Non-invasive assessment of endocardial displacement quantification maps, devices and media

Abstract

The application discloses an endocardial displacement quantification atlas noninvasive evaluation method, device and medium, comprising the following steps: acquiring a cardiac cine magnetic resonance imaging sequence and performing standardization processing, extracting each phase endocardial closure contour, and constructing a starting point, direction and arc length normalized closed curve relying on an anatomical anchor point. Taking the end diastolic contour as a reference, matching the target phase contour corresponding point in the cycle arc length neighborhood, projecting the displacement to the local normal direction of the reference contour, and obtaining the endocardial normal displacement trajectory; according to the clinical segment mapping relationship, generating an endocardial displacement quantification atlas containing displacement intensity, systolic phase and confidence channel. An end-to-end training verification is carried out by using a combined loss function which fuses a multi-class loss and a physiological constraint, and a deployable evaluation system is built. The application can quickly and quantitatively evaluate the endocardial displacement and segment wall motion without contrast agent, and overcomes the defects of the prior art, such as dependence on contrast agent, poor interpretability, unclear reliability, dependence on centroid and fixed angle, etc.

Description

Technical Field

[0001] This invention relates to the fields of medical image processing, quantitative analysis of cardiac function, and computer-aided assessment technology. Specifically, it relates to a non-invasive assessment method, computer equipment, and computer-readable storage medium for endocardial displacement quantification maps generated based on cardiac cine magnetic resonance imaging sequences, anatomical anchor point constraints for endocardial closure contour trajectory modeling, and normal displacement quantification maps. Background Technology

[0002] Myocardial infarction, ischemic myocardial injury, and related myocardial motion disorders can lead to decreased local myocardial contractility, ventricular wall motion asynchrony, and an increased risk of ventricular remodeling. Clinically, a comprehensive assessment is usually conducted combining contrast-enhanced magnetic resonance imaging (MRI), interpretation of ventricular wall motion, overall cardiac function indicators, and physician experience. Contrast-enhanced MRI relies on contrast agents, has a lengthy procedure, and is difficult for patients with contraindications to contrast agents to undergo this type of examination. Non-invasive, rapid, and interpretable quantitative assessment of endocardial motion using standard cardiac cine MRI sequences has clinical value.

[0003] Existing cardiac cine magnetic resonance imaging (MRI) techniques mainly include multi-task segmentation and motion estimation, deformation field estimation based on image registration, radial ray scanning based on the centroid of the ventricle and blood pool, contour analysis based on polar coordinate image unfolding, and end-to-end depth classification. Multi-task segmentation and motion estimation methods typically output segmentation masks and deformation vector fields simultaneously through a shared encoder or latent space, easily leading to black-box judgments dependent on network latent features. Image registration-based techniques typically output pixel-level or voxel-level deformation fields, requiring additional post-processing for clinical segment correspondences. Techniques based on the centroid of the ventricle and blood pool or radial ray scanning rely on fixed centers and angles, making them susceptible to overall cardiac translation, sectional tilt, local non-radial motion, and inconsistent anatomical origins.

[0004] To reduce the risk of similarity to the aforementioned technical approaches and improve the verifiability of the results, this invention does not use a deformable vector field as the core output, does not employ a parallel segmentation and motion estimation architecture with a shared latent space, does not use the centroid of the blood pool as the radial ray scanning center, and does not simply unfold the endocardial contour into a polar coordinate image with a fixed angle as the horizontal axis. This invention represents the endocardial contour as a closed curve with anatomical anchor points, curve direction, and period arc length coordinates. It establishes phase correspondences within the period arc length neighborhood of the closed curve and projects the corresponding displacements onto the local in-normal direction of the end-diastolic reference contour, further generating a three-channel endocardial displacement quantification map containing displacement intensity, systolic phase, and confidence level. Summary of the Invention

[0005] The purpose of this invention is to provide a non-invasive assessment method, device, and medium for quantifying endocardial displacement atlases, in order to solve the technical problems in the prior art, such as insufficient interpretability of end-to-end motion estimation models, strong dependence of centroid-based radial scanning on geometric centers, difficulty in stably mapping deformation field-based outputs to clinical segment interpretation systems, and unclear sources of atlas reliability.

[0006] To achieve the above objectives, this invention provides a non-invasive assessment method for endocardial displacement quantification atlases, which is executed by a computer device. The method generates endocardial displacement quantification atlases through standardization processing, endocardial closure contour extraction, anatomical anchor point calibration, periodic arc length correspondence, reference contour normal displacement projection, segmental robust convergence, and reliability verification.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A non-invasive assessment method for quantifying endocardial displacement involves the following steps performed using a computer device. S1. Obtain cardiac cine magnetic resonance imaging (MRI) sequences and perform normalization processing on the cardiac cine MRI sequences to obtain normalized phase image sequences. S2. Perform endocardial probability estimation and closed contour extraction on each phase image in the standardized phase image sequence, and convert the endocardial closed contour corresponding to each phase image into a closed curve with the same contour starting point, the same contour direction and normalized arc length coordinates based on anatomical anchor points. S3. Using the closed curve corresponding to the end-diastolic phase as a reference contour, determine the correspondence between the corresponding point on the target phase closed curve and the reference point on the reference contour within the periodic arc length neighborhood of the normalized arc length coordinates, and project the displacement of the corresponding point relative to the reference point onto the local in-normal direction of the reference point to obtain the endocardial normal displacement trajectory. S4. Calculate point-level and segment-level credibility based on endocardial probability, contour correspondence residuals, temporal smoothness, and physiological range constraints. S5. According to the preset clinical segment mapping relationship, the endocardial normal displacement trajectory is segmentally converged to generate an endocardial displacement quantification map including displacement intensity channel, systolic phase channel and confidence channel, and motion abnormality auxiliary prompts are generated based on segment displacement value, systolic phase and the segment confidence. S6. Output the endocardial displacement quantification map, segment displacement values, segment reliability, and motion abnormality auxiliary prompts.

[0008] Further, step S1 includes: acquiring multiple phase images of the same examination object on at least one short-axis plane that cover the cardiac cycle; performing pixel spacing resampling, image matrix cropping or padding, left ventricular region localization, and orientation unification processing on the phase images to obtain resampled images; normalizing the intensity of the resampled images based on the mean and standard deviation of pixel intensity within the cardiac neighborhood mask to obtain the standardized phase image sequence; and during the training phase, classifying all phase images and all plane images of the same examination object into the same data subset.

[0009] Further, step S2 includes: performing endocardial probability estimation on each standardized phase image to obtain an endocardial region probability map; obtaining a binary endocardial region based on the endocardial region probability map and a preset endocardial probability threshold; performing connected component filtering, hole filling, and boundary smoothing on the binary endocardial region to obtain an endocardial closed contour; resampling the endocardial closed contour into N ordered contour points; detecting or inferring anatomical anchor points, using the anatomical anchor points as the contour starting point, and reordering the N ordered contour points according to a preset anatomical direction to obtain the closed curve.

[0010] Furthermore, the anatomical anchor point is determined by the right ventricular insertion point, the interventricular septum reference point, the relative position point of the papillary muscle, or the starting point of the equivalent curve formed by the above structures; the end-diastolic phase and the end-systolic phase are determined according to the enclosed area of ​​each closed curve, wherein the phase with the largest enclosed area is the end-diastolic phase, and the phase with the smallest enclosed area is the end-systolic phase.

[0011] Further, step S3 includes: selecting multiple reference points on the normalized arc length coordinates of the reference contour; calculating the unit tangent vector of each reference point based on adjacent reference points, and adjusting the candidate normal vector obtained by rotating the unit tangent vector to point in the direction of the left ventricular cavity to obtain the unit in-cell normal vector; constructing a periodic arc length search window centered on the normalized arc length coordinates of the target reference point, and handling arc length wrapping by utilizing the adjacency relationship of the beginning and end of the closed curve when the periodic arc length search window crosses the endpoints of the normalized arc length coordinates; and determining the corresponding point on the target phase closed curve within the periodic arc length search window.

[0012] Further, determining the corresponding point within the periodic arc length search window includes: determining the corresponding point within the periodic arc length search window that minimizes the weighted cost function based on a weighted cost function of the spatial distance between contour points and the reasonableness of displacement amplitude; calculating the projection of the displacement vector between the corresponding point and the corresponding reference point onto the unit normal vector of the corresponding reference point to obtain the original normal displacement; performing time consistency correction on the original normal displacement to obtain the corrected normal displacement; and calculating the point-level displacement amplitude, the point-level contraction peak phase, and the point-level rebound index based on the corrected normal displacement.

[0013] Further, step S4 includes: calculating the point-level confidence level based on point-level boundary uncertainty, point-level contour corresponding residual, point-level time non-smoothing penalty, and point-level physiological range penalty; weighting and converging the endocardial normal displacement trajectory based on the point-level confidence level of multiple points within the segment to obtain the segmental end-systolic displacement, segmental displacement amplitude, and segmental systolic peak phase; standardizing the segmental end-systolic displacement, segmental displacement amplitude, and segmental systolic peak phase based on at least one of population reference statistics and individual historical reference information; mapping the standardized segmental displacement amplitude to the displacement intensity channel, mapping the standardized segmental systolic peak phase to the systolic phase channel, and mapping the segment confidence level to the confidence level channel to obtain the atlas elements corresponding to each clinical segment.

[0014] Furthermore, the generation of motion anomaly auxiliary prompts in step S5 includes: generating a review prompt when the segment confidence is lower than a preset confidence threshold; generating a contraction anomaly prompt when the segment confidence is not lower than the preset confidence threshold and the segment contraction end displacement is lower than a preset displacement threshold; generating a contraction asynchrony prompt when the segment confidence is not lower than the preset confidence threshold and the segment contraction peak phase is later than a preset contraction phase threshold; and generating a no-anomaly prompt in other cases.

[0015] Furthermore, step S5 also includes: rendering the atlas elements corresponding to each clinical segment onto a standard segment base map, wherein the color of each clinical segment is determined by the displacement intensity channel, the transparency or border style of each clinical segment is determined by the confidence channel, and labeling each clinical segment with at least one of the segment end-contraction displacement, the segment displacement amplitude, the segment contraction peak phase, and the segment confidence; and generating an assessment report based on the atlas elements, segment displacement values, segment confidence, and motion abnormality auxiliary prompts of each clinical segment.

[0016] Furthermore, during the training phase, a training sample set is constructed, including standardized cardiac cine magnetic resonance imaging sequences, endocardial reference masks, anatomical anchor reference information, and clinical reference labels or segmental motion scores. A joint loss function is used to train endocardial probability estimation, anatomical anchor localization, and motion anomaly assistance cues. This joint loss function includes endocardial probability estimation loss, anatomical anchor localization loss, anomaly assistance cues classification loss, temporal smoothing loss, and physiological range constraint loss. During the application phase, it is not necessary to input endocardial reference masks, anatomical anchor reference information, clinical reference labels, or segmental motion scores.

[0017] As can be seen from the above technical solution, this invention discloses a non-invasive assessment method, device, and medium for endocardial displacement quantification atlases. The method first acquires cardiac cine magnetic resonance imaging sequences and performs standardization processing; extracts the endocardial closed contours of each phase, and establishes closed curves with the same starting point, direction, and normalized arc length coordinates based on anatomical anchor points; using the end-diastolic contour as a reference, determines the corresponding point of the target phase contour within the periodic arc length neighborhood, and projects the displacement onto the local in-normal direction of the reference contour point to obtain the endocardial normal displacement trajectory; generates an endocardial displacement quantification atlas containing displacement intensity channels, systolic phase channels, and confidence channels according to a preset clinical segment mapping relationship; and completes end-to-end training and validation using a joint loss function that integrates endocardial probability estimation loss, anatomical anchor point localization loss, abnormal auxiliary cue classification loss, temporal smoothing loss, and physiological range constraint loss to construct a deployable assessment system. This invention eliminates the need for contrast agents and enables rapid, quantitative, and visual assessment of endocardial displacement and segmental ventricular wall motion. It addresses the problems of existing technologies, such as reliance on contrast agents, insufficient interpretability, unclear sources of image reliability, and strong dependence on centroid and fixed angles. It provides a safe and convenient quantitative method for early screening, accurate assessment, and auxiliary verification of myocardial infarction and ischemic cardiomyopathy.

[0018] This invention presents a non-invasive assessment method for endocardial displacement quantification atlases. Through endocardial probability estimation, anatomical anchor calibration, and periodic arc length coordinate modeling, it deeply mines endocardial motion information in cardiac cine sequences, improving the interpretability and clinical acceptability of myocardial motion quantification. After image standardization, probabilistic map-based closed contour extraction and anatomical anchor redirection ensure that contours at each phase and level have the same starting point, direction, and normalized arc length coordinates, laying a spatial consistency foundation for periodic neighborhood correspondence. Using the local inward normal of the end-diastolic reference contour as the displacement projection direction, combined with the periodic arc length search window and the cost function of fusion edge probability, it stably captures the true normal displacement trajectory under contrast agent-free conditions, significantly reducing the correspondence errors caused by overall cardiac translation, sectional tilt, and non-radial motion. Temporal consistency correction and physiological range constraints effectively suppress contour noise and abnormal beating, improving the physiological rationality of the displacement trajectory. Segmental convergence introduces point-level credibility weighting, integrating population and individual references to generate three-channel maps of displacement intensity, contraction phase, and credibility. This solves the problem of traditional deformation fields being difficult to stably map to a 17-segment interpretation system, and provides a clear source of credibility for each map. During the training phase, temporal smoothing loss and physiological range constraint loss are used to guide the model to converge quickly based on physical laws, improving the accuracy of displacement trajectories. After rigorous validation and comparison with traditional methods, this invention demonstrates superior motion quantification performance and segmental anomaly indication capabilities. This not only overcomes the problems of poor interpretability, dependence on the blood pool centroid, and susceptibility to geometric shifts in existing models, but also achieves non-invasive assessment of endocardial displacement without contrast agents and with verifiability, providing a safe, efficient, and traceable quantitative tool for the diagnosis and follow-up of ventricular wall motion abnormalities. Attached Figure Description

[0019] Figure 1 This is a flowchart of the non-invasive assessment method for quantifying endocardial displacement according to an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of the system module structure according to an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram illustrating the establishment of the anatomical anchor point constraint contour coordinates in an embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram illustrating the calculation of endocardial displacement based on the periodic arc length window and the normal direction in an embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram of a three-channel endocardial displacement quantification map according to an embodiment of the present invention.

[0024] Figure 6 This is a schematic diagram of the segment reliability and motion abnormality auxiliary prompt report in an embodiment of the present invention.

[0025] Figure 7 This is a comparison chart (polar coordinate form) of the predicted myocardial infarction area in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0027] like Figure 1 and Figure 2 As shown in the figure, this embodiment of the invention provides a non-invasive rapid assessment system for cardiac function generated by a non-invasive assessment method for endocardial displacement quantification maps. The specific steps are as follows: S1. Obtain cardiac cine magnetic resonance imaging sequences and perform standardization processing to obtain standardized phase image sequences (related dataset, with permission from the parties involved, the related data is only used for research purposes); S2. For each phase image in the standardized phase image sequence, perform endocardial probability estimation and closed contour extraction, and convert the endocardial closed contour of each phase into a closed curve with the same starting point, the same direction and normalized arc length coordinates based on anatomical anchor points. S3. Using the endocardial closure contour corresponding to the end-diastolic phase as a reference contour, determine the correspondence between the target phase contour point and the reference contour point in the periodic arc length neighborhood of the normalized arc length coordinate, and project the displacement of the corresponding point relative to the reference contour point onto the local in-normal direction of the reference contour point to obtain the endocardial normal displacement trajectory. S4. According to the preset clinical segment mapping relationship, the endocardial normal displacement trajectory is segmentally converged to generate an endocardial displacement quantification map including displacement intensity channel, systolic phase channel and confidence channel; S5. Calculate segment reliability based on endocardial probability, contour correspondence residual, temporal smoothness and physiological range constraints, and generate motion abnormality auxiliary prompts based on segment displacement values, contraction phase and segment reliability. S6. Output the endocardial displacement quantification map, segment displacement values, segment reliability, and motion abnormality auxiliary prompts.

[0028] Definitions of terms and symbols: To avoid the same symbol representing different meanings, the following uniform symbols are used in this specification: This represents the total number of phases in the cardiac cycle; Indicates the phase number; Indicates the level number; This represents the set of levels involved in the analysis; This indicates the number of resampling points for each endocardial closure contour; Indicates the outline point number; Indicates the clinical segment number; This represents a probability map of the endocardial region. Indicates a binary endocardial region; Indicates the closed outline of the endocardium; Represents the normalized arc length coordinates; Indicates the reference contour point at the end of diastole; Indicates the point corresponding to the target phase; Represents the unit tangent vector; Represents the unit normal vector; Indicates the original normal displacement; This represents the corrected normal displacement; Indicates point-level confidence level; Indicates segment reliability; Indicates the displacement at the end of segmental contraction; Indicates the segmental displacement amplitude; Indicates the peak phase of segmental contraction; Indicates the first The spectral elements of each segment; This represents the total training loss.

[0029] Parameter description: Each parameter in this invention has a defined range of values ​​and can be optimized and determined on a training or validation set, and is not limited to the following example values. In one embodiment, the parameters are selected from the range shown in the table below, with preferred values ​​as follows:

[0030] The following is a detailed explanation: Step 1: Image sequence acquisition and normalization processing The specific process is as follows: Step 1.1: Acquire multiple cardiac cycle phase images of the same subject at at least on at least one short-axis plane, denoted as: , in, Indicates the phase number, Indicates the total number of phases. Indicates the layer number, This represents the set of levels involved in the analysis; Step 1.2: Process the image Pixel spacing resampling, image matrix cropping or padding, left ventricular region localization, and orientation alignment processing are performed to obtain the resampled image. , , in, This represents the interpolation resampling function based on the target pixel spacing. and Representing direction and Target pixel spacing in the direction; Step 1.3: Resample the image Intensity normalization is performed. For each image, a cardiac neighborhood mask is selected. Calculate the average value of pixels within the range and standard deviation A standardized phase image is obtained. , ; Step 1.4: During the training phase, obtain reference information corresponding to the sample. The reference information includes at least one of endocardial reference mask, anatomical anchor reference information, segmental wall motion score, and overall cardiac function classification. During the application phase, the reference information is not required input. Step 1.5: Divide the training samples into training set, validation set, and test set. If the same object to be inspected contains multiple layers or multiple phase images, then all images of the same object to be inspected are assigned to the same data subset.

[0031] Step 2: Endocardial probability estimation and closed contour extraction with anatomical anchor point constraints like Figure 3 As shown, the specific process is as follows: Step 2.1: For each normalized phase image Endocardial probability estimation was performed to obtain an endocardial region probability map. ,in, The endocardial probability estimation module is only used to acquire the contour and does not output the motion deformation vector field; Step 2.2: Based on the aforementioned endocardial region probability map Obtain binary endocardial region .

[0032] ; Step 2.3: For the binary endocardial region By performing connected component filtering, hole filling, and boundary smoothing, the endocardial closure contour is obtained. ; Step 2.4: Close the endocardial contour. Resampling Let be an ordered contour point, denoted as . , in, ; Step 2.5: Detect or infer anatomical anchor points The anatomical anchor points are determined by the right ventricular insertion point, the interventricular septum reference point, the relative position of the papillary muscles, or the starting point of the equivalent curve formed by the above structures. , in, Indicates the candidate coordinates of the anchor point. Indicates the anatomical anchor points projected onto the closure contour of the endocardium; Step 2.6: Using the anatomical anchor points As the starting point of the closed contour, the contour points are reordered according to the preset anatomical direction, and the normalized arc length coordinates are calculated. , , , , Among them, when hour ,when hour ; Step 2.7: Determine the end-diastolic phase and end-systolic phase based on the area enclosed by the endocardial closure contour. The endocardial closure contour... The area enclosed is denoted as , , .

[0033] Step 3: Calculation of normal displacement trajectory based on the neighborhood of the periodic arc length like Figure 4 As shown, the specific process is as follows: Step 3.1: For layer l, with the end-diastolic phase The corresponding endocardial closure contour is used as a reference contour, and the normalized arc length coordinates are selected. One reference point: , in, ; Step 3.2: Calculate the reference point unit tangent vector and unit normal vector , Equation (7): , ; Step 3.3: For the target phase , at the target phase endocardial closure contour above Establish a periodic arc length search window centered on the arc. , , , when At that time, the periodic arc length search window is composed of intervals. and interval Composition; when At that time, the periodic arc length search window is composed of intervals. and interval Composition; when and At that time, the periodic arc length search window is an interval ; Step 3.4: In the periodic arc length search window Internally determine the corresponding point , , , , in, The solution is found in the period arc length search window. The process is conducted internally, and the computer equipment is no larger than [a certain size]. arc length step size Sampling, or using a one-dimensional bounded search method to determine... Minimum arc length coordinates; both methods preserve The defined periodic distance remains unchanged; in, This represents the periodic arc length coordinates on the endocardial closure contour of the target phase. Contour points obtained by linear interpolation; when When the value is close to 0 or 1, the interpolation point is determined by the two adjacent points at the beginning and end of the closed curve. , and The three values ​​are the search weights for the corresponding points, and they belong to the range [0,1] and their sum is 1. , representing the equivalent pixel spacing of layer l; Step 3.5: Calculate the reference point In the target phase normal displacement on , ; Step 3.6: Adjust the normal displacement Perform time consistency correction to obtain the corrected normal displacement. , , in, The time smoothing weight has a value range of 0 to 10; Step 3.7: Based on the corrected normal displacement Calculate the displacement amplitude at the point level Point-level contraction peak phase and point-level rebound index , , , , .

[0034] Step 4: Clinical segment convergence and generation of three-channel endocardial displacement quantification map like Figure 5 As shown, the specific process is as follows: Step 4.1: Establish a segment mapping table: , in, Indicates the first The set of contour points corresponding to each clinical segment , in, For the first The set of layers corresponding to each clinical segment and The arc length boundary is determined by the anatomical anchor point; Step 4.2: Based on point-level confidence Weighted convergence of points within the segment yields the displacement at the end of segmental contraction. Segmental displacement amplitude and the peak phase of segmental contraction , , , , , in, Represents the imaginary unit. and They represent the real part and the imaginary part, respectively. Indicates to Take the mold; Step 4.3: Calculate the segmental standardized displacement fraction , , ， ， , , in, ; Step 4.4: Generate a three-channel endocardial displacement quantification map The graph includes a displacement intensity channel, a contraction phase channel, and a confidence channel. ; Step 4.5: Transfer the spectral elements Rendered onto a standard segment base map. The color of each segment is determined by the displacement intensity channel, while the segment transparency or border is determined by the confidence channel. Segment numerical annotations include... , , and At least one of them, , in, Indicates the color level threshold. Indicates the first Each color level This indicates the total number of color levels.

[0035] Step 5: Segment Reliability Calculation and Motion Anomaly Auxiliary Hint Generation like Figure 6 As shown, the specific process is as follows: Step 5.1: Calculate point-level boundary uncertainty Point-level corresponding residuals Point-level time unsmoothing penalty and point-level physiological range punishment , , , , ; Step 5.2: Calculate point-level confidence level Segment credibility , , when When not empty, ;when When empty, ; Step 5.3: Based on segment credibility Segmental contraction at the end of the displacement and the peak phase of segmental contraction Generate segmental auxiliary prompts , like ,but Review prompt; like and ,but Error message; like and ,but Error message; In other cases, No abnormality was detected; Step 5.4: Based on each segment An assessment report is generated, which includes a quantitative atlas of endocardial displacement, segmental displacement values, segmental reliability, segments requiring review, and auxiliary prompts for motion abnormalities.

[0036] Step Six: Model Training, Evaluation, and System Deployment The specific process is as follows: Step 6.1: Construct the training sample set: , in, Indicates the first For example, a standardized cardiac cine magnetic resonance imaging sequence. Indicates the first For example, endocardial reference mask. Indicates the first Example anatomical anchor point reference information, Indicates the first Examples include clinical reference labels or segmental motion scores; Step 6.2: When training the endocardial probability estimation module, use the endocardial probability estimation loss. , , in, For pixels Reference tags, For pixels The predicted probability, The number of pixels involved in the calculation; Step 6.3: When training the anatomical anchor point detection module, use the anatomical anchor point localization loss. , , in, Indicates the first One reference anchor point, Indicates the first One prediction anchor point, Indicates the number of anchor points; Step 6.4: When training the motion anomaly assistance cue model, use the anomaly assistance cue classification loss. , , in, Indicates the first Example sample reference label, Indicates the first Example sample anomaly probability; Step 6.5: When training or calibrating the displacement trajectory, use time-smoothing loss. and physiological range constraint loss , .

[0037] , in, Indicates the number of layers involved in the analysis; Step 6.6: Total System Loss Function The weighted sum of the losses of each subtask. ; Step 6.7: Optimize parameters using the training set, select model parameters or thresholds using the validation set, and evaluate system performance using the test set. Evaluation metrics include endocardial contour extraction accuracy, anchor point positioning error, consistency between segment displacement and reference interpretation, accuracy of auxiliary prompts, recall, F1 score, and segment confidence distribution. Step 6.8: During deployment, the computer device receives the cardiac cine magnetic resonance imaging sequence to be evaluated, executes steps 1 to 5, and outputs the endocardial displacement quantification map, segment displacement values, segment confidence, segments requiring review, and motion abnormality auxiliary prompts.

[0038] Step 7: When applying the system, input the cardiac cine magnetic resonance imaging sequence into the final non-invasive rapid cardiac function assessment system, and output the endocardial displacement quantification map after calculation.

[0039] For example: like Figure 7 As shown, Sample 1 and Sample 2 are the MRI images of two patients, respectively. The left column shows the gadolinium contrast-enhanced images, where the highlighted areas clearly show the necrotic myocardium; the middle column shows the myocardial infarction areas marked by manual delineation and confirmation by professional cardiac radiologists based on the gadolinium contrast-enhanced images, serving as a reference standard for evaluation; the right column shows the myocardial infarction area prediction results obtained by the model proposed in this embodiment of the invention based on the conventional cardiac cine MRI sequence before the patient's contrast agent injection, with the results output in polar coordinates.

[0040] In summary, the embodiments of this invention provide a non-invasive assessment method, device, and medium for endocardial displacement quantification atlases. The method first acquires cardiac cine magnetic resonance imaging sequences and performs standardization processing; then, endocardial probability estimation is performed on each phase image, and endocardial closed contours with the same starting point, direction, and normalized arc length coordinates are extracted based on anatomical anchor points; using the end-diastolic contour as a reference contour, corresponding points between the target phase contour and the reference contour are determined within the periodic arc length neighborhood of the normalized arc length coordinates, and the corresponding displacement is projected onto the local in-normal direction of the reference contour point to obtain the endocardial normal displacement trajectory; subsequently, the normal displacement trajectory is segmentally converged according to a preset clinical segment mapping relationship to generate an endocardial displacement quantification atlas containing displacement intensity channels, systolic phase channels, and confidence channels; segment confidence is calculated based on endocardial probability, contour correspondence residuals, temporal smoothness, and physiological range constraints, and motion abnormality auxiliary prompts are generated based on segment displacement values, systolic phases, and segment confidence. During the training phase, a joint loss function integrating endocardial probability estimation loss, anatomical anchor point localization loss, abnormal auxiliary cue classification loss, temporal smoothing loss, and physiological range constraint loss was used to complete end-to-end training and validation, constructing a deployable assessment system. This invention eliminates the need for contrast agents, enabling rapid, quantitative, and visual assessment of endocardial displacement and segmental wall motion. It solves the problems of existing technologies, such as reliance on contrast agents, insufficient interpretability, unclear sources of atlas reliability, and strong dependence on the centroid and fixed angles. It provides a safe and convenient quantitative method for early screening, accurate assessment, and auxiliary verification of myocardial infarction and ischemic cardiomyopathy.

[0041] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0042] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0043] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the non-invasive assessment methods for endocardial displacement quantification atlases described in the above embodiments.

[0044] It is understood that the system provided in the embodiments of the present invention corresponds to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

[0045] This application also provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, communication interface, and memory communicate with each other via the communication bus. Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements the above-mentioned non-invasive assessment method of endocardial displacement quantification map.

[0046] The communication bus mentioned in the aforementioned electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc.

[0047] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0048] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0049] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0050] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0052] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0053] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A non-invasive assessment method for quantifying endocardial displacement, characterized in that, Includes the following steps: S1. Obtain cardiac cine magnetic resonance imaging (MRI) sequences and perform normalization processing on the cardiac cine MRI sequences to obtain normalized phase image sequences. S2. Perform endocardial probability estimation and closed contour extraction on each phase image in the standardized phase image sequence, and convert the endocardial closed contour corresponding to each phase image into a closed curve with the same contour starting point, the same contour direction and normalized arc length coordinates based on anatomical anchor points. S3. Using the closed curve corresponding to the end-diastolic phase as a reference contour, determine the correspondence between the corresponding point on the target phase closed curve and the reference point on the reference contour within the periodic arc length neighborhood of the normalized arc length coordinates, and project the displacement of the corresponding point relative to the reference point onto the local in-normal direction of the reference point to obtain the endocardial normal displacement trajectory. S4. Calculate point-level and segment-level credibility based on endocardial probability, contour correspondence residuals, temporal smoothness, and physiological range constraints. S5. According to the preset clinical segment mapping relationship, the endocardial normal displacement trajectory is segmentally converged to generate an endocardial displacement quantification map including displacement intensity channel, systolic phase channel and confidence channel, and motion abnormality auxiliary prompts are generated based on segment displacement value, systolic phase and the segment confidence. S6. Output the endocardial displacement quantification map, segment displacement values, segment reliability, and motion abnormality auxiliary prompts.

2. The non-invasive assessment method for quantifying endocardial displacement according to claim 1, characterized in that, Step S1 includes: acquiring multiple phase images of the same examination subject at at least one short-axis plane that cover the cardiac cycle; performing pixel spacing resampling, image matrix cropping or padding, left ventricular region localization, and orientation unification processing on the phase images to obtain resampled images; normalizing the intensity of the resampled images based on the mean and standard deviation of pixel intensity within the cardiac neighborhood mask to obtain the standardized phase image sequence; and during the training phase, classifying all phase images and all plane images of the same examination subject into the same data subset.

3. The non-invasive assessment method for quantifying endocardial displacement according to claim 1, characterized in that, Step S2 includes: performing endocardial probability estimation on each standardized phase image to obtain an endocardial region probability map; obtaining a binary endocardial region based on the endocardial region probability map and a preset endocardial probability threshold; performing connected component filtering, hole filling, and boundary smoothing on the binary endocardial region to obtain an endocardial closed contour; resampling the endocardial closed contour into N ordered contour points; detecting or inferring anatomical anchor points, using the anatomical anchor points as the contour starting point, and reordering the N ordered contour points according to a preset anatomical direction to obtain the closed curve.

4. The non-invasive assessment method for quantifying endocardial displacement according to claim 3, characterized in that, The anatomical anchor points are determined by the right ventricular insertion point, the interventricular septum reference point, the relative position point of the papillary muscles, or the starting point of the equivalent curve formed by the above structures; the end-diastolic phase and the end-systolic phase are determined according to the enclosed area of ​​each closed curve, wherein the phase with the largest enclosed area is the end-diastolic phase, and the phase with the smallest enclosed area is the end-systolic phase.

5. The non-invasive assessment method for quantifying endocardial displacement according to claim 1, characterized in that, Step S3 includes: selecting multiple reference points on the normalized arc length coordinates of the reference contour; calculating the unit tangent vector of each reference point based on adjacent reference points, and adjusting the candidate normal vector obtained by rotating the unit tangent vector to point in the direction of the left ventricular cavity to obtain the unit in-cell normal vector; constructing a periodic arc length search window centered on the normalized arc length coordinates of the target reference point, and handling arc length wrapping by utilizing the adjacency relationship of the beginning and end of the closed curve when the periodic arc length search window crosses the endpoints of the normalized arc length coordinates; and determining the corresponding point on the target phase closed curve within the periodic arc length search window.

6. The non-invasive assessment method for quantifying endocardial displacement according to claim 5, characterized in that, Determining the corresponding point within the periodic arc length search window includes: determining the corresponding point within the periodic arc length search window that minimizes the weighted cost function based on a weighted cost function of the spatial distance between contour points and the reasonableness of displacement amplitude; calculating the projection of the displacement vector between the corresponding point and the corresponding reference point onto the unit normal vector of the corresponding reference point to obtain the original normal displacement; performing time consistency correction on the original normal displacement to obtain the corrected normal displacement; and calculating the point-level displacement amplitude, the point-level contraction peak phase, and the point-level rebound index based on the corrected normal displacement.

7. The non-invasive assessment method for quantifying endocardial displacement according to claim 1, characterized in that, Step S4 includes: calculating the point-level confidence level based on point-level boundary uncertainty, point-level contour corresponding residual, point-level time non-smoothing penalty, and point-level physiological range penalty; weighting and converging the endocardial normal displacement trajectory based on the point-level confidence level of multiple points within the segment to obtain the segmental end-systolic displacement, segmental displacement amplitude, and segmental systolic peak phase; standardizing the segmental end-systolic displacement, segmental displacement amplitude, and segmental systolic peak phase based on at least one of population reference statistics and individual historical reference information; mapping the standardized segmental displacement amplitude to the displacement intensity channel, mapping the standardized segmental systolic peak phase to the systolic phase channel, and mapping the segment confidence level to the confidence level channel to obtain the atlas elements corresponding to each clinical segment.

8. The non-invasive assessment method for quantifying endocardial displacement according to claim 7, characterized in that, Step S5 further includes: rendering the atlas elements corresponding to each clinical segment onto a standard segment base map, wherein the color of each clinical segment is determined by the displacement intensity channel, the transparency or border style of each clinical segment is determined by the confidence channel, and labeling each clinical segment with at least one of the segment end-contraction displacement, the segment displacement amplitude, the segment contraction peak phase, and the segment confidence; and generating an assessment report based on the atlas elements, segment displacement values, segment confidence, and motion abnormality auxiliary prompts of each clinical segment.

9. A computer device, characterized in that, It includes a processor and a memory, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the non-invasive assessment method for endocardial displacement quantification atlas according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to perform the non-invasive assessment method for endocardial displacement quantification atlas according to any one of claims 1 to 8.

Images