A method for standardizing pelvic floor contour based on pelvic coordinate correction and landmark zoning

By using pelvic coordinate correction and anatomical landmark partitioning, the problems of template deformity and mismatch in pelvic organ morphology analysis were solved, achieving standardization and stability of pelvic organ two-dimensional contours and improving the accuracy and reliability of deformable template construction.

CN123118558APending Publication Date: 2026-11-17KUNMING UNIV OF SCI & TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611358122.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-03
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing techniques for morphological analysis of pelvic floor organs struggle to establish stable anatomical correspondences between coordinate unification, topology preservation, anatomical partitioning, and partition resampling, leading to template malformation, local area mismatch, and distortion of statistical morphological analysis results.

Method used

The method of maintaining partitions through pelvic coordinate correction and anatomical landmark topology includes establishing a pelvic reference coordinate system, organ topology identification, anatomical landmark partitioning, and equal arc length resampling, combined with multi-dimensional quality control, to ensure the stability of contour data under different functional states and organ topology types.

Benefits of technology

It reduces the impact of coordinate pose differences among different subjects or states, avoids incorrect closure of open curves and local region mismatch, improves the stability of local region correspondence and batch calculation, reduces pollution of abnormal labeled data, and avoids initial template bias.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118558A_ABST
    Figure CN123118558A_ABST
Patent Text Reader

Abstract

This invention relates to the field of medical image data processing and pelvic floor organ morphology analysis, specifically a method for pelvic floor contour standardization based on pelvic coordinate correction and landmark point partitioning. The method involves acquiring labeled and exported 2D multi-organ contour data, bony reference landmarks, and organ anatomical landmarks from the same subject under at least two functional states on MRI; establishing a pelvic reference coordinate system based on the bony landmarks and performing contour coordinate correction; determining closed contour or open curve topology rules according to organ type; dividing the bladder and uterus into multiple open polyline partitions based on landmarks (BNP, BNA, BNA, BNP ...
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of medical image data processing, morphological analysis of pelvic floor organs, two-dimensional contour standardization, deformation template construction preprocessing, and computer-aided medical image analysis, and particularly to a method for two-dimensional contour standardization of pelvic floor organs based on pelvic coordinate correction and anatomical landmark partitioning. Background Technology

[0002] Morphological analysis of pelvic floor organs often uses pelvic floor magnetic resonance images, ultrasound images, or two-dimensional projected contours reconstructed from three-dimensional images as input data. However, due to the influence of imaging position, pelvic tilt, manual annotation habits, and differences in organ topology, the original contour data cannot usually be directly used for deformation template construction. In order to further carry out deformation template construction and morphological statistical analysis, it is usually necessary to transform the organ contours of different subjects to the same coordinate system and the same data structure.

[0003] In existing technologies, common processing methods include uniform arc-length resampling of the entire contour, rigid registration of the entire organ, or direct input of non-rigid deformation registration algorithms. These methods suffer from problems such as inconsistent annotation starting points, inconsistent point order and direction, and inconsistent sampling density. However, such methods tend to ignore the topological features and anatomical partitioning information of different organs. If the processed contour is directly input into the deformation template construction process, it may lead to template distortion, local region mismatch, curve intersection, and distortion of statistical morphological analysis results, thereby affecting the accuracy and verifiability of the results.

[0004] For the bladder contour, although it can be regarded as a closed structure as a whole, if only the entire closed contour is uniformly resampled between different cases, especially between samples with different degrees of filling and large differences in shape, the same numbered sampling point may fall in different anatomical regions, resulting in mismatches in the anterior wall, posterior wall, top and bladder neck regions in subsequent templates. For the uterine open curve, if it is forcibly closed, it will introduce non-existent connecting edges, causing connecting lines that do not conform to the actual anatomical structure to appear in subsequent template construction or PCA visualization.

[0005] Therefore, a standardized method is needed for the analysis of two-dimensional contour deformation of multiple organs in the pelvic floor. This method should establish a complete technical link between coordinate unification, topology preservation, anatomical partitioning, partition resampling, and quality control. It should be able to establish two-dimensional contour data of multiple organs in the pelvic floor with stable anatomical correspondence under different functional states, different organ topology types, and different original contour point numbers, while preventing samples with topological errors and incorrect marker points from entering the statistical shape modeling process. Summary of the Invention

[0006] The purpose of this invention is to provide a two-dimensional contour standardization method for pelvic organs based on pelvic coordinate correction and anatomical landmark topological relationship preservation, in order to solve the above-mentioned technical problems.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Step S1: Construct a dataset of pelvic floor organ contours and landmarks and complete the data reading. The pelvic floor MRI two-dimensional multi-organ contour data, bony reference landmarks, and organ anatomical landmarks, labeled and exported, are obtained from the same subject in at least two functional states. Preferably, the pelvic floor MRI two-dimensional multi-organ contour data includes at least the closed contour of the bladder and at least one open contour of the uterus, rectum, and levator palpebrae superioris. The contour data and landmark data are stored in Markups JSON files or other structured data files containing physical coordinates, point sequence information, and object labels.

[0008] Furthermore, points in the original medical image can be first read into three-dimensional physical coordinates or image coordinates, and then projected onto the midsagittal plane of the pelvic floor to form two-dimensional coordinates. The two-dimensional coordinate system is defined as follows: the x-axis is along the front-back direction of the image, pointing in front of the subject as the positive direction; the y-axis is along the top-bottom direction of the image, pointing to the side of the head or above as the positive direction.

[0009] Step S2: Establish pelvic reference coordinates and perform contour coordinate correction. A pelvic reference coordinate system is established based on the bony reference landmarks. The coordinate transformation parameters corresponding to each functional state are calculated, and the two-dimensional contour point set and organ anatomical landmarks are transformed to the pelvic reference coordinate system.

[0010] Furthermore, the establishment and transformation calculation of the pelvic reference coordinate system are as follows: Using the lower edge of the pubic symphysis as the origin The sacrococcygeal reference point is denoted as Connect the two points to form the SCIPP direction vector: in For the first Sacrococcygeal reference points in a functional state This refers to the lower edge of the pubic symphysis in this state.

[0011] Calculate its direction angle based on the SCIPP direction vector: Let the target state correction angle be... The effective rotation angle is: Construct a two-dimensional rotation matrix: Two-dimensional contour points and organ anatomical landmarks Perform coordinate transformation as follows: After the transformation, the pelvic reference axis corresponding to this state is aligned with the preset unified analysis axis.

[0012] Preferably, the functional state includes at least two of the resting state, Valsalva state, and anal sphincter contraction state; when the data of different functional states come from different image sequences, the two-dimensional contour points and landmark points of each functional state are mapped to the common physical coordinate system of the same subject based on the spatial location information of each image sequence.

[0013] Furthermore, the coordinate transformation supports two output modes: The first is the state correction version: a state transformation matrix is ​​constructed for each functional state to reduce the average pelvic tilt difference between different states. Second, a unified bony coordinate system: a unified reference transformation matrix is ​​constructed by selecting the reference state of the same subject (preferably the resting state, but can also be specified by the user). The contours of the subject in different states are uniformly output to the same bony coordinate system using the following formula to preserve dynamic changes in pelvic pose: in, , and These are the origin of the coordinate system, the pelvic reference direction angle, and the reference correction angle, respectively, under the reference state.

[0014] Step S3: Organ topology identification and determination of closure / open processing rules The topological processing rules for the two-dimensional contours are determined based on organ type, wherein the two-dimensional contour of the bladder is determined as a closed contour, and the two-dimensional contours of the uterus, rectum, and levator palpebrae are determined as open curves.

[0015] Preferably, the open curve rule strictly preserves the original start and end points, does not add closed connecting edges from the start to the end point, and avoids introducing virtual connecting lines that do not conform to the anatomical structure.

[0016] Step S4: Preserve partitioning based on the topological relationships of anatomical landmarks Based on the position of the anatomical landmarks of the organs on the corresponding two-dimensional contours, the two-dimensional contours are divided into topologically preserved partitions, and the open or closed topological types of each organ contour, the arrangement order of each anatomical partition, and the shared endpoint relationships of adjacent partitions are recorded.

[0017] When marker point m is an independent annotation point, calculate its distance to the i-th point on the contour. Euclidean distance: The contour point with the smallest distance is selected as the corresponding point of the marker; if the smallest distance exceeds the preset matching threshold... If so, it is determined that the marker point matching is abnormal.

[0018] Furthermore, the rules for zoning each organ are as follows: 1. Bladder division: based on anatomical landmarks As a unified starting point, according to → , →BNP, BNP→BNA and BNA→ The points sequentially form four open Polyline regions: the top region, the posterior wall region, the bladder neck and bottom region, and the anterior wall region. The original closed contour can be completely reconstructed by connecting the beginning and end of each region.

[0019] in, It is the junction between the top of the bladder and the anterior wall. BNP is the junction of the top and posterior wall of the bladder, BNA is the posterior point of the bladder neck, and each landmark is the shared endpoint of adjacent partitions.

[0020] 2. Uterine division: Following the open curve point sequence, according to the starting point... → , → as well as →End point The point sequence directions sequentially form three open polyline partitions corresponding to the lower anterior wall, the fundus of the uterus, and the lower posterior wall; among them, It is the junction of the lower segment of the anterior wall of the uterus and the fundus of the uterus. It is the junction between the fundus of the uterus and the lower segment of the posterior uterine wall.

[0021] 3. Rectum and Levator Plate: Maintain the original open curve topology and treat them as independent Polyline objects without performing additional partitioning.

[0022] Preferably, for the closed contour of the bladder, making The corresponding contour point is located at index 0 of the point sequence. Forward and reverse candidate point sequences are constructed respectively, and the unique one that satisfies the condition is selected. → →BNP→BNA→ The sequence of anatomical dissections is used as the standard point sequence; if there is no sequence that meets the conditions or if there are multiple sets of sequences that meet the conditions, it is considered a quality control failure.

[0023] Step S5: Resampling of fixed points with equal arc length in each anatomical region The cumulative arc length of each open Polyline partition is calculated, and each open Polyline partition is resampled with equal arc length according to the preset number of sampling points for the corresponding organ and the corresponding anatomical partition, so that the same organ and the same anatomical partition of different subjects have the same number of sampling points, point order direction and partition label.

[0024] Furthermore, the resampling calculation process is as follows: Let the original ordered contour point sequence of the r-th partition be... The cumulative arc length from the starting point of the partition to the i-th point is: The total arc length of this partition is the cumulative arc length at the endpoint: Let N be the number of target resampling points in this partition. r Then the target arc length corresponding to the j-th resampling point is: Based on the original arc length interval where the target arc length is located Perform linear interpolation to obtain the coordinates of the resampling points: Preferably, the same fixed number of sampling points is used for the same anatomical region of the same organ in different subjects and under different functional states; the number of sampling points is determined in advance based on the average arc length ratio of the region in the reference training set, and cannot be adjusted for individual samples after determination.

[0025] Step S6: Multi-dimensional quality control scoring and sample grading screening Quality control is implemented based on the original two-dimensional contour, organ anatomical landmark matching results, and open Polyline partitions after resampling; according to the preset serious error judgment rules and comprehensive quality score results, the corresponding samples are classified into qualified sample set, warning sample set, or failed sample set, and the organ, anatomical partition, point sequence location, and abnormality type that trigger the quality abnormality are recorded.

[0026] Furthermore, quality control covers multiple dimensions before and after resampling. Before resampling, the integrity of marker points, the order of marker point arrangement, the contour topology type, and the number of original contour points in the partition are checked. After resampling, the partition length, curvature change, duplicate points, and self-intersection are checked.

[0027] Preferably, the overall quality score is obtained by weighting five individual score items: in, Scoring the integrity of the marker points For point-order scoring, Scoring based on partition length Scoring for curvature stability, Scoring for self-intersection and repetition points. to Let be the weighting coefficient, and satisfy: The calculation methods for each score are as follows: 1) Marker Completeness Scoring: Let the total number of required markers be... The number of missing items is ,but 2) Marker point order scoring: Let the number of inversions in the marker point index sequence be... The maximum possible number of inversions is ,but 3) Partition Length Scoring: Let the mean partition length of the group be _____. The standard deviation is The current sample partition length is The deviation is The individual score is: Where k is the tolerance threshold. To prevent positive numbers with a denominator of zero.

[0028] 4) Curvature stability score: For three consecutive points within the partition , , Discrete curvature is Count the number of curvature outliers The total number of points is ,but 5) Self-intersection and duplicate point scoring: Let the number of duplicate points be... The number of self-intersections is ,but Where a is the deduction value for a single duplicate point, and b is the deduction value for a single self-intersection.

[0029] Furthermore, the sample grading rules are as follows: If the quality score is greater than or equal to the first threshold, it is included in the qualified sample set; When the second threshold is less than or equal to the quality score but less than the first threshold, the sample is classified into the warning sample set. If the quality score is less than the second threshold, or if a pre-defined serious error exists, the sample is classified into the failure sample set.

[0030] Among them, the preset serious errors include missing organ landmarks, landmark order not conforming to anatomical rules, incorrect closure of open curves, and self-intersection of contours; when self-intersection exists, regardless of the overall score, it is directly judged as a failed sample.

[0031] Step S7: Output standardized multi-object data and quality control results The open Polyline partitions in the qualified sample set are combined into standardized multi-object Polyline data according to organ labels, anatomical partition labels and sampling point numbers; the data in the warning sample set is output to the review process, and step S6 is re-executed after correction; only the data that is re-determined to be qualified is output as standardized multi-object Polyline data; the data in the failed sample set is prevented from directly entering the pelvic floor multi-organ deformation template or statistical shape model, and a quality control report is output.

[0032] Preferably, the standardized multi-object Polyline data includes organ identifiers, anatomical partition identifiers, functional status identifiers, sampling point numbers, and sampling point coordinates; the quality control report includes sample quality labels, abnormal organs, abnormal partitions, abnormal types, and abnormal point locations.

[0033] Furthermore, standardized data from qualified samples can be directly used as input for multi-object deformation templates or statistical shape models; warning samples must be manually reviewed and corrected, and re-qualified before they can enter the subsequent process; failed samples are prohibited from directly entering the modeling stage and the reasons for the anomalies must be marked for problem localization.

[0034] Preferably, an initial deformation template can be selected based on qualified standardized data using a Medoid strategy. The calculation process is as follows: The multi-object contour distance between any two subjects i and j is: Where M is the number of objects. Let m be the weight of the m-th object. This represents the outline distance of the corresponding object.

[0035] The cumulative distance from the i-th candidate to all other candidates in the same group is: The subject with the smallest cumulative distance was selected as the Medoid center sample: Using all partitions of the sample as the initial template (Polyline) can avoid bias caused by manually specifying the template.

[0036] Compared with the prior art, the present invention has the following beneficial effects: First, the pelvic reference coordinate correction is combined with the standardization of organ contour partitioning to reduce the impact of coordinate posture differences among different subjects or in different states on morphological analysis. Secondly, organ topology recognition avoids incorrect closure of open curves, reducing distortion of contour curves and mismatch in local areas; Third, by resampling anatomical landmarks, the stability of local region correspondence is improved, so that the same anatomical region of the same organ has a more stable local correspondence among different subjects. Fourth, by screening abnormal samples before constructing the deformation template through quality control, the stability of batch calculation is improved and the pollution of subsequent deformation template construction by abnormal annotation data is reduced. Fifth, by using Medoid samples as the initial template, the bias of the initial template caused by random samples or manually specified samples can be avoided. The multi-object Polyline output provides standard input for subsequent deformation template construction, statistical morphological analysis and visualization verification. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the pelvic floor contour standardization method for pelvic coordinate correction and marker point partitioning according to the present invention. Figure 2 This is a schematic diagram of the bladder closed contour open section and the uterus open three sections of the present invention; Figure 3 This is a flowchart illustrating the quality control and abnormal sample triage process for standardized input data in this invention. Detailed Implementation

[0038] The following is in conjunction with the appendix Figure 1 To be continued Figure 3 The present invention will be further illustrated by specific embodiments. It should be noted that these embodiments are implemented based on the technical solution of the present invention, and detailed implementation methods and specific operation processes are given, but the protection scope of the present invention is not limited to the following embodiments.

[0039] Example 1 This embodiment performs full-process standardization processing on the annotated contour data of a set of pelvic floor MRI midsagittal plane images. The data comes from Markups JSON files exported by 3D Slicer medical image annotation software, and each subject includes annotation results for three functional states: resting state, Valsalva maneuver, and anal sphincter contraction state. For details, please refer to the appendix. Figure 1 As shown, a method for pelvic floor contour standardization based on pelvic coordinate correction and landmark point partitioning includes the following steps: Step S1: Construct a dataset of pelvic floor organ contours and landmarks and complete the reading.

[0040] Parse the Markups JSON file to read the organ contour point sequence, the lower border of the pubic symphysis, the sacrococcygeal reference point, and the anatomical landmarks of each organ. Represent each contour point as a two-dimensional coordinate vector. The x-axis runs along the front-to-back direction of the image (positive for those pointing forward to the subject), and the y-axis runs along the top-to-bottom direction of the image (positive for those pointing towards the head). The three-dimensional physical coordinates are first projected onto the midsagittal plane of the pelvic floor, and the two-dimensional coordinates within the sagittal plane are used in subsequent calculations.

[0041] The data retrieved included: bladder closure contour (average 180-260 original control points), uterine opening curve (average 120-180 points), rectal opening curve (average 100-150 points), levator palpebrae superioris opening curve (average 60-100 points), and bony landmarks (O2 at the lower border of the pubic symphysis). s Sacral-coccygeal reference point s and organ anatomical landmarks (bladder) BNP, BNA; uterus , ).

[0042] The contour data and marker point data are stored in a MarkupsJSON file containing physical coordinates, point sequence information, and object labels.

[0043] Step S2: Establish pelvic reference coordinates and perform contour coordinate correction.

[0044] Using the lower edge of the pubic symphysis as the origin The sacrococcygeal reference point is denoted as Connect the two points to form the SCIPP direction vector: Where s is the functional state number (s=1 resting state, s=2 Valsalva state, s=3 anal sphincter state). The orientation angle is calculated based on the SCIPP orientation vector: In this embodiment, the pre-stored state-angle parameter table sets the target correction angle for all three functional states to 0°, meaning the SCIPP axis is uniformly rotated to the horizontal direction. The effective rotation angle is: Construct a two-dimensional rotation matrix: Two-dimensional contour points and organ anatomical landmarks Perform coordinate transformation as follows: After transformation, the pelvic reference axis is aligned with the horizontal analysis axis in all states.

[0045] This embodiment supports two coordinate output modes: one is the state correction version, which constructs a state transformation matrix for each functional state. T s The first step is to correct pelvic tilt in each state to reduce the average pelvic posture difference between different functional states; the second step is to unify the bony coordinate system by selecting the resting state of the same subject as the reference state to construct a unified reference transformation matrix. The contours of the subject in the three states are then uniformly output to the same bony coordinate system using the following formula: in, , and These are the coordinate origin, pelvic reference direction angle, and reference correction angle in the resting state, respectively. This mode is used to preserve the dynamic pelvic pose changes of the subject in different functional states. The reference state can also be specified by the user as Valsalva state or anal contraction state. When the data of the three functional states come from different MRI sequences, the contour points and landmark points are first mapped to the common physical coordinate system of the same subject based on the DICOM spatial positioning information of each sequence, and then the above coordinate transformation is performed.

[0046] Step S3: Organ topology identification and determination of closure / open processing rules.

[0047] Based on organ labels, topological processing rules are determined. The bladder's two-dimensional contour is defined as a closed contour, while the uterus, rectum, and levator palpebrae superioris' two-dimensional contours are defined as open curves. The open curve rules strictly preserve the original start and end points, without adding closed connecting edges from the start to the end point, thus avoiding the introduction of virtual connecting lines that do not conform to the anatomical structure.

[0048] Step S4: Preserve partitioning based on the topological relationships of anatomical landmarks.

[0049] The system is partitioned based on the positional sequence of anatomical landmarks on the corresponding two-dimensional contour. The matching rule between landmarks and contour points is as follows: when a landmark is already used as a contour control point, its index position is directly read; when a landmark is an independent annotation point, the distance from landmark m to the i-th point on the contour is calculated. Euclidean distance: The contour point with the smallest distance is selected as the corresponding point of the marker point. In this embodiment, the marker point matching distance threshold is used. =2.0mm. If the minimum distance exceeds this threshold, it is determined to be an abnormal marker matching.

[0050] Bladder partitioning: Reading the bladder closure contour point sequence B={b 0 , b 1 , …, b {n-1} } Identify , Four markers: BNP, BNA, etc. It is the junction between the top of the bladder and the anterior wall. The point where the top of the bladder meets the posterior wall is the junction of the bladder neck; BNP is the posterior point of the bladder neck, and BNA is the anterior point of the bladder neck. The corresponding contour point is located at index 0 of the point sequence. Forward and reverse candidate point sequences are constructed respectively, and the unique one that satisfies the condition is selected. → →BNP→BNA→ The anatomical sequence is used as the standard point sequence. The four open sections of the bladder closure contour and the three open sections of the uterine opening curve are shown in the appendix. Figure 2 As shown, Figure 2 Different gray colors are used to distinguish the anatomical regions, and the locations of shared anatomical landmarks are marked. The clockwise index of landmarks on the normal bladder outline should meet the following requirements. idx < idx <idxBNP< idx BNA. If there is no sequence that meets the conditions or if there are multiple sequences that meet the conditions, the quality control is considered to have failed.

[0051] The standard bladder point sequence was divided into four segments to form four open polyline partitions. → (Top area) →BNP (posterior wall region), BNP→BNA (bladder neck and base region), BNA→ (Anterior wall region). The original closed contour can be completely reconstructed by connecting the beginning and end of each section, and adjacent sections share the corresponding anatomical landmarks as endpoints.

[0052] Uterine partitioning: Reading the uterine opening curve point sequence U={u 0 , u 1 , …, u m } Keep the original starting point u0 and ending point u m Not closed. Recognition. (The junction of the lower segment of the anterior uterine wall and the fundus of the uterus) and Two landmarks: (the junction of the fundus of the uterus and the lower segment of the posterior uterine wall). If or If a sample is missing, it will be added to the warning or failed sample set; if the order is different from the preset anatomical direction ( idx U0< idx < idx < idx If U1) is inconsistent, record the abnormal point sequence and generate a verification map. If the sequence meets the conditions, the uterus is divided into... → (Lower segment of the anterior wall) → (Fundus of the uterus) → (Lower section of the rear wall) Three open Polyline partitions.

[0053] Rectum and levator plate: Maintain the original open curve topology and treat them as independent Polyline objects without performing additional partitioning.

[0054] Each Polyline object retains an independent object number, organ label, anatomical partition label, partition start point marker, partition end point marker, and preset number of resampling points.

[0055] Step S5: Resample fixed points of equal arc length in each anatomical region.

[0056] Perform equal-arc-length resampling on all open Polyline partitions. Let the original ordered contour point sequence of the r-th partition be... The cumulative arc length from the starting point of the partition to the i-th point is: The total arc length of this partition is the cumulative arc length at the endpoint: Let N be the number of target resampling points in this partition. r Then the target arc length corresponding to the j-th resampling point is: Based on the original arc length interval where the target arc length is located Perform linear interpolation to obtain the coordinates of the resampling points: The r-th partition after resampling is represented as .

[0057] In this embodiment, the fixed number of sampling points for each anatomical region is uniformly set in advance as follows: 50 points each for the top, posterior wall, neck and bottom, and anterior wall of the bladder; 50 points each for the lower segment of the anterior wall, bottom, and lower segment of the posterior wall of the uterus; 200 points for the rectal opening curve; and 200 points for the levator palpebrae superioris opening curve. The above number of sampling points is predetermined based on the average arc length ratio of each anatomical region in the reference training dataset, and is used uniformly for all subjects and all functional states. It should not be recalculated for individual samples.

[0058] Step S6: Multi-dimensional quality control scoring and sample grading screening.

[0059] Perform multi-dimensional quality checks on each sample before and after resampling. Before resampling, check the integrity of marker points, the order of marker point arrangement, the contour topology type, and the number of original contour points in each partition; after resampling, check the partition length, curvature changes, duplicate points, and self-intersections.

[0060] The overall quality score is obtained by weighting the scores of the five individual components: in, to Let be the weighting coefficient, and satisfy: In this embodiment, the weights are set as follows: marker integrity w1=0.25, point order w2=0.20, partition length w3=0.20, curvature stability w4=0.15, and self-intersection and repeating points w5=0.20.

[0061] The scoring methods for each item are as follows: (1) Marker completeness score: Let the total number of required markers be... The number of missing items is ,but: For example, the bladder requires four essential landmarks; if one is missing, then... =75 points.

[0062] (2) Marker point order scoring: Let the number of inversions in the marker point index sequence be... The maximum possible number of inversions is ,but: (3) Partition length score: Let the mean partition length of the group be 1. The standard deviation is The current sample partition length is Then the deviation is: in =1 10 -12 mm; Individual score: .

[0063] Tolerance threshold k=3.0 (4) Curvature stability score: for three consecutive points within the partition , , The discrete curvature is: when > =0.16mm -1 or the change in curvature at adjacent points | - |>Δ =0.10mm -1 When the curvature outlier is reached, it is recorded as a curvature outlier. The number of outliers is counted. Total points ,but: (5) Self-intersection and duplicate point scoring: Let the number of duplicate points be . (Non-shared endpoint pairs with a distance less than 0.5 mm), the number of self-intersections is ,but: The penalty for a single duplicate point is a=5, and the penalty for a single self-intersection is b=25. Endpoint duplications formed by adjacent partitions sharing anatomical landmarks are not counted as abnormal duplicate points.

[0064] Sample grading rules: In this embodiment, a first threshold T1 = 85 points and a second threshold T2 = 60 points are set. When Q >= 85, the sample is classified into the qualified sample set; when 60 <= Q < 85, the sample is classified into the warning sample set; when Q < 60 or a preset serious error exists, the sample is classified into the failed sample set. Preset serious errors include missing organ landmarks, landmark order not conforming to anatomical rules, incorrect closure of open curves, and self-intersection of contours; when... When the score is greater than 0, regardless of the overall score, it is directly judged as a failed sample. When the number of original contour points in any anatomical region is lower than the minimum number of valid points x... min When the score is 5, the sample is directly identified as abnormal and no further comprehensive score is calculated. The loop process for calculating the quality control score, the three-level sample triage (pass / warning / failure), and the re-quality control of warning samples after manual review is as follows: Figure 3 As shown, Figure 3 The system uses different path colors to distinguish three processing branches: direct output of qualified samples, re-entry of warning samples, and modeling of failed samples.

[0065] The quality control parameters used in this embodiment are summarized in Table 1.

[0066] Table 1. Implementation parameters for quality control and abnormal sample triage used in Example 1 Step S7: Output standardized multi-object data and quality control results.

[0067] For qualified samples, standardized multi-object Polyline data is output, including organ identifiers, anatomical region identifiers, functional state identifiers, sampling point numbers, and two-dimensional coordinates of the sampling points. The output format supports both Polyline JSON and VTK. A coordinate transformation parameter file is also output, recording the origin coordinates, SCIPP orientation angle, state correction angle, and rotation matrix for each functional state, facilitating result traceability.

[0068] The quality control report includes sample quality labels, abnormal organs, abnormal anatomical regions, abnormal types, and abnormal point sequence locations. Warning samples are output to a manual review process where operators check and correct the abnormal locations. After correction, step S6 is repeated, and the sample can only be output after being re-evaluated as acceptable. Failed samples are prevented from directly entering the multi-object deformation template or statistical shape model; the reason for failure and the location of the abnormality are output to facilitate problem localization.

[0069] Standardized data from qualified samples can be further processed using the Medoid strategy to select an initial deformation template: Standardized multi-object Polyline data from n subjects within the same functional state are input, and the multi-object contour distance between any two subjects i and j is: Where M=9, (4 bladder regions + 3 uterine regions + rectum + levator ani plate), since each region has already undergone fixed-point resampling, D m Use the root mean square distance between the corresponding points: The cumulative distance from the i-th candidate to all other candidates in the same group is: The subject with the smallest cumulative distance was selected as the Medoid center sample: Use all its partitions of Polyline as the initial template The multi-object deformation template construction module performs deformation calculations and optimizations, and outputs organ-level or partition-level average deformation templates for the corresponding functional states. The multi-object deformation template construction can employ methods based on Current distance, Varifold distance, or other distances without corresponding shapes. Its objective function can be expressed as... Where v i,t For the velocity field, For the endpoint deformation mapping, δ is the matching term weight parameter.

[0070] Example 2 This embodiment is an application of the method of the present invention under a simplified configuration. The difference from Embodiment 1 is that it only includes two functional states: resting state and Valsalva state; it only processes two organs: the bladder and the uterus; the coordinate output uses a unified bony coordinate system; and the number of sampling points in each partition uses a different configuration. The remaining steps are the same as in Embodiment 1 and will not be repeated here.

[0071] Step S1: Data Reading: In this embodiment, the data is derived from a structured JSON file exported from pelvic floor ultrasound image annotations. Each subject's data includes only two functional states: resting state (s=1) and Valsalva state (s=2). The read organ contours only include the bladder closure contour and uterine opening curve, excluding the rectum and levator palpebrae superioris. Bony landmarks are read at the lower edge of the pubic symphysis (O). s and sacrococcygeal reference point S s Anatomical landmarks of organs for reading the bladder , BNP, BNA and the uterus , The two-dimensional coordinate system is defined in accordance with that in Example 1.

[0072] Step S2 Coordinate Correction (Unified Bony Coordinate System): This embodiment uses the unified bone coordinate system output mode. The resting state is selected as the reference state, and the lower edge of the pubic symphysis in the resting state is taken as the origin of the unified coordinate system. Calculate the resting state SCIPP orientation angle Set reference correction angle Construct a unified reference transformation matrix: All contour points and landmarks in both the resting and Valsalva states are uniformly transformed to the same bony coordinate system: , s=1,2 In this mode, the contours of the two functional states share the same bony coordinate origin and reference axis, allowing direct comparison of organ displacement and morphological changes in the same subject under resting and Valsalva maneuvers, preserving dynamic pelvic pose information. Compared to the state-corrected version in Example 1, this example does not individually correct the rotation angle for each state, but instead uses the resting state as a reference for unified alignment, making it suitable for analysis scenarios that focus on the dynamic displacement of organs.

[0073] Step S3: Topology identification: The bladder is defined as a closed contour, and the uterus as an open curve. This embodiment does not involve the rectum or levator palpebrae superioris.

[0074] Step S4 Anatomical Landmark Division: The rules for the four divisions of the bladder and the three divisions of the uterus are completely consistent with those in Example 1. The bladder is... To ensure a unified starting point, according to → →BNP→BNA→ The uterus is sequentially divided into four open polyline regions: top, posterior wall, bladder neck and base, and anterior wall; → → → The sequence is divided into three open polyline partitions: the lower anterior wall, the bottom, and the lower posterior wall. The marker matching distance threshold is set to δmark = 1.5 mm in this embodiment (the threshold is appropriately tightened due to the difference in resolution between ultrasound images and MRI).

[0075] Step S5: Equal Arc Length Resampling: The resampling calculation method is the same as in Example 1, but this example uses a different fixed set of sampling point configurations, suitable for ultrasound data scenarios with small sample sizes and short contours. The number of sampling points for each region is set as follows: 40 points each for the top, posterior wall, bladder neck and bottom, and anterior wall of the bladder; 40 points each for the lower segment of the anterior wall, bottom, and lower segment of the posterior wall of the uterus. That is, in this example, the total number of subjects M=7 (4 regions of the bladder + 3 regions of the uterus), with a uniform 40 points for each region, for a total of 280 sampling points. The number of sampling points is also predetermined based on the average arc length ratio of the regions in the ultrasound reference training set, and is used uniformly for all subjects.

[0076] Step S6 Quality Control: The quality control dimensions and scoring formula are consistent with Example 1. Due to the reduced number of organs, the comprehensive quality score Q is still obtained by weighting five individual scores, with the weights adjusted as follows: marker integrity w1=0.30, point order w2=0.25, partition length w3=0.20, curvature stability w4=0.15, and self-intersection and duplicate points w5=0.10. The first threshold T1=80 points, and the second threshold T2=55 points. The remaining parameters (curvature threshold, duplicate point threshold, deduction parameters, etc.) are the same as in Table 1 of Example 1. The sample grading rules and serious error judgment rules are consistent with Example 1.

[0077] Step S7 Output and Medoid Initial Template: For qualified samples, standardized multi-object Polyline data (7 partitions) are output using a unified bony coordinate system. Since this embodiment only includes two functional states, the Medoid initial template selection is performed independently within the resting state group and the Valsalva state group. The multi-object contour distance between n subjects within each group is: Where D m The root mean square distance between corresponding points is used (each partition has 40 points, one-to-one correspondence). The sample with the smallest cumulative distance is selected as the Medoid center, and initial templates for the resting state and Valsalva state are constructed respectively, so that the average deformation between the two states can be calculated.

[0078] This embodiment demonstrates that the method of the present invention can be flexibly adapted to different imaging modalities (MRI / ultrasound), different numbers of functional states (two / three), different organ combinations (all organs / partial organs), and different sampling density configurations, and can output standardized multi-object contour data with stable anatomical correspondences, exhibiting good versatility and scalability.

[0079] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for standardizing the pelvic floor contour through pelvic coordinate correction and landmark point partitioning, characterized in that, Includes the following steps: Step S1: Obtain labeled and exported pelvic floor MRI two-dimensional multi-organ contour data, bony reference landmarks and organ anatomical landmarks for the same subject in at least two functional states. The bony reference landmarks include at least the lower edge of the pubic symphysis and the sacrococcygeal reference point. The multi-organ contour data includes at least the closed contour of the bladder and at least one open contour of the uterus, rectum and levator palpebrae superioris. Step S2: Using the lower edge of the pubic symphysis as the origin and the direction of the line connecting the lower edge of the pubic symphysis and the sacrococcygeal reference point as the pelvic reference direction, establish a pelvic reference coordinate system, calculate the coordinate transformation parameters corresponding to each functional state, and transform the two-dimensional contour point set and organ anatomical landmark points to the pelvic reference coordinate system. Step S3: Determine the topology processing rules for the two-dimensional contours according to the organ type. The two-dimensional contours of the bladder are determined as closed contours, and the two-dimensional contours of the uterus, rectum, and levator palpebrae are determined as open curves. The open curves retain the original start and end points and do not add closed connecting edges from the start to the end point. Step S4: Based on the position of the anatomical landmarks of the organs on the corresponding two-dimensional contours, the bladder closed contour and the uterine open curve are divided into topologically preserved regions. The bladder is divided into four open Polyline regions: the top region, the posterior wall region, the bladder neck and bottom region, and the anterior wall region. The uterus is divided into three open Polyline regions: the lower anterior wall, the bottom of the uterus, and the lower posterior wall. The rectum and levator palpebrae are preserved as independent Polyline objects with their original open curves. The arrangement order of each anatomical region and the shared endpoint relationship between adjacent regions are recorded. Step S5: Calculate the cumulative arc length of each open Polyline partition, and resample each open Polyline partition with equal arc length according to the preset fixed number of sampling points for the corresponding organ and the corresponding anatomical partition, so that the same organs and the same anatomical partitions of different subjects have the same number of sampling points, point order direction and partition label. Step S6: Based on the original two-dimensional contour, organ anatomical landmark matching results, and the resampled open Polyline partition, quality control is implemented. The quality control includes at least landmark integrity check, landmark arrangement order check, partition length check, curvature stability check, and self-intersection and duplicate point check. According to the preset serious error judgment rules and comprehensive quality score results, the corresponding samples are classified into qualified sample set, warning sample set, or failed sample set. Step S7: Compile each open Polyline partition in the qualified sample set into standardized multi-object Polyline data according to organ label, anatomical partition label and sampling point number and output it. Output the data in the warning sample set to the review process and re-execute step S6 after correction to prevent the data in the failed sample set from directly entering the pelvic floor multi-organ deformation template or statistical shape model, and output a quality control report.

2. The method according to claim 1, characterized in that, The coordinate transformation in step S2 is calculated as follows: O is the point at the lower edge of the pubic symphysis. s Let the sacrococcygeal reference point be denoted as S. s Connect the two points to form the SCIPP direction vector U. s =S s -O s Calculate its direction angle θ s =atan2(u s,y , u s,x ); Let the target state correction angle be α. s The effective rotation angle is φ. s =α s -θ s Construct a two-dimensional rotation matrix R s =R(φ s ); Perform coordinate transformation on two-dimensional contour points and organ anatomical landmarks using the following formula: p' s,i =R s (p s,i -O s ); The coordinate transformation supports two output modes: one is the state correction version, which constructs a state transformation matrix for each functional state; the other is the unified bony coordinate version, which selects the reference state of the same subject to construct a unified reference transformation matrix and outputs the contours of the subject in different states to the same bony coordinate system.

3. The method according to claim 1, characterized in that, The specific rules for bladder partitioning in step S4 are as follows: Bladder anatomical landmark A B As a unified starting point, the A B This is the junction between the top of the bladder and the anterior wall, according to A. B →P B P B →BNP, BNP→BNA, and BNA→A B The point sequence direction sequentially forms four open Polyline regions corresponding to the top region, posterior wall region, bladder neck and bottom region, and anterior wall region; Where P B BNP is the junction of the top and posterior wall of the bladder, BNA is the posterior point of the bladder neck, and each landmark is a shared endpoint of adjacent regions. Make A B The corresponding contour point is located at index 0 of the point sequence. Forward and reverse candidate point sequences are constructed respectively, and the unique point that satisfies A is selected. B →P B →BNP→BNA→A B The sequence of anatomical dissections is used as the standard point sequence; if there is no sequence that meets the conditions or if there are multiple sets of sequences that meet the conditions, it is considered a quality control failure.

4. The method according to claim 1, characterized in that, The specific rules for uterine partitioning in step S4 are as follows: Following the point sequence of the uterine opening curve, from the starting point U0 to A U A U →P U and P U →The point sequence direction of endpoint U1 forms three open Polyline partitions corresponding to the lower anterior wall segment, the fundus of the uterus, and the lower posterior wall segment in sequence; Where A U P is the junction of the lower segment of the anterior uterine wall and the fundus of the uterus. U It is the junction of the fundus of the uterus and the lower segment of the posterior uterine wall; When A U or P U If any markers are missing or their order does not match the preset anatomical direction, it will be recorded as a quality control anomaly.

5. The method according to claim 1, characterized in that, Step S4 also includes a matching step between marker points and contour points: When the marker is an independent annotation point, calculate its Euclidean distance d to the i-th point on the contour. i =‖mp i The contour point with the smallest distance is selected as the corresponding point of the marker point; if the smallest distance exceeds the preset matching threshold, it is determined that the marker point matching is abnormal.

6. The method according to claim 1, characterized in that, The calculation method for equal arc length resampling in step S5 is as follows: Let the original ordered contour point sequence of the r-th partition be... The cumulative arc length from the starting point of the partition to the i-th point is The total arc length of this partition is L. r = ; Let N be the number of target resampling points in this partition. r Then the target arc length corresponding to the j-th resampling point is Linear interpolation is performed based on the original arc length interval where the target arc length is located to obtain the coordinates p' of the resampling point. r,j ; For the same anatomical region of the same organ, the same fixed number of sampling points is used in different subjects and under different functional states. The number of sampling points is determined in advance based on the average arc length ratio of the region in the reference training set.

7. The method according to claim 1, characterized in that, In step S6, the overall quality score is obtained by weighting the five individual scores: Q = w1Q mark +w2Q order +w3Q length +w4Q curve +w5Q self Q mark Scoring for marker integrity, Q order For point-order scoring, Q length Scoring based on partition length, Q curve Q is used to score curvature stability. self For self-intersecting and repeated point scores, w1 to w5 are weighting coefficients and satisfy Σw i =1.

8. The method according to claim 7, characterized in that, The sample grading rule in step S6 is as follows: When the quality score Q ≥ the first threshold, it is assigned to the qualified sample set; when the second threshold ≤ Q < the first threshold, it is assigned to the warning sample set; when Q < the second threshold or a preset serious error exists, it is assigned to the failed sample set. The preset serious errors include missing organ landmarks, landmark order not conforming to anatomical rules, incorrect closure of open curves, and self-intersection of contours; when self-intersection exists, regardless of the overall score, it is directly judged as a failed sample.

9. The method according to claim 1, characterized in that, Following step S7, the method further includes a step of selecting an initial deformation template based on qualified standardized data using a Medoid strategy: The multi-object contour distance between any two subjects i and j is Where M is the number of objects, w m Let D be the weight of the m-th object. m The distance to the outline of the corresponding object; The cumulative distance from the i-th candidate to all other candidates in the same group is The subject with the smallest cumulative distance was selected as the Medoid center sample, and all partitions of the sample (Polyline) were used as the initial template.

10. The method according to claim 1, characterized in that, The functional states include at least two of the following: resting state, Valsalva state, and anal contraction state; the preset fixed sampling points for the four open Polyline partitions of the bladder and the three open Polyline partitions of the uterus are all 40 to 60 points, and the preset fixed sampling points for the open curves of the rectum and levator palpebrae superioris are all 150 to 250 points.