alignment method of claim 1, wherein, The specific steps of implementing centroid alignment in S8 are as follows: Step 1: Calculate the single curve centroid; Step 2: Calculate the global centroid; Step 3: Calculate the local centroid; Step 4: Centroid alignment; Move the own centroid to coincide with the local centroid to achieve the alignment of all samples of the same physiological state of all curves; move the own centroid to coincide with the global centroid to achieve the alignment of all samples of all curves.

Images