Abnormality detection method and system for cardiac ultrasound data

Abstract

The invention discloses an anomaly detection method and system for cardiac ultrasonic data, and relates to the field of ultrasonic data monitoring, and the method comprises the steps: obtaining an inter-frame dynamic feature value and an image embedding feature value of each frame; identifying each systolic-diastolic fluctuation period in the original heart B ultrasonic image sequence; determining whether to enter a rhythm abnormality judgment process or not based on the structural period deviation judgment result; and implementing corresponding operation based on the rhythm abnormality judgment result. According to the invention, by fusing dual-channel feature extraction and an intelligent period recognition mechanism, and adopting a two-stage strategy of structural deviation and rhythm fluctuation to carry out anomaly judgment, the rhythm stability and the anomaly recognition accuracy are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of ultrasonic data monitoring, and in particular to a method and system for detecting abnormalities in cardiac ultrasonic data. Background Art

[0002] As a non-invasive, real-time imaging medical examination method, cardiac ultrasound has been widely used for structural assessment and functional diagnosis of the cardiovascular system. By observing the dynamic changes of various cardiac structures during contraction and relaxation, the current clinical interpretation of cardiac ultrasound images still relies mainly on manual experience. Existing technologies often focus on anatomical changes, but lack analysis of time series features such as motion rhythm and cycle stability, making it difficult to identify subtle rhythm fluctuations. When comparing different contraction-diastole cycles, existing systems have difficulty achieving standardized alignment and difference measurement of inter-frame image embedding features, resulting in inaccurate judgment of cycle consistency. In the presence of mild structural deviations or local rhythm disorders, traditional systems have difficulty effectively identifying, labeling, or automatically correcting. In summary, existing cardiac ultrasound data analysis methods still have significant shortcomings. There is an urgent need for an automated anomaly detection method and system that integrates image structure embedding, motion dynamic feature analysis, and periodic rhythm modeling to improve the intelligent interpretation level of cardiac ultrasound data and the efficiency of diagnostic assistance.

[0003] Existing technologies, such as the invention patent with the public account CN114469176B, are a method and related device for detecting fetal heart ultrasound images. The method includes obtaining a multimodal echocardiogram; determining the fetal heart morphological structure, fetal heart rhythm indicators and fetal heart atrioventricular movement patterns based on the obtained echocardiogram; and determining the type of fetal arrhythmia based on the fetal heart morphological structure, fetal heart rhythm indicators and fetal heart atrioventricular movement patterns.

[0004] Existing technology, such as the invention patent with the public number CN107847173B, is an ultrasound sequencing system and method, including a catheter configured for delivery to a body cavity defined by surrounding tissue; multiple ultrasound transducers connected to the distal end of the catheter; and an electronic module configured to selectively turn on / off each ultrasound transducer according to a predetermined activation sequence, and process the signals received from each ultrasound transducer to generate at least a 2D display of the surrounding tissue. Users can selectively calculate and display various aspects of cardiac activity. Users can display dipole density (DDM), charge density (CDM), or voltage (VV). The shape and position of the chamber (surface) can be displayed, as well as the potential recorded on the electrodes.

[0005] Existing technologies in ultrasound data monitoring and processing focus on extracting standard slices using variational autoencoders and semi-supervised learning. However, these technologies focus solely on the static features of ultrasound images and lack the exploration of temporal relationships between frames, making them incapable of detecting anomalies due to cyclical cardiac changes. Furthermore, existing technologies analyze ultrasound images at a macroscopic level and are unable to identify subtle rhythm abnormalities. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the present invention provides a method and system for detecting abnormalities in cardiac ultrasound data. To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for detecting abnormalities in cardiac ultrasound data, comprising:

[0007] A dual-channel feature extraction and detection is performed on the original cardiac B-ultrasound image sequence to obtain the inter-frame dynamic feature value and image embedding feature value of each frame in the original cardiac B-ultrasound image sequence.

[0008] Each systolic-diastolic fluctuation cycle in the original cardiac B-ultrasound image sequence is identified based on the inter-frame dynamic features and image embedding feature values ​​of each frame.

[0009] The mean value of the structural alignment error of each systolic-diastolic fluctuation cycle is obtained, and the structural cycle deviation judgment result of each systolic-diastolic fluctuation cycle is obtained. Based on the structural cycle deviation judgment result, it is decided whether to enter the rhythm abnormality judgment process.

[0010] The image embedding feature parameters of each systolic-diastolic fluctuation cycle are analyzed, and the rhythm abnormality judgment results of each systolic-diastolic fluctuation cycle are obtained after introducing the influence of the structural cycle deviation judgment results. Corresponding operations are performed based on the rhythm abnormality judgment results.

[0011] As a preferred technical solution, the inter-frame dynamic feature value and image embedding feature value of each frame in the original cardiac B-ultrasound image sequence are obtained. The specific process is as follows:

[0012] In the inter-frame dynamic channel, the motion information parameters of each frame in the original cardiac B-ultrasound image sequence are extracted, including the local optical flow mean, boundary displacement intensity, frame difference energy density and displacement direction mutation rate. The unit weight factors corresponding to the motion information parameters are extracted from the database, including the local optical flow mean unit weight factor, boundary displacement intensity unit weight factor, frame difference energy density unit weight factor and displacement direction mutation rate unit weight factor. The motion information parameters of each frame in the original cardiac B-ultrasound image sequence are corrected and coupled with the corresponding unit weight factors to obtain the inter-frame dynamic feature values ​​of each frame in the original cardiac B-ultrasound image sequence.

[0013] In the image embedding feature channel, the embedding information parameters of each frame in the original B-ultrasound sequence are extracted, including the embedding norm, characteristic entropy and principal component projection amplitude. The unit weight factors corresponding to the embedding information parameters are extracted from the database, including the embedding norm unit weight factor, characteristic entropy unit weight factor and principal component projection amplitude unit weight factor. The embedding information parameters of each frame in the original B-ultrasound sequence are corrected and coupled with the corresponding unit weight factors to obtain the image embedding feature values ​​of each frame in the original cardiac B-ultrasound image sequence.

[0014] As a preferred technical solution, each systolic-diastolic fluctuation cycle in the original cardiac B-ultrasound image sequence is identified based on the inter-frame dynamic features and image embedding feature values ​​of each frame. The specific process is as follows:

[0015] The inter-frame dynamic features of each frame are synchronously aligned with the image embedding features on the time axis to construct a unified period-aware feature curve.

[0016] By performing first-order derivative analysis on the periodic perception characteristic curve, the changing trend of the characteristic trend is identified, and the positive and negative changes of its derivative are further detected to obtain the trend reversal point, and the potential peak and potential valley values ​​are preliminarily identified.

[0017] Based on the trend reversal point, the local extreme point is determined in combination with the peak-valley detection algorithm. At the same time, the wavelet transform method is introduced to extract the dominant fluctuation rhythm from the period perception characteristic curve. The peak-valley detection results are combined with the wavelet main frequency information to obtain the preliminary candidate boundary points of each contraction-diastole fluctuation cycle, which are recorded as the preliminary candidate boundary points.

[0018] Arrange the preliminary candidate points of each boundary in ascending order according to the frame number, calculate the inter-frame distance between each boundary candidate point and the adjacent points, and if the inter-frame time is less than the set time redundancy threshold, these points are considered to be repeated responses to the same boundary event, and the average frame position is selected as the merged preliminary candidate point of the boundary.

[0019] Calculate the cycle length between two preliminary candidate boundary points, and count the mean and standard deviation of all cycle lengths. Eliminate abnormal cycle segments and record the remaining cycle segments as candidate cycle segments. Align the embedded trajectory of each candidate cycle segment on the time axis and compare its trajectory similarity with the candidate cycles of the same sequence. Filter based on trajectory similarity and eliminate candidate cycle segments below the preset trajectory similarity threshold. Use the cycle boundary points of each filtered candidate cycle segment as the final cycle division result. Output the start and end frame numbers of each systolic-diastolic cycle, and number each frame within each systolic-diastolic cycle according to the standardized position. Output the standardized frame position number of each frame of each systolic-diastolic cycle.

[0020] As a preferred technical solution, the mean value of the structural alignment error of each contraction-diastole fluctuation cycle is obtained, specifically including:

[0021] The image embedding amount of each frame in each contraction-diastole fluctuation cycle is obtained, and the image embedding amounts of each frame with the same standardized frame position number in each contraction-diastole fluctuation cycle are averaged and used as the image embedding amount reference value of the standardized frame position number. The image embedding amount of each frame of a certain contraction-diastole fluctuation cycle is subtracted from the corresponding image embedding amount reference value to obtain the image embedding amount difference of each frame in the contraction-diastole fluctuation cycle. The image embedding amount difference of each frame is averaged to obtain the mean value of the structural alignment error of the cycle.

[0022] As a preferred technical solution, the structural cycle deviation judgment results of each contraction-diastole fluctuation cycle are obtained, specifically including:

[0023] The mean value of the structural alignment error of each contraction-diastole fluctuation cycle is compared with the structural alignment error mean threshold pre-stored in the database. If the mean value of the structural alignment error of a contraction-diastole fluctuation cycle is greater than or equal to the structural alignment error mean threshold, the structural periodic deviation judgment result of the contraction-diastole fluctuation cycle is that there is a structural periodic deviation.

[0024] If the mean value of the structural alignment error of a certain contraction-diastole fluctuation cycle is less than the structural alignment error mean threshold value, the structural cycle deviation judgment result of the contraction-diastole fluctuation cycle is that there is no structural cycle deviation.

[0025] As a preferred technical solution, whether to enter the rhythm abnormality judgment process is determined based on the structural period deviation judgment result, specifically including:

[0026] If there is a structural cycle deviation in a certain contraction-diastole fluctuation cycle, the image embedding amount difference of each frame of the contraction-diastole fluctuation cycle is obtained, and each frame with an image embedding amount difference greater than or equal to the image embedding amount difference threshold is screened as a deviation frame, and the consecutive number of deviation frames of the contraction-diastole fluctuation cycle is counted. If the consecutive number of deviation frames is greater than or equal to the preset consecutive number threshold, the structural cycle deviation of the contraction-diastole fluctuation cycle is defined as a serious deviation, and an early warning is issued, and the rhythm abnormality judgment process is not entered.

[0027] If the continuous number of deviation frames is less than the preset continuous number threshold, the structural periodic deviation of the contraction-diastole fluctuation cycle is defined as a mild deviation, the continuous number of deviation frames and the distribution skewness are obtained, and the continuous number of deviation frames is put into the mapping set of continuous number-image enhancement parameters pre-stored in the database for mapping and matching to obtain the image enhancement parameters of the contraction-diastole fluctuation cycle, and the image enhancement of the contraction-diastole fluctuation cycle is performed with the image enhancement parameters of the contraction-diastole fluctuation cycle, and the distribution skewness is put into the mapping set of distribution skewness-timing position adjustment parameters pre-stored in the database to obtain the timing position adjustment parameters of the contraction-diastole fluctuation cycle, and the timing position adjustment parameters of the contraction-diastole fluctuation cycle are used to adjust the timing positions of each deviation frame of the contraction-diastole fluctuation cycle. After completing the image enhancement and timing position adjustment, the rhythm abnormality judgment process is entered.

[0028] As a preferred technical solution, the image embedding feature parameters of each systolic-diastolic fluctuation cycle are analyzed, and the influence of the structural cycle deviation judgment result is introduced to obtain the rhythm abnormality judgment result of each systolic-diastolic fluctuation cycle, specifically including:

[0029] The embedded information parameters of each frame within each systolic-diastolic fluctuation cycle were extracted, including the embedding norm, characteristic entropy, and principal component projection amplitude of each frame. The embedded information parameters were arranged in frame sequence within the cycle for time series analysis. The index parameters of the internal characteristic curve of each systolic-diastolic fluctuation cycle were calculated, including the first-order derivative, local volatility, and principal component reconstruction error.

[0030] If the structural periodic deviation judgment result of a certain contraction-diastole fluctuation cycle is that there is a structural periodic deviation, the continuous number of deviation frames of the contraction-diastole fluctuation cycle is subtracted from the continuous number threshold to obtain the continuous number difference, and the continuous number difference-penalty factor mapping set pre-stored in the database is put into the mapping matching to obtain the penalty factor of the contraction-diastole fluctuation cycle.

[0031] The index parameters of the internal characteristic curve are modified based on the penalty factor of the contraction-diastole fluctuation cycle.

[0032] If any indicator of a certain contraction-diastole fluctuation cycle exceeds the corresponding preset indicator threshold, the rhythm abnormality judgment result of the contraction-diastole fluctuation cycle is determined to be rhythm abnormality.

[0033] If all indicators of a certain systolic-diastolic fluctuation cycle do not exceed the corresponding preset indicator thresholds, the rhythm abnormality judgment result of the systolic-diastolic fluctuation cycle is determined to be normal rhythm.

[0034] As a preferred technical solution, corresponding operations are performed based on the rhythm abnormality judgment results. The specific processing conditions are:

[0035] If the rhythm abnormality judgment result of the systolic-diastolic fluctuation cycle is a rhythm abnormality, then all the systolic-diastolic fluctuation cycles with rhythm abnormality judgment results are counted. If the number of consecutive cycles is greater than the preset abnormality threshold, the rhythm abnormality is determined to be a global rhythm abnormality, and the main frequency matching loss mechanism is performed.

[0036] If the number of consecutive cycles is less than or equal to the preset abnormal threshold, the rhythm abnormality is determined to be a local cycle abnormality, and all contraction-diastole fluctuation cycles with rhythm abnormality judgment results as rhythm abnormality are marked as abnormal cycles, and an early warning is issued.

[0037] As an optimal technical solution, a main frequency matching loss mechanism is implemented, specifically including:

[0038] By performing Fourier transform on the inter-frame dynamic parameter sequence of each contraction-diastole fluctuation cycle, the main frequency of each frame is extracted, and the spectrum multi-peak ratio of each contraction-diastole fluctuation cycle is obtained. If the spectrum multi-peak ratio of a certain contraction-diastole fluctuation cycle is greater than the preset spectrum multi-peak ratio threshold, it is marked as an abnormal continuous contraction-diastole fluctuation cycle and an early warning is issued.

[0039] If the spectrum multi-peak ratio of all contraction-diastole fluctuation cycles is less than or equal to the spectrum multi-peak ratio threshold, the main frequency of each frame in each contraction-diastole fluctuation cycle is compared with the main frequency with the highest proportion in the entire cycle. Based on the comparison result, the loss factor of each frame is constructed, and the loss factor of each frame is compared with the loss factor threshold pre-stored in the database. If the loss factor of each frame does not exceed the threshold, it is put into the correct rhythm frequency generation algorithm, and the auxiliary algorithm generates the rhythm frequency and corrects each frame.

[0040] If the loss factor of a certain frame in a certain systolic-diastolic fluctuation cycle exceeds the threshold, an early warning is issued.

[0041] In addition, a cardiac ultrasound data anomaly detection system includes:

[0042] The feature extraction module is used to perform dual-channel feature extraction and detection on the original cardiac B-ultrasound image sequence to obtain the inter-frame dynamic feature value and image embedding feature value of each frame in the original cardiac B-ultrasound image sequence.

[0043] The cycle recognition module is used to recognize each systolic-diastolic fluctuation cycle in the original cardiac B-ultrasound image sequence based on the inter-frame dynamic features and image embedding feature values ​​of each frame.

[0044] The structural cycle deviation judgment module is used to obtain the mean value of the structural alignment error of each contraction-diastole fluctuation cycle, obtain the structural cycle deviation judgment result of each contraction-diastole fluctuation cycle, and decide whether to enter the rhythm abnormality judgment process based on the structural cycle deviation judgment result.

[0045] The rhythm abnormality judgment module analyzes the image embedding feature parameters of each contraction-diastole fluctuation cycle, obtains the rhythm abnormality judgment results of each contraction-diastole fluctuation cycle after introducing the influence of the structural cycle deviation judgment results, and performs corresponding operations based on the rhythm abnormality judgment results.

[0046] Compared with the prior art, the embodiments of the present invention have at least the following beneficial effects:

[0047] (1) The present invention provides a method for detecting abnormalities in cardiac ultrasound data. Through a "dual-channel feature extraction" architecture, inter-frame dynamic parameters and image embedding information are simultaneously collected, and a unit weight factor is introduced for coupling processing, effectively improving the pertinence and noise resistance of feature expression. Furthermore, a unified cycle perception curve is constructed through feature alignment, and multiple methods such as derivative analysis, wavelet extraction, and extreme value screening are used to ensure the accuracy and continuity of systolic-diastolic cycle identification. In addition, the method systematically filters abnormal cycle segments, standardizes the cycle boundary frame sequence, and eliminates cycles with low similarity, thereby enhancing the robustness of cycle division and providing a stable foundation for subsequent analysis.

[0048] (2) The present invention constructs a judgment logic for the two stages of structural periodic deviation and rhythm anomaly. First, the periodic structural deviation is judged by the mean value of the structural alignment error, and then it is judged as severe or mild deviation based on the number of deviation frames. In the case of severe deviation, a warning is directly issued and subsequent judgment is skipped; in the case of mild deviation, the image enhancement and timing adjustment stage is entered before rhythm anomaly analysis. The rhythm anomaly judgment is further introduced into the principal component reconstruction error, derivative change and volatility, and correction is made in combination with the penalty factor of structural deviation. This progressive judgment path can effectively distinguish local short-term anomalies from potential systemic fluctuation problems, and realize a more comprehensive and reasonable judgment strategy.

[0049] (3) The present invention further classifies abnormalities into local period abnormalities or global rhythm abnormalities based on their continuity, and issues warnings or enters the main frequency matching loss mechanism respectively. The main frequency mechanism obtains spectrum information through Fourier transform and constructs a loss factor by combining the spectrum multi-peak ratio and main frequency comparison. If the abnormal condition is not triggered, it further assists in rhythm frequency repair. This mechanism realizes the closed-loop regulation capability after the abnormality occurs, and can adaptively adjust the rhythm trend of the abnormal signal. In addition, this method corrects the main frequency of the periodic frame frame by frame, providing stable and structured frequency information.

[0050] Of course, any product implementing the present invention does not necessarily need to achieve all of the above advantages at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Schematic diagram of the method of the present invention.

[0052] Figure 2Schematic diagram of the system module of the present invention.

[0053] Figure 3 It is a logical flow diagram of the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0055] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "thickness", "top", "middle", "length", "inside", "around" and the like indicating orientation or positional relationship are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0056] See also Figure 1 and Figure 3 As shown, an embodiment of the present invention provides a method for detecting abnormalities in cardiac ultrasound data, comprising:

[0057] A dual-channel feature extraction and detection is performed on the original cardiac B-ultrasound image sequence to obtain the inter-frame dynamic feature value and image embedding feature value of each frame in the original cardiac B-ultrasound image sequence.

[0058] The inter-frame dynamic feature value and image embedding feature value of each frame in the original cardiac B-ultrasound image sequence are obtained. The specific process is as follows:

[0059] In the inter-frame dynamic channel, motion information parameters for each frame in the original cardiac B-ultrasound image sequence are extracted, including the local optical flow mean, boundary displacement intensity, frame difference energy density, and displacement direction mutation rate, and the local optical flow mean is calculated. Optical flow estimation is performed between adjacent frames (using methods such as Farneback optical flow, RAFT, or PWC-Net) to obtain the motion vector for each pixel. Each frame is divided into multiple subregions, and the optical flow vectors within each region are averaged to obtain the local optical flow mean distribution for that region. By summarizing the optical flow means of multiple regions within a frame, a low-dimensional motion representation reflecting the overall displacement trend of that frame is formed.

[0060] Evaluate boundary displacement strength. Use an edge detection algorithm (such as Canny or Sobel) to extract anatomical structure boundaries in each image frame. Then, match and measure the movement distance of these boundaries between adjacent frames. In this embodiment of the present invention, the matching and measurement strategy uses optical flow tracking to obtain the movement distance of each edge point between adjacent frames. The movement distances of each edge point between adjacent frames are averaged to obtain the average movement distance of the structure boundary, which is the boundary displacement strength. This represents the degree of active deformation and movement of cardiac tissue and is particularly suitable for identifying critical moments of contraction and relaxation.

[0061] Calculate the frame difference energy density. For each pair of adjacent frames, a pixel-level difference image is calculated. These difference images are squared and averaged to obtain the energy density value of the inter-frame image difference. This value is the average squared value of the pixel intensity variation and quantifies the overall intensity of the image change between two adjacent frames. A larger value indicates a more dramatic image change between two adjacent frames, indicating a strong jump or rhythm shift in cardiac motion.

[0062] Extract the displacement direction mutation rate. Extract the movement direction of the pixel points from the optical flow vector field, and construct a direction histogram of the entire frame. By tracking these direction changes in the time dimension, identify whether there is a sharp directional turn in the local area in multiple consecutive frames. In an embodiment of the present invention, the specific judgment conditions include: in the direction histogram of the entire frame, the direction at time t+1 in the time dimension and the direction at time t change in the time dimension are greater than or equal to the preset direction change threshold, then it is determined that there is a sharp directional turn. The mutation rate is obtained by counting the pixel points in the entire frame whose direction change is greater than the preset threshold, and calculating its proportion of all pixel points in the entire frame for quantification. It is an important indicator for judging rhythm disturbances or abnormal movement patterns.

[0063] Unit weight factors corresponding to motion information parameters are extracted from the database, including the unit weight factor of local optical flow mean, the unit weight factor of boundary displacement intensity, the unit weight factor of frame difference energy density, and the unit weight factor of displacement direction mutation rate. The motion information parameters of each frame in the original cardiac B-ultrasound image sequence are corrected and coupled with the corresponding unit weight factors to obtain the inter-frame dynamic feature values ​​of each frame in the original cardiac B-ultrasound image sequence, specifically including: ;

[0064] in, is the inter-frame dynamic feature value of the t-th frame, is the local optical flow mean of the t-th frame, is the boundary displacement intensity of the t-th frame, is the frame difference energy density of the t-th frame, is the displacement direction mutation rate of the t-th frame, is the unit weight factor of the local optical flow mean, is the unit weight factor of boundary displacement intensity, is the unit weight factor of the frame difference energy density, is the unit weight factor of the mutation rate in the displacement direction, is the frame number in the original cardiac B-ultrasound image sequence, , is the total number of frames in the original cardiac B-ultrasound image sequence.

[0065] It should be noted that the local optical flow mean unit weighting factor is used to normalize the local displacement intensity between frames (in pixels / frame), making it dimensionless in subsequent feature fusion. This factor also controls the proportion of local global displacement in the dynamic feature representation. When the overall motion trend of the heart is strong but the structural deformation is weak, increasing the weight of this factor can help enhance its sensitivity to rhythm judgment.

[0066] The boundary displacement intensity unit weighting factor measures the average distance cardiac anatomical structure boundaries move between adjacent frames, also expressed in pixels per frame. This factor, while eliminating unit weights, assigns a higher or lower impact to structural boundary deformations. This is useful for enhancing the dynamic representation of changes in the systolic-diastolic rhythm on key structures such as the cardiac wall and valves.

[0067] The frame difference energy density unit weighting factor normalizes pixel-level image intensity changes (measured in squared grayscale values) to a unit value. It is primarily used to weight the intensity of image changes relative to dynamic features. This factor is crucial for identifying cardiac motion jumps, abnormal interruptions, or sudden rhythm changes.

[0068] The unit weighting factor for the displacement direction mutation rate is used to control the influence of the degree of pixel directional mutation (itself a dimensionless ratio). In the original image sequence, this parameter reflects the degree of dramatic changes in the motion direction of a local region and is a significant indicator of abnormal motion patterns, rhythm disturbances, and premature beats. By adjusting this factor, the contribution of directional anomalies to the overall inter-frame dynamic characteristics can be controlled.

[0069] It should also be noted that the motion information parameters of each frame, including the local optical flow mean, boundary displacement intensity, frame difference energy density, and displacement direction mutation rate, are correlated. The local optical flow mean reflects the overall pixel-level motion trend of the image, while the boundary displacement intensity focuses on the degree of deformation at the anatomical edge. Both typically increase simultaneously during strong contraction or relaxation, reflecting the consistency between global cardiac motion and local structural changes. Local optical flow is significantly affected by global motion (e.g., slight probe movement), while boundary displacement intensity is more sensitive to changes in cardiac chamber volume and valve motion, making it more representative of morphological changes. Violent image motion (e.g., rapid contraction) is often accompanied by increased pixel motion amplitude and enhanced image grayscale variation, manifesting as an increase in optical flow amplitude and inter-frame difference. Therefore, in the absence of other interferences (e.g., probe jitter or noise mutation), the local optical flow mean and frame difference energy density may exhibit a certain degree of positive correlation under specific motion patterns. Optical flow describes displacement velocity (vector field), while frame difference energy emphasizes pixel intensity changes (brightness difference). Therefore, frame difference is sensitive to image noise, while optical flow is more stable. During the regular contraction-relaxation process of a rhythm, the directional mutation rate is typically low. Abnormal motion (such as premature beats and tremors) can cause discontinuities in the optical flow direction, increasing the mutation rate. This can be accompanied by abnormal changes in the optical flow amplitude and increased image brightness differences, leading to fluctuations in the optical flow distribution and increased frame difference energy. The directional mutation rate is a morphological "mutation" metric. Unlike the "intensity" properties of other parameters, it is particularly sensitive to abnormal, non-periodic motion and is an important supplement for identifying abnormalities such as arrhythmias.

[0070] In the image embedding feature channel, embedding information parameters for each frame in the original B-ultrasound sequence are extracted, including the embedding norm, feature entropy, and principal component projection amplitude. In this channel, high-dimensional features are extracted from each frame of the original B-ultrasound sequence using a deep image encoder (such as a convolutional neural network (CNN), a visual Transformer, or a hybrid encoder) to obtain an embedding vector reflecting cardiac structural information. Inter-frame embedding information parameters with diagnostic or recognition value are further extracted from these embedding vectors to characterize the dynamic changes and temporal consistency of the inter-frame structure. Specifically, these parameters include the following aspects:

[0071] The embedding norm calculates the L2 norm of the embedded vector extracted for each frame, which is used to measure the overall structural feature strength of the frame. Variations in the embedding norm reflect the periodic fluctuations in the structural complexity of the heart during different phases (such as systole and diastole), helping to identify frames with weak function or ambiguous structure.

[0072] Feature entropy measures the structural complexity and stability of the current frame in the embedding space by normalizing the embedding vector of each frame and calculating its information entropy. High feature entropy indicates rich structural information or the potential for uncertainty, such as motion blur or unusual deformation. By comparing entropy fluctuations between consecutive frames, structural jumps and inconsistent areas can be identified.

[0073] The principal component projection amplitude is obtained by performing principal component analysis (PCA) on a sequence of embedded vectors to obtain the projection results of the first few principal components. The projection amplitude reflects the main direction and amplitude of structural changes in the embedded space and can be used to measure the regularity and integrity of the entire periodic structural activity. Sudden changes in the principal component trajectory or a sharp decrease in amplitude may indicate abnormal periodic motion or localized functional impairment.

[0074] Extract the unit weight factors corresponding to the embedding information parameters from the database, including the embedding norm unit weight factor, the characteristic entropy unit weight factor, and the principal component projection amplitude unit weight factor. Modify and couple the embedding information parameters of each frame in the original B-ultrasound sequence with the corresponding unit weight factors to obtain the image embedding feature values ​​of each frame in the original cardiac B-ultrasound image sequence, specifically including: ;

[0075] in, is the image embedding feature value of the t-th frame, is the embedding norm of the t-th frame, is the feature entropy of the t-th frame, is the principal component projection amplitude of the t-th frame, is the embedding norm unit weight factor, is the characteristic entropy unit weight factor, is the unit weight factor of the principal component projection amplitude, is the frame number in the original cardiac B-ultrasound image sequence, , is the total number of frames in the original cardiac B-ultrasound image sequence.

[0076] It should be noted that the embedding norm unit weight factor is used to eliminate the overall magnitude (i.e., the L2 norm, measured in units of vector modulus) of each frame's embedding vector, rendering this feature dimensionless across different encoder architectures and scale distributions. This factor also controls the contribution of the image's overall structural strength to the periodic dynamics. When the cardiac structure exhibits significant overall contraction or relaxation, while local features exhibit less pronounced changes, appropriately increasing this factor's weight can help enhance the significance of that frame in the analysis of periodic structural fluctuations.

[0077] The feature entropy unit weighting factor normalizes the information entropy value (in bits / frame or nats / frame) of the image embedding vector to remove bias related to the number of embedding dimensions or the normalization strategy. Feature entropy reflects the complexity and uncertainty of the embedded features. Setting this factor adjusts the weight given to these changes in complexity in periodicity detection and anomaly recognition. Increasing this factor can improve the model's sensitivity to structurally chaotic regions, particularly when identifying blurred frames, frames with abnormal motion, or transient transitions.

[0078] The principal component projection amplitude unit weighting factor is used to eliminate the dimensional effects of different amplitude components during principal component analysis, ensuring comparability of principal direction changes across different cycles or lesion conditions. This factor also controls the dynamic representation weight of the principal structural deformation direction (e.g., the systolic axis). When the integrity of the cardiac structure's systolic / diastolic trajectory decreases or the projection amplitude changes suddenly, adjusting this factor can effectively enhance its representation during abnormal cycle detection and improve the accuracy of identifying frames with reduced function.

[0079] It should also be noted that there is a certain correlation between the embedding information parameters, including the embedding norm, feature entropy, and principal component projection amplitude. The embedding norm represents the overall "strength" or "amplitude" of the embedding vector, reflecting the strength of the energy activated in the high-dimensional feature space of the image frame. The feature entropy describes the "complexity" or "uniformity" of the embedding vector distribution and is used to measure the uncertainty of the embedded feature dimensions of the image frame. The principal component projection amplitude reflects the concentration or dominant trend of the embedded features along the main direction (such as the principal axis of the PCA), highlighting the main deformation pattern. In the presence of noise or invalid features in the image, the entropy may increase while the norm does not increase significantly, indicating redundant distribution or abnormal perturbations. When image features are concentrated in a specific direction (such as a significant deformation trend of an anatomical structure), the norm is high and the principal component projection amplitude is large. If the image is noisy or features are divergent, the norm may be high but the principal component amplitude may be low, indicating a non-concentrated energy distribution. When the feature distribution is concentrated in a few principal components (large principal projection amplitude), the entropy may be low (concentrated distribution). If the entropy is high but the principal component amplitude is low, it indicates scattered features, unclear structure, or interference.

[0080] Each systolic-diastolic fluctuation cycle in the original cardiac B-ultrasound image sequence is identified based on the inter-frame dynamic feature value and image embedding feature value of each frame.

[0081] The inter-frame dynamic feature values ​​and image embedding feature values ​​of each frame are synchronously aligned on the time axis to construct a unified periodic perception feature curve.

[0082] It should be noted that the original cardiac B-ultrasound image sequence has a natural frame sequence number, and the frame sequence number can be used as a time index. Since the inter-frame dynamic features and image embedding features are extracted frame by frame based on the image sequence, the frame number can be directly used as the time axis reference to correspond the features of the two channels one-to-one in the time dimension.

[0083] Because the dimensions and numerical ranges of the two types of features differ significantly, each type of feature needs to be normalized. In this embodiment of the present invention, an excitation function is used to map the inter-frame dynamic feature values ​​and image embedding feature values ​​of each frame to the interval [0, 1]. The inter-frame dynamic feature values ​​are multiplied by the inter-frame dynamic feature weighting factor, and the image embedding feature values ​​are multiplied by the image embedding feature weighting factor. Coupled processing is then performed and plotted along the frame sequence number to obtain a periodic perception feature curve. The periodic perception feature curve integrates the joint information of motion trend and structural expression, possessing a well-defined periodic fluctuation structure, which serves as the basis for subsequent peak and valley detection, wavelet analysis, and period boundary determination.

[0084] By performing first-order derivative analysis on the periodic perception characteristic curve, the changing trend of the characteristic trend is identified, and the positive and negative changes of its derivative are further detected to obtain the trend reversal point, and the potential peak and potential valley values ​​are preliminarily identified.

[0085] It should be noted that the purpose of performing first-order derivative analysis on the periodic perception characteristic curve is to identify the trend of the characteristic over time, so as to locate the key fluctuation nodes in the periodic rhythm. The specific operation process is as follows: Obtain the periodic perception characteristic curve , the curve represents the characteristic value of each frame image after fusion on the time axis. Perform a first-order difference operation to approximate its first-order derivative , specifically: ;

[0086] The derivative reflects the growth and decay trend of the eigenvalue. When , it means the feature is in an upward trend. , it indicates that the feature is in a downward trend.

[0087] Furthermore, the positive and negative changes of the derivative are identified to obtain trend reversal points, i.e., turning points from positive to negative and turning points from negative to positive, wherein the turning points from positive to negative usually correspond to the local peak value of the characteristic curve, which corresponds to the apex of cardiac contraction in an embodiment of the present invention, and the turning points from negative to positive usually correspond to the local valley value of the characteristic curve, which corresponds to the diastolic low point in an embodiment of the present invention.

[0088] Based on the trend reversal point, the local extreme point is determined in combination with the peak-valley detection algorithm. At the same time, the wavelet transform method is introduced to extract the dominant fluctuation rhythm from the period perception characteristic curve. The peak-valley detection results are combined with the wavelet main frequency information to obtain the preliminary candidate boundary points of each contraction-diastole fluctuation cycle, which are recorded as the preliminary candidate boundary points.

[0089] It should be noted that at the identified trend reversal points, a peak-valley detection algorithm is used to further precisely locate local maxima and local minima. Local maxima reflect the systolic peak, while local minima reflect the diastolic trough. Ultimately, a set of time point sequences is obtained, representing the local systolic peak and diastolic trough, respectively, constituting a preliminary extreme value time series. To enhance the ability to identify the dominant rhythm, a wavelet transform is introduced to perform multi-scale frequency domain analysis on the periodic perception characteristic curve. The specific process includes: selecting a mother wavelet, in this embodiment of the present invention, a Morlet wavelet is selected, extracting the amplitude response and power spectral density at the dominant wavelet frequency, and analyzing the dominant frequency components with concentrated energy in the time-frequency diagram. The dominant frequency components are used to reflect the length of the basic systolic-diastolic cycle, obtaining an approximate period range for period boundary constraints.

[0090] The peak and valley detection results are integrated with the wavelet dominant frequency rhythm for analysis, specifically including: under the constraint of the dominant frequency period, the extreme value pairs with too small or too large intervals between adjacent extreme value points are screened out, and the start and end frames of each candidate period segment are used as a preliminary candidate boundary point.

[0091] Arrange the preliminary candidate points of each boundary in ascending order according to the frame number, calculate the inter-frame distance between each boundary candidate point and the adjacent points, and if the inter-frame time is less than the set time redundancy threshold, these points are considered to be repeated responses to the same boundary event, and the average frame position is selected as the merged preliminary candidate point of the boundary.

[0092] Calculate the cycle length between two preliminary candidate boundary points, and count the mean and standard deviation of all cycle lengths. Eliminate abnormal cycle segments and record the remaining cycle segments as candidate cycle segments. Align the embedded trajectory of each candidate cycle segment on the time axis and compare its trajectory similarity with the candidate cycles of the same sequence. Filter based on trajectory similarity and eliminate candidate cycle segments below the preset trajectory similarity threshold. Use the cycle boundary points of each filtered candidate cycle segment as the final cycle division result. Output the start and end frame numbers of each systolic-diastolic cycle, and number each frame within each systolic-diastolic cycle according to the standardized position. Output the standardized frame position number of each frame of each systolic-diastolic cycle.

[0093] The mean value of the structural alignment error of each systolic-diastolic fluctuation cycle is obtained, and the structural cycle deviation judgment result of each systolic-diastolic fluctuation cycle is obtained. Based on the structural cycle deviation judgment result, it is decided whether to enter the rhythm abnormality judgment process.

[0094] Obtain the mean structural alignment error for each systolic-diastolic fluctuation cycle, specifically including:

[0095] The image embedding amount of each frame in each contraction-diastole fluctuation cycle is obtained, and the image embedding amounts of each frame with the same standardized frame position number in each contraction-diastole fluctuation cycle are averaged and used as the image embedding amount reference value of the standardized frame position number. The image embedding amount of each frame of a certain contraction-diastole fluctuation cycle is subtracted from the corresponding image embedding amount reference value to obtain the image embedding amount difference of each frame in the contraction-diastole fluctuation cycle. The image embedding amount difference of each frame is averaged to obtain the mean value of the structural alignment error of the cycle.

[0096] It should be noted that the image embedding value refers to the numerical expression of the high-dimensional feature vector in the embedding space after each frame of cardiac ultrasound image is input into a deep coding network (such as CNN, ViT, etc.). It is used to represent the compressed expression of the structural information in that frame of image, and can reflect deep information such as the image's texture features, boundary shape, and anatomical structure layout. Through a deep network, in this embodiment of the present invention, CLIP (Contrastive Language-Image Pretraining), each frame of two-dimensional cardiac ultrasound image is input into the model, and the output of the middle layer or the second-to-last layer is extracted as the image embedding vector.

[0097] Based on the structural period deviation judgment result, it is decided whether to enter the rhythm abnormality judgment process, which includes:

[0098] If there is a structural cycle deviation in a certain contraction-diastole fluctuation cycle, the image embedding amount difference of each frame of the contraction-diastole fluctuation cycle is obtained, and each frame with an image embedding amount difference greater than or equal to the image embedding amount difference threshold is screened as a deviation frame, and the consecutive number of deviation frames of the contraction-diastole fluctuation cycle is counted. If the consecutive number of deviation frames is greater than or equal to the preset consecutive number threshold, the structural cycle deviation of the contraction-diastole fluctuation cycle is defined as a serious deviation, and an early warning is issued, and the rhythm abnormality judgment process is not entered.

[0099] If the continuous number of deviation frames is less than the preset continuous number threshold, the structural periodic deviation of the contraction-diastole fluctuation cycle is defined as a mild deviation, the continuous number of deviation frames and the distribution skewness are obtained, and the continuous number of deviation frames is put into the mapping set of continuous number-image enhancement parameters pre-stored in the database for mapping and matching to obtain the image enhancement parameters of the contraction-diastole fluctuation cycle, and the image enhancement of the contraction-diastole fluctuation cycle is performed with the image enhancement parameters of the contraction-diastole fluctuation cycle, and the distribution skewness is put into the mapping set of distribution skewness-timing position adjustment parameters pre-stored in the database to obtain the timing position adjustment parameters of the contraction-diastole fluctuation cycle, and the timing position adjustment parameters of the contraction-diastole fluctuation cycle are used to adjust the timing positions of each deviation frame of the contraction-diastole fluctuation cycle. After completing the image enhancement and timing position adjustment, the rhythm abnormality judgment process is entered.

[0100] It should be noted that the distribution skewness is an indicator used in statistics to measure the degree of asymmetry of data distribution. It describes the direction and degree of skewness of the data distribution relative to the mean. In an embodiment of the present invention, in the contraction-diastole fluctuation cycle, the distribution skewness of the deviation frames is calculated for the frames that structurally deviate, and is used to describe whether the distribution of these deviation frames in the time series is uniform, or whether they are biased towards a certain part of the cycle. If the distribution skewness of the deviation frames is close to zero, it means that the deviation frames are relatively evenly distributed throughout the cycle; if the skewness is positive, it means that the deviation frames are mainly concentrated in the second half of the cycle; if the skewness is negative, it means that the deviation frames are mainly concentrated in the first half of the cycle.

[0101] The structural cycle deviation judgment results of each contraction-diastole fluctuation cycle are obtained, including:

[0102] The mean value of the structural alignment error of each contraction-diastole fluctuation cycle is compared with the structural alignment error mean threshold pre-stored in the database. If the mean value of the structural alignment error of a contraction-diastole fluctuation cycle is greater than or equal to the structural alignment error mean threshold, the structural periodic deviation judgment result of the contraction-diastole fluctuation cycle is that there is a structural periodic deviation.

[0103] If the mean value of the structural alignment error of a certain contraction-diastole fluctuation cycle is less than the structural alignment error mean threshold value, the structural cycle deviation judgment result of the contraction-diastole fluctuation cycle is that there is no structural cycle deviation.

[0104] The image embedding feature parameters of each systolic-diastolic fluctuation cycle are analyzed, and the rhythm abnormality judgment results of each systolic-diastolic fluctuation cycle are obtained after introducing the influence of the structural cycle deviation judgment results. Corresponding operations are performed based on the rhythm abnormality judgment results.

[0105] The image embedding feature parameters of each systolic-diastolic fluctuation cycle are analyzed, and the influence of the structural cycle deviation judgment result is introduced to obtain the rhythm abnormality judgment result of each systolic-diastolic fluctuation cycle, including:

[0106] The embedded information parameters of each frame within each systolic-diastolic fluctuation cycle were extracted, including the embedding norm, characteristic entropy, and principal component projection amplitude of each frame. The embedded information parameters were arranged in frame sequence within the cycle for time series analysis. The index parameters of the internal characteristic curve of each systolic-diastolic fluctuation cycle were calculated, including the first-order derivative, local volatility, and principal component reconstruction error.

[0107] It is important to note that the first-order derivative reflects the rate of change of the characteristic curve over time, that is, the slope of the inter-frame embedded parameter change. It reveals the dynamic changes in cardiac structure during the systolic-diastolic cycle, such as rapid rise or fall phases. The first-order derivative is calculated using numerical difference methods for the embedded parameter sequence within the cycle.

[0108] Local volatility describes the amplitude and instability of the characteristic curve within the systolic-diastolic fluctuation cycle, reflecting subtle changes in structural characteristics and possible abnormal jumps within the cycle. Large volatility generally indicates dramatic local structural changes, possibly corresponding to abnormal cardiac motion or noise. It is calculated by calculating the standard deviation within the systolic-diastolic fluctuation cycle.

[0109] The principal component reconstruction error measures the error in reconstructing the original embedded feature sequence using a selected number of principal components. It reflects the ability of principal component analysis (PCA) to explain dynamic changes in periodic structures. Large errors may indicate that important abnormal patterns are not captured by the principal components. The reconstruction error is calculated by performing PCA on the embedded information parameters, selecting the principal components, and reconstructing the original data.

[0110] If the structural periodic deviation judgment result of a certain contraction-diastole fluctuation cycle is that there is a structural periodic deviation, the continuous number of deviation frames of the contraction-diastole fluctuation cycle is subtracted from the continuous number threshold to obtain the continuous number difference, and the continuous number difference-penalty factor mapping set pre-stored in the database is put into the mapping matching to obtain the penalty factor of the contraction-diastole fluctuation cycle. The matching rule is that the smaller the continuous number difference, the larger the penalty factor.

[0111] The index parameter of the internal characteristic curve is corrected based on the penalty factor of the contraction-diastole fluctuation cycle, and the correction process includes multiplying the penalty factor and the index parameter.

[0112] If any indicator of a certain contraction-diastole fluctuation cycle exceeds the corresponding preset indicator threshold, the rhythm abnormality judgment result of the contraction-diastole fluctuation cycle is determined to be rhythm abnormality.

[0113] If all indicators of a certain systolic-diastolic fluctuation cycle do not exceed the corresponding preset indicator thresholds, the rhythm abnormality judgment result of the systolic-diastolic fluctuation cycle is determined to be normal rhythm.

[0114] Implement corresponding operations based on the results of rhythm abnormality judgment. The specific processing conditions are as follows:

[0115] If the rhythm abnormality judgment result of the systolic-diastolic fluctuation cycle is a rhythm abnormality, then all the systolic-diastolic fluctuation cycles with rhythm abnormality judgment results are counted. If the number of consecutive cycles is greater than the preset abnormality threshold, the rhythm abnormality is determined to be a global rhythm abnormality, and the main frequency matching loss mechanism is performed.

[0116] If the number of consecutive cycles is less than or equal to the preset abnormal threshold, the rhythm abnormality is determined to be a local cycle abnormality, and all contraction-diastole fluctuation cycles with rhythm abnormality judgment results as rhythm abnormality are marked as abnormal cycles, and an early warning is issued.

[0117] Implement the main frequency matching loss mechanism, including:

[0118] By performing Fourier transform on the inter-frame dynamic parameter sequence of each contraction-diastole fluctuation cycle, the main frequency of each frame is extracted, and the spectrum multi-peak ratio of each contraction-diastole fluctuation cycle is obtained. If the spectrum multi-peak ratio of a certain contraction-diastole fluctuation cycle is greater than the preset spectrum multi-peak ratio threshold, it is marked as an abnormal continuous contraction-diastole fluctuation cycle and an early warning is issued.

[0119] It should be noted that Fourier transforming the inter-frame dynamic parameter sequence of each systolic-diastolic fluctuation cycle converts the time domain into the frequency domain, extracting the energy distribution of the sequence across different frequency components. The dominant frequency is the frequency component with the highest energy in the spectrum, corresponding to the heart's primary rhythm during that cycle. A stable and clear dominant frequency indicates a well-regularized rhythmic movement within that cycle; a blurred or shifted dominant frequency may indicate abnormal structural activity. The multi-peak ratio (MPR) is the ratio of the sum of the energy of other larger frequency components in the spectrum, excluding the dominant frequency, to the total energy. It measures whether there are multiple competing rhythms or motion instability within that cycle. The Fourier transform can reveal seemingly subtle rhythmic variations and is particularly effective for highly periodic signals such as cardiac motion. As an abnormality indicator, the MPR can identify cycles that appear to have normal periodicity but possess complex or abnormal rhythms.

[0120] If the spectrum multi-peak ratio of all contraction-diastole fluctuation cycles is less than or equal to the spectrum multi-peak ratio threshold, the main frequency of each frame in each contraction-diastole fluctuation cycle is compared with the main frequency with the highest proportion in the entire cycle. Based on the comparison result, the loss factor of each frame is constructed, and the loss factor of each frame is compared with the loss factor threshold pre-stored in the database. If the loss factor of each frame does not exceed the threshold, it is put into the correct rhythm frequency generation algorithm, and the auxiliary algorithm generates the rhythm frequency and corrects each frame.

[0121] It should be noted that the main frequency with the highest proportion in the entire cycle is extracted by performing histogram statistics on the main frequencies of all frames to find the main frequency with the highest proportion. Constructing the loss factor of each frame specifically includes: comparing the main frequency of each frame with the main frequency with the highest proportion, and calculating its loss factor, specifically: ,in, is the loss factor between the main frequency of the i-th frame and the main frequency with the highest proportion, is the main frequency of the i-th frame, The loss factor of each frame is fed into the correct rhythm frequency generation algorithm, which in this embodiment of the present invention is a frequency smoothing interpolation + multi-cycle regression algorithm, to perform rhythm curve fitting and generate a smooth periodic rhythm frequency curve. This is used to reposition slightly offset frames based on the rhythm frequency curve.

[0122] If the loss factor of a certain frame in a certain systolic-diastolic fluctuation cycle exceeds the threshold, an early warning is issued.

[0123] like Figure 2 As shown, in this embodiment, the present invention provides a cardiac ultrasound data abnormality detection system, comprising:

[0124] The feature extraction module is used to perform dual-channel feature extraction and detection on the original cardiac B-ultrasound image sequence to obtain the inter-frame dynamic feature value and image embedding feature value of each frame in the original cardiac B-ultrasound image sequence.

[0125] The cycle recognition module is used to recognize each systolic-diastolic fluctuation cycle in the original cardiac B-ultrasound image sequence based on the inter-frame dynamic features and image embedding feature values ​​of each frame.

[0126] The structural cycle deviation judgment module is used to obtain the mean value of the structural alignment error of each contraction-diastole fluctuation cycle, obtain the structural cycle deviation judgment result of each contraction-diastole fluctuation cycle, and decide whether to enter the rhythm abnormality judgment process based on the structural cycle deviation judgment result.

[0127] The rhythm abnormality judgment module analyzes the image embedding feature parameters of each contraction-diastole fluctuation cycle, obtains the rhythm abnormality judgment results of each contraction-diastole fluctuation cycle after introducing the influence of the structural cycle deviation judgment results, and performs corresponding operations based on the rhythm abnormality judgment results.

[0128] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0129] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all details in detail, nor do they limit the invention to specific implementation methods. Obviously, many modifications and changes can be made based on the contents of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can better understand and utilize the present invention. As long as they do not deviate from the structure of the present invention or exceed the scope defined by the present invention, they should fall within the scope of protection of the present invention.

Claims

1. A method for detecting abnormalities in cardiac ultrasound data, characterized in that: include: Perform dual-channel feature extraction and detection on the original cardiac B-ultrasound image sequence to obtain the inter-frame dynamic feature value and image embedding feature value of each frame in the original cardiac B-ultrasound image sequence; Identify each systolic-diastolic fluctuation cycle in the original cardiac B-ultrasound image sequence based on the inter-frame dynamic features and image embedding feature values ​​of each frame; Obtaining the mean value of the structural alignment error of each systolic-diastolic fluctuation cycle, obtaining the structural cycle deviation judgment result of each systolic-diastolic fluctuation cycle, and determining whether to enter the rhythm abnormality judgment process based on the structural cycle deviation judgment result; The image embedding feature parameters of each systolic-diastolic fluctuation cycle are analyzed, and the rhythm abnormality judgment results of each systolic-diastolic fluctuation cycle are obtained after introducing the influence of the structural cycle deviation judgment results. Corresponding operations are performed based on the rhythm abnormality judgment results.

2. The method for detecting abnormalities in cardiac ultrasound data according to claim 1, wherein: The specific process of obtaining the inter-frame dynamic feature value and image embedding feature value of each frame in the original cardiac B-ultrasound image sequence is as follows: In the inter-frame dynamic channel, the motion information parameters of each frame in the original cardiac B-ultrasound image sequence are extracted, including the local optical flow mean, boundary displacement intensity, frame difference energy density and displacement direction mutation rate. The unit weight factors corresponding to the motion information parameters are extracted from the database, including the local optical flow mean unit weight factor, the boundary displacement intensity unit weight factor, the frame difference energy density unit weight factor and the displacement direction mutation rate unit weight factor. The motion information parameters of each frame in the original cardiac B-ultrasound image sequence are respectively corrected and coupled with the corresponding unit weight factors to obtain the inter-frame dynamic feature values ​​of each frame in the original cardiac B-ultrasound image sequence; In the image embedding feature channel, the embedding information parameters of each frame in the original B-ultrasound sequence are extracted, including the embedding norm, characteristic entropy and principal component projection amplitude. The unit weight factors corresponding to the embedding information parameters are extracted from the database, including the embedding norm unit weight factor, characteristic entropy unit weight factor and principal component projection amplitude unit weight factor. The embedding information parameters of each frame in the original B-ultrasound sequence are corrected and coupled with the corresponding unit weight factors to obtain the image embedding feature values ​​of each frame in the original cardiac B-ultrasound image sequence.

3. The method for detecting abnormalities in cardiac ultrasound data according to claim 1, wherein: The specific process of identifying each systolic-diastolic fluctuation cycle in the original cardiac B-ultrasound image sequence based on the inter-frame dynamic features and image embedding feature values ​​of each frame is as follows: The inter-frame dynamic features of each frame are synchronously aligned with the image embedding features on the time axis to construct a unified period-aware feature curve; By performing first-order derivative analysis on the cycle perception characteristic curve, we can identify the changing trend of the characteristic trend, further detect the positive and negative changes of its derivative to obtain the trend reversal point, and preliminarily identify the potential peak and potential valley; Based on the trend reversal point, the local extreme point is determined by combining the peak-valley detection algorithm. At the same time, the wavelet transform method is introduced to extract the dominant fluctuation rhythm from the period perception characteristic curve. The peak-valley detection results are combined with the wavelet dominant frequency information to obtain the preliminary candidate points of the boundary of each systolic-diastolic fluctuation cycle, which are recorded as the preliminary candidate points of each boundary. Arrange the preliminary candidate points of each boundary in ascending order according to the frame number, calculate the inter-frame distance between each candidate point and the adjacent points, and if the inter-frame time is less than the set temporal redundancy threshold, these points are considered to be repeated responses to the same boundary event, and the average frame position is selected as the merged preliminary candidate point of the boundary. Calculate the cycle length between two preliminary candidate boundary points, and count the mean and standard deviation of all cycle lengths. Eliminate abnormal cycle segments and record the remaining cycle segments as candidate cycle segments. Align the embedded trajectory of each candidate cycle segment on the time axis and compare its trajectory similarity with the candidate cycles of the same sequence. Filter based on trajectory similarity and eliminate candidate cycle segments below the preset trajectory similarity threshold. Use the cycle boundary points of each filtered candidate cycle segment as the final cycle division result. Output the start and end frame numbers of each systolic-diastolic cycle, and number each frame within each systolic-diastolic cycle according to the standardized position. Output the standardized frame position number of each frame of each systolic-diastolic cycle.

4. The method for detecting abnormalities in cardiac ultrasound data according to claim 1, wherein: The obtaining of the mean value of the structural alignment error of each contraction-diastole fluctuation cycle specifically includes: The image embedding amount of each frame in each contraction-diastole fluctuation cycle is obtained, and the image embedding amounts of each frame with the same standardized frame position number in each contraction-diastole fluctuation cycle are averaged and used as the image embedding amount reference value of the standardized frame position number. The image embedding amount of each frame of a certain contraction-diastole fluctuation cycle is subtracted from the corresponding image embedding amount reference value to obtain the image embedding amount difference of each frame in the contraction-diastole fluctuation cycle. The image embedding amount difference of each frame is averaged to obtain the mean value of the structural alignment error of the cycle.

5. The method for detecting abnormalities in cardiac ultrasound data according to claim 1, wherein: The structural period deviation judgment result of each contraction-diastole fluctuation period is obtained, specifically including: The mean value of the structural alignment error of each contraction-diastole fluctuation cycle is compared with the structural alignment error mean threshold value stored in the database. If the mean value of the structural alignment error of a contraction-diastole fluctuation cycle is greater than or equal to the structural alignment error mean threshold value, the structural period deviation of the contraction-diastole fluctuation cycle is judged to exist. If the mean value of the structural alignment error of a certain contraction-diastole fluctuation cycle is less than the structural alignment error mean threshold value, the structural cycle deviation judgment result of the contraction-diastole fluctuation cycle is that there is no structural cycle deviation.

6. The method for detecting abnormalities in cardiac ultrasound data according to claim 1, wherein: The determination of whether to enter the rhythm abnormality determination process based on the structural period deviation determination result specifically includes: If a structural cycle deviation exists in a certain systolic-diastolic fluctuation cycle, the image embedding amount difference of each frame of the systolic-diastolic fluctuation cycle is obtained, and each frame with an image embedding amount difference greater than or equal to the image embedding amount difference threshold is selected as a deviation frame. The consecutive number of deviation frames of the systolic-diastolic fluctuation cycle is counted. If the consecutive number of deviation frames is greater than or equal to a preset consecutive number threshold, the structural cycle deviation of the systolic-diastolic fluctuation cycle is defined as a severe deviation, and an early warning is issued, and the rhythm abnormality judgment process is not entered; If the continuous number of deviation frames is less than the preset continuous number threshold, the structural periodic deviation of the contraction-diastole fluctuation cycle is defined as a mild deviation, the continuous number of deviation frames and the distribution skewness are obtained, and the continuous number of deviation frames is put into the mapping set of continuous number-image enhancement parameters pre-stored in the database for mapping and matching to obtain the image enhancement parameters of the contraction-diastole fluctuation cycle, and the image enhancement of the contraction-diastole fluctuation cycle is performed with the image enhancement parameters of the contraction-diastole fluctuation cycle, and the distribution skewness is put into the mapping set of distribution skewness-timing position adjustment parameters pre-stored in the database to obtain the timing position adjustment parameters of the contraction-diastole fluctuation cycle, and the timing position adjustment parameters of the contraction-diastole fluctuation cycle are used to adjust the timing positions of each deviation frame of the contraction-diastole fluctuation cycle. After completing the image enhancement and timing position adjustment, the rhythm abnormality judgment process is entered.

7. The method for detecting abnormalities in cardiac ultrasound data according to claim 1, wherein: The analysis of the image embedding feature parameters of each systolic-diastolic fluctuation cycle and the introduction of the influence of the structural cycle deviation judgment result to obtain the rhythm abnormality judgment result of each systolic-diastolic fluctuation cycle specifically includes: The embedded information parameters of each frame within each systolic-diastolic fluctuation cycle were extracted, including the embedding norm, characteristic entropy, and principal component projection amplitude of each frame. The embedded information parameters were arranged in a frame sequence within the cycle, and time series analysis was performed to calculate the index parameters of the internal characteristic curve of each systolic-diastolic fluctuation cycle, including the first-order derivative, local volatility, and principal component reconstruction error. If the structural cycle deviation judgment result of a certain systolic-diastolic fluctuation cycle is that there is a structural cycle deviation, the continuous number of deviation frames of the systolic-diastolic fluctuation cycle is subtracted from the continuous number threshold to obtain a continuous number difference, and the continuous number difference-penalty factor mapping set pre-stored in the database is input for mapping matching to obtain the penalty factor of the systolic-diastolic fluctuation cycle; The index parameters of the internal characteristic curve are modified based on the penalty factor of the contraction-diastole fluctuation cycle; If any indicator of a certain systolic-diastolic fluctuation cycle exceeds the corresponding preset indicator threshold, the rhythm abnormality judgment result of the systolic-diastolic fluctuation cycle is determined to be rhythm abnormality; If all indicators of a certain systolic-diastolic fluctuation cycle do not exceed the corresponding preset indicator thresholds, the rhythm abnormality judgment result of the systolic-diastolic fluctuation cycle is determined to be normal rhythm.

8. The method for detecting abnormalities in cardiac ultrasound data according to claim 1, wherein: The corresponding operation is performed based on the rhythm abnormality judgment result, and the specific processing conditions are: If the rhythm abnormality judgment result of the systolic-diastolic fluctuation cycle is a rhythm abnormality, all the systolic-diastolic fluctuation cycles with rhythm abnormality judgment results as rhythm abnormality are counted. If the number of consecutive cycles is greater than the preset abnormality threshold, the rhythm abnormality is determined to be a global rhythm abnormality, and the main frequency matching loss mechanism is performed; If the number of consecutive cycles is less than or equal to the preset abnormal threshold, the rhythm abnormality is determined to be a local cycle abnormality, and all contraction-diastole fluctuation cycles with rhythm abnormality judgment results as rhythm abnormality are marked as abnormal cycles, and an early warning is issued.

9. The method for detecting abnormalities in cardiac ultrasound data according to claim 8, wherein: The main frequency matching loss mechanism specifically includes: By performing Fourier transform on the inter-frame dynamic parameter sequence of each systolic-diastolic fluctuation cycle, the dominant frequency of each frame is extracted, and the spectrum multi-peak ratio of each systolic-diastolic fluctuation cycle is obtained. If the spectrum multi-peak ratio of a systolic-diastolic fluctuation cycle is greater than a preset spectrum multi-peak ratio threshold, it is marked as an abnormal continuous systolic-diastolic fluctuation cycle and an early warning is issued. If the spectrum multipeak ratios of all systolic-diastolic fluctuation cycles are less than or equal to the spectrum multipeak ratio threshold, then the main frequency of each frame in each systolic-diastolic fluctuation cycle is compared with the main frequency with the highest proportion in the entire cycle. Based on the comparison results, a loss factor is constructed, and the loss factor of each frame is compared with the loss factor threshold pre-stored in the database. If the loss factor of each frame does not exceed the threshold, it is put into the correct rhythm frequency generation algorithm, and the auxiliary algorithm generates the rhythm frequency and corrects each frame; If the loss factor of a certain frame in a certain systolic-diastolic fluctuation cycle exceeds the threshold, an early warning is issued.

10. A system using the method for detecting abnormalities in cardiac ultrasound data according to any one of claims 1 to 9, characterized in that: A feature extraction module is used to perform dual-channel feature extraction and detection on the original cardiac B-ultrasound image sequence to obtain inter-frame dynamic feature values ​​and image embedding feature values ​​of each frame in the original cardiac B-ultrasound image sequence; A cycle recognition module is used to identify each systolic-diastolic fluctuation cycle in the original cardiac B-ultrasound image sequence based on the inter-frame dynamic features and image embedding feature values ​​of each frame; A structural cycle deviation judgment module is used to obtain the mean value of the structural alignment error of each systolic-diastolic fluctuation cycle, obtain the structural cycle deviation judgment result of each systolic-diastolic fluctuation cycle, and decide whether to enter the rhythm abnormality judgment process based on the structural cycle deviation judgment result; The rhythm abnormality judgment module analyzes the image embedding feature parameters of each contraction-diastole fluctuation cycle, obtains the rhythm abnormality judgment results of each contraction-diastole fluctuation cycle after introducing the influence of the structural cycle deviation judgment results, and performs corresponding operations based on the rhythm abnormality judgment results.

Images