Urine flow dynamics pattern clustering based bladder function typing and abnormality recognition method
By introducing pattern clustering with strategy substitution points and substitution consistency constraints into urodynamic analysis, the problem of misjudgment of mechanism interruption and substitution phenomena during urination in existing technologies is solved, and accurate classification and abnormal identification of bladder function are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ENSHI TUJIA & MIAO AUTONOMOUS PREFECTURE CENT HOSPITAL
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-21
AI Technical Summary
Existing urodynamic analysis methods lack the ability to characterize dynamic changes in mechanisms when dealing with interruptions and substitutions during voiding, resulting in insufficient accuracy and stability in classification and a tendency to misinterpret signal changes in different voiding stages or pathological mechanisms.
By introducing a strategy to locate the interruption point of the urination mechanism, and combining it with a pattern clustering algorithm with alternative consistency constraints, the urodynamic examination sequence is decomposed into fracture pattern units, and clustered based on time and mechanism consistency to identify abnormal fractures.
It enables accurate localization of the interruption and substitution of mechanisms during urination, improves the reliability and stability of the typing results, and can accurately identify bladder dysfunction in complex scenarios.
Smart Images
Figure CN122432862A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of pattern classification technology, and in particular to a method for bladder function classification and abnormality identification based on urodynamic pattern clustering. Background Technology
[0002] Urodynamic testing is an important tool for assessing bladder storage and voiding functions. By recording multi-channel signals such as urinary flow rate, bladder pressure, abdominal pressure, and electromyography, the voiding process is analyzed to aid in the diagnosis of various lower urinary tract dysfunctions, including bladder outlet obstruction, detrusor muscle weakness, and sphincter dyssynergia. Current clinical analysis methods mainly rely on the interpretation of urinary flow rate curve morphology and several key parameters, such as maximum urinary flow rate, voiding time, pressure level, and residual urine volume, combined with the physician's experience for comprehensive judgment.
[0003] With the development of data processing technology, some solutions have attempted to introduce pattern recognition or clustering algorithms to automatically classify urodynamic data. However, existing technologies generally still model based on the overall signal morphology or static feature combination, and their analysis objects are usually the complete urination process or signal segments within a fixed time window. They lack the ability to characterize the dynamic mechanism changes during urination, resulting in limited classification accuracy and stability in complex scenarios.
[0004] Specifically, the phenomenon of "mechanism interruption and substitution" is common during urination. This means that when the original urination mechanism fails at a certain moment, the body compensates or substitutes by engaging abdominal pressure, intermittent urination, or changes in muscle coordination. This type of mechanism switching usually only occurs within a local timeframe, but existing methods often analyze overall curves or global features, making it difficult to pinpoint the specific location of the mechanism change or distinguish the sequential relationship between different mechanisms during the same urination process. Secondly, existing pattern clustering or classification methods primarily rely on signal morphological similarity for merging, neglecting the phase attributes of urination behavior on the time axis and the physiological coordination relationships between different signals. In practical applications, different urination stages may exhibit morphologically similar but semantically different signal changes. Direct clustering can easily lead to erroneous merging of data from different stages or different pathological mechanisms, thus affecting the reliability of the classification results. Summary of the Invention
[0005] This invention provides a bladder function classification and abnormality identification method based on urodynamic pattern clustering. It introduces the characterization of the location of urination mechanism interruption, alternative behavior and their sequential relationship in urodynamic analysis, and on this basis, constructs a clustering and classification method with time constraints and mechanism constraints, while realizing the effective distinction between real abnormality and normal regulatory behavior.
[0006] A method for bladder function classification and abnormality identification based on urodynamic pattern clustering includes the following steps:
[0007] S1. Input the urodynamic examination sequence of the target subject, identify the location where the original urination mechanism is interrupted during urination, generate the corresponding strategy substitution point sequence, and output the fracture mode unit with the strategy substitution point sequence as a constraint.
[0008] S2. Input the fracture pattern unit, and based on the existing pattern clustering algorithm, introduce the substitution consistency constraint so that only fracture pattern units with the same substitution direction and fracture position participate in the same type merging, and output the fracture pattern cluster and the corresponding bladder function classification result.
[0009] S3. Input the bladder function classification result and the fracture pattern unit, identify abnormal fractures that do not conform to the bladder function classification result, and output the abnormal identification result based on the continuity relationship of the abnormal fractures before and after replacement.
[0010] Optionally, the urodynamic examination sequence is decomposed into time-series data of multi-channel signals, including urinary flow rate signal, bladder pressure signal, abdominal pressure signal, and electromyography signal.
[0011] Optionally, S1 further includes a preset interruption detection threshold. Based on the preset interruption detection threshold, a mutation point is detected for each channel signal. Any location where a mutation occurs and causes the urination mechanism to deviate from the normal physiological curve is marked as the original interruption location. For each original interruption location, a strategy substitution point is generated by interpolation or pattern matching using adjacent normal urination segments. All strategy substitution points are arranged in chronological order to form a strategy substitution point sequence.
[0012] Optionally, each strategy substitution point in the strategy substitution point sequence serves as a segmentation boundary, dividing the original urodynamic examination sequence into multiple sub-segments, with each sub-segment output as a break mode unit.
[0013] Optionally, S2 further includes parsing each fracture pattern unit into a feature vector including fracture location labels and strategy substitution point direction labels, transforming the original complex signal fragments in the fracture pattern unit into a structured representation, so that the pattern clustering algorithm can effectively merge based on location and mechanism consistency.
[0014] Optionally, the substitution consistency constraint includes: allowing two fracture pattern units to participate in the same category of merging calculation only when the interruption time windows corresponding to the fracture location labels of the two fracture pattern units overlap or are adjacent and the policy substitution point direction labels of the two are the same; performing clustering on all fracture pattern units that satisfy the substitution consistency constraint, outputting one or more fracture pattern clusters, and mapping the dominant fracture pattern feature in each fracture pattern cluster to the corresponding bladder function classification result.
[0015] Optionally, the break location label indicates the start and end break locations of the break mode unit in the original urodynamic examination sequence, and the strategy substitution point direction label indicates the substitution operation type corresponding to the strategy substitution point within the break mode unit.
[0016] Optionally, S3 includes calculating the matching degree between the feature vector of each fracture mode unit and the typical pattern features in the bladder function classification result. If the matching degree is lower than a preset classification threshold, the current fracture mode unit is marked as an abnormal fracture.
[0017] Optionally, for each fracture pattern unit marked as the abnormal fracture, the pre-substitution segment and post-substitution segment corresponding to it in the strategy substitution point sequence are extracted, and the urodynamic parameter continuity index between the end of the pre-substitution segment and the beginning of the post-substitution segment is calculated. When the continuity index exceeds a preset abnormal continuity threshold, the abnormal fracture is determined to have a real abnormality in a pathological or physiological sense, and the abnormal identification result is output according to the position of the abnormal fracture in the original sequence and the continuity relationship between the pre-substitution and post-substitution segments.
[0018] Optionally, the continuity indicators include at least one of pressure change rate, urine flow rate jump amplitude, and electromyographic signal mutation degree; the abnormality identification results include at least the abnormality type, the time interval of the abnormality occurrence, and the severity level of the abnormality.
[0019] The beneficial effects of this invention are:
[0020] This invention, through strategic substitution point localization, decomposes continuous urodynamic examination sequences into fracture pattern units with clear physiological significance, transforming the traditional overall curve analysis of the urination process into an analysis of mechanistic changes. Compared to existing technologies that rely on the morphology of the urinary flow rate curve or a few statistical indicators for classification, this invention can clearly distinguish when different urination mechanisms are interrupted and how they are replaced, fundamentally avoiding misjudgments due to similar morphologies but different causes. Especially in complex scenarios involving multiple intertwined mechanisms such as detrusor weakness, abdominal pressure compensation, and synergistic mismatch, it can accurately identify the location and evolution of fractures in various mechanisms.
[0021] This invention, based on traditional pattern clustering algorithms, introduces a substitution consistency constraint. Only fracture pattern units with similar fracture locations and consistent substitution operation types are allowed to participate in clustering within the same category. This elevates the clustering criterion from simple signal morphology similarity to a joint constraint of temporal location consistency and voiding mechanism consistency. This solves the problems of mixed data from different voiding stages and mis-merging of different pathological mechanisms when existing clustering methods are applied to this invention, ensuring clear physiological stage and mechanism consistency within the same fracture pattern cluster. The resulting fracture pattern clusters more closely resemble actual voiding behavior patterns, and their dominant fracture pattern features stably reflect the corresponding bladder function state, significantly improving the reliability and stability of the classification results.
[0022] In the process of anomaly identification, this invention first compares the feature matching degree with the classification results to screen out the fracture pattern units that do not conform to the dominant pattern; then, it uses the continuity index to quantitatively evaluate the continuity of the signal before and after the strategy substitution point, thereby determining whether the fracture constitutes a real physiological or pathological abnormality. This two-layer judgment mechanism can effectively distinguish between normal strategy adjustment behavior during urination and abnormal fractures with actual clinical significance, avoiding misjudging short-term fluctuations or slight compensation as abnormalities. Attached Figure Description
[0023] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;
[0025] Figure 2 This is a schematic diagram illustrating the process of obtaining the fracture mode clusters and corresponding bladder function classification results in an embodiment of the present invention. Detailed Implementation
[0026] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.
[0027] like Figures 1-2 As shown, the bladder function classification and abnormality identification method based on urodynamic pattern clustering includes the following steps:
[0028] S1. Input the urodynamic examination sequence of the target subject, identify the location where the original urination mechanism is interrupted during urination, generate the corresponding strategy substitution point sequence, and output the break mode unit with the strategy substitution point sequence as a constraint.
[0029] S11, the urodynamic examination sequence is decomposed into multi-channel time-series data of urinary flow rate signal, bladder pressure signal, abdominal pressure signal, and electromyography signal in chronological order, as shown below: ;in, This indicates the urodynamic testing sequence. Represents a time series index. The urine flow rate signal indicates the change in the amount of urine excreted per unit time, and is used to reflect whether urination is smooth. This is a bladder pressure signal, indicating pressure changes within the bladder, primarily reflecting the contraction of the detrusor muscle. This is an abdominal pressure signal, indicating changes in the pressure applied to the abdomen, used to determine whether compensatory behaviors such as straining during urination are present. Electromyography (EMG) signals indicate the electrical activity of the pelvic floor or sphincter muscles, reflecting whether the muscles are coordinated and relaxed during urination.
[0030] S12. During urination, signals such as urine flow rate, bladder pressure, abdominal pressure, and electromyography (EMG) have a synergistic relationship. Under normal circumstances, bladder pressure rises—spontaneous relaxation—urine flow appears and gradually stabilizes. When this synergistic relationship is broken, it means that the urination mechanism has been interrupted or switched. Therefore, at each time point, a change amplitude is calculated for each type of signal, and the curve shape is converted into change intensity. All signals are compared at the same time. If the change of a certain signal suddenly becomes very large, it indicates that an abnormal fluctuation may have occurred at that moment. However, this invention does not treat all sudden changes as abnormal. Instead, it sets an interruption detection threshold. Only when the change intensity at a certain moment exceeds the interruption detection threshold, and this change can cause urination behavior to deviate from the normal physiological pattern, such as sudden interruption of urine flow, abnormal pressure jump, or abnormal activation of EMG, is it identified as the original interruption position, avoiding misjudging normal small fluctuations as abnormalities.
[0031] Specifically, based on a preset interruption detection threshold, abrupt change points are detected for each channel signal, and the intensity of change in each channel signal is calculated. ;in, Indicates the first Each channel signal, and , Indicates time interval, Indicates the first Each channel at time The intensity of change.
[0032] The original interrupt location is marked when the following conditions are met: ;in, This indicates the interruption detection threshold, which is set based on the actual signal characteristics. Specifically, it is determined based on baseline fluctuations. First, the natural fluctuation range during normal urination is statistically analyzed, and then the threshold is set to 2 to 3 times this fluctuation range to distinguish between normal fluctuations and abnormal abrupt changes. This indicates that the maximum value is taken for the change intensity of all channels.
[0033] When the criteria for deviation of the urination mechanism from the normal physiological curve are met simultaneously, the original set of interruption locations is obtained:
[0034] ;in, Represents the original set of interrupt locations. Indicates the first The original interrupt location.
[0035] S13. For each original interruption location, a strategy substitution point is generated by pattern matching using adjacent normal urination segments. The strategy substitution point is understood as the key time point at which the urination mechanism changes. That is, before the strategy substitution point, one urination mode is used, such as detrusor muscle dominance; after the strategy substitution point, another mode is used, such as abdominal pressure compensation.
[0036] For example:
[0037] The contraction changes from normal urination to straining;
[0038] From continuous urination to intermittent urination;
[0039] Sphincter interference can occur during coordinated urination.
[0040] Therefore, the strategic substitution point is the location where the urination behavior undergoes a strategic switch.
[0041] Specifically, it is expressed as follows: This formula represents finding a time point within a certain time range near the original interruption location that most closely resembles a normal urination state, and using this time point as the true strategy replacement point. When an original interruption location is detected, this point is often just a signal abrupt change, not necessarily the exact location where the urination mechanism has actually changed; it may occur earlier or later. Therefore, candidate time windows are selected near the interruption point, and then each time point within the window is compared to see which of the multi-channel signal states corresponding to each time point is closer to the state that should exist during normal urination. Finally, the time point that is closest to the normal urination pattern is selected as the strategy replacement point. Indicates the first A strategy alternative point, Indicates the original interrupt location Nearby candidate time windows, specifically based on the original interruption location Centered on the candidate time window, the time window is set to 2 seconds forward and 2 seconds backward. This represents a reference pattern generated from adjacent normal urination segments. The pattern difference measurement function is described, and the specific measurement steps are as follows:
[0042] 1. Input the multi-channel signal status at the current time point and the reference mode for normal urination status.
[0043] 2. Align the current signal with the reference mode on the time scale and normalize each channel.
[0044] 3. Calculate the following differences respectively:
[0045] Differences in urine flow rate: Are there sudden drops or discontinuities in flow rate?
[0046] Bladder pressure differences: Do they conform to normal driving trends?
[0047] Differences in abdominal pressure: Is there any abnormal compensation?
[0048] Electromyographic differences: whether there is abnormal contraction or lack of relaxation;
[0049] 4. Each type of difference yields a difference value.
[0050] 5. The differences between the channels are weighted and fused to obtain an overall difference result. The smaller the difference, the closer it is to the normal urination pattern. The difference value at this time point is taken as the degree of pattern deviation at that time.
[0051] Then, all strategy substitution points are arranged in chronological order to form a strategy substitution point sequence:
[0052] ;in, This indicates the strategy substitution point sequence. This indicates the number of points that can be replaced by the strategy.
[0053] It should also be noted that the reference pattern is not a fixed template, but rather a normal urination pattern adaptively generated from the current data. Details are as follows:
[0054] 1. Locate stable urination segments before and after the original interruption location. Screening criteria include continuous urine flow, smooth pressure changes, and normal electromyographic coordination.
[0055] 2. Extract adjacent normal segments: Select a normal segment before the break point and select a segment that recovers to normal after the break point.
[0056] 3. Map multiple normal segments to a uniform time length to eliminate the influence of differences in urination speed.
[0057] 4. Average or weighted merge multiple normal segments to form a standard normal urination pattern, which represents the urination behavior that should occur if no abnormality occurs.
[0058] S14, using each strategy substitution point in the strategy substitution point sequence as a segmentation boundary, divides the original urodynamic examination sequence into multiple sub-segments: ;in, Indicates the first One fracture mode unit, Indicates the first The location of each strategy substitution point on the timeline. Indicates the first The location of each strategy substitution point on the timeline. Indicates time interval Sequence truncation.
[0059] All sub-segments constitute a set of fracture mode units: ;in, Represents the set of fracture mode elements. Indicates the number of fracture mode elements.
[0060] S2. Input the fracture pattern unit, and based on the existing pattern clustering algorithm, introduce the substitution consistency constraint so that only fracture pattern units with the same substitution direction and fracture position participate in the same type merging, and output the fracture pattern cluster and the corresponding bladder function classification result.
[0061] S21, each fracture pattern unit is transformed from the original signal segment into a structured label representation for subsequent clustering. Specifically, each fracture pattern unit is parsed into a feature vector containing fracture location labels and policy substitution point direction labels, represented as: ;in, Indicates the first Feature vectors of each fracture mode unit The label indicates the fracture location, which is the timeline position, i.e., the time of fracture. This indicates a strategy alternative operation type label.
[0062] The fracture location label is indicated as follows: ;in, Indicates the first The starting interrupt position corresponding to each fracture mode unit. Indicates the first The termination / interruption position corresponding to each fracture mode unit.
[0063] The direction label for the strategy substitution point is represented as follows: ;in, Indicates the first The strategy substitution operation type for each fracture mode unit. This represents the set of preset strategy substitution types. The strategy substitution operation type refers to the category of new urination methods adopted after the original urination mechanism fails during urination, including abdominal pressure compensation type, intermittent urination type, coordination mismatch type, low drive attenuation type, and outlet blockage response type.
[0064] Specifically, the analysis of fracture mode elements is achieved through the following steps:
[0065] 1. Input a fracture mode unit, obtain the start and end interrupt positions corresponding to the unit, and output the fracture position label start time and end time.
[0066] 2. Within the fracture mode unit, find the strategy substitution point, extract the signal state before substitution and the signal state after substitution respectively, and focus on the trend of urine flow change, bladder pressure change, abdominal pressure participation, and whether electromyography is abnormal.
[0067] 3. Input a fracture mode unit, locate its internal strategy substitution point, and extract the pre-substitution state and post-substitution state using this strategy substitution point as the boundary. The pre-substitution state characterizes the original urination mechanism, and the post-substitution state characterizes the urination mechanism after substitution. Compare the pre-substitution and post-substitution states to extract multi-channel coordinated change relationships. These coordinated changes include at least the following determination information:
[0068] Whether the trend of urinary flow rate change is interrupted or re-established;
[0069] Does bladder pressure continuously provide the driving force?
[0070] Whether abdominal pressure changes from being uninvolved to being involved or increased;
[0071] Whether electromyography (EMG) changes from a relaxed state to an activated state.
[0072] Based on the above changes, a description of the changes in the urination mechanism of this fracture mode unit is formed.
[0073] The description of the change in the urination mechanism is input into a preset set of rules for determining alternative operation types. The rules are then used to match the corresponding alternative operation types. These rules include:
[0074] If abdominal pressure changes from a low level to significant involvement, and urine flow depends on abdominal pressure for maintenance, it is determined to be abdominal pressure compensatory type.
[0075] If the urine flow is interrupted and then restarts within a short period of time, it is classified as intermittent urination.
[0076] If abnormal activation of electromyographic signals occurs during urination, it is determined to be a synergistic mismatch type.
[0077] If the bladder pressure gradually decreases and the urine flow weakens synchronously, it is determined to be a low-drive attenuation type;
[0078] If bladder pressure is elevated but urine flow is restricted, it is classified as an outlet block response type.
[0079] If urine flow is restored but depends on abnormal drivers for maintenance, it is determined to be a pseudo-restoration compensatory type;
[0080] The matched alternative operation type is output as the policy alternative operation type label for the fracture pattern unit, which is used for subsequent fracture pattern clustering and bladder function classification.
[0081] 4. Constructing feature vectors: Combine the two types of information to form fracture location labels and strategy substitution operation type labels, thus completing the construction of feature vectors.
[0082] S22. Traditional pattern clustering algorithms cluster all samples and then automatically group them. However, in the scenario of this invention, two curves may appear to follow the same pattern, but one occurs at the beginning of urination and the other at the end, resulting in completely different meanings. Alternatively, they may have similar shapes, but one represents abdominal pressure compensation and the other represents detrusor muscle attenuation, indicating fundamentally different mechanisms. When applying this invention to traditional clustering, the time position or physiological mechanism may be ignored, easily leading to erroneous clustering that results in similar shapes but fundamentally different outcomes.
[0083] Therefore, based on existing pattern clustering algorithms, an alternative consistency constraint is added.
[0084] Condition 1 requires that the time positions be consistent or close: the two break mode units must occur near the same stage in the urination process, specifically in that the time windows overlap or are adjacent in time. Setting this condition avoids mixing up abnormalities in the initiation stage and abnormalities in the end of emptying.
[0085] Condition 2: The alternative operation type must be the same: the two break mode units must belong to the same type of voiding strategy change, such as both being abdominal pressure compensation, both being intermittent voiding, or both being synergistic mismatch, to avoid mixing different pathological mechanisms into one category.
[0086] Specifically, two fracture mode units are defined. and The decision function for a function to participate in merging functions of the same type is:
[0087] ;
[0088] in, Represents fracture mode unit and Does the alternative consistency constraint satisfy? This indicates that the interrupt time windows corresponding to two break location labels overlap. Adjacent means that the two interrupt time windows are immediately adjacent on the time axis and the interval does not exceed a preset time threshold, which is 0.6-0.8 seconds. This indicates that the policy substitution point direction labels of the two fracture mode units are the same.
[0089] Only if the following conditions are met: At that time, the fracture mode unit is allowed. and It participates in the merging calculation of the same category.
[0090] S23, perform clustering on all fracture mode units that satisfy the alternative consistency constraint to obtain a fracture mode cluster set: ;in, Represents the set of fracture pattern clusters. Indicates the first A cluster of fracture patterns, Indicates the number of fracture pattern clusters; perform clustering as follows:
[0091] 1. Input all fracture mode units. Based on the aforementioned substitution consistency constraint, perform pairwise judgment on any two fracture mode units, retain unit pairs that satisfy the conditions of consistent or adjacent time positions and the same strategy substitution operation type, and output the set of mergeable relations that satisfy the constraints.
[0092] 2. Treat each fracture mode unit as a node, and establish a connection between two units that satisfy the substitution consistency constraint to form a constrained connection structure, which can be understood as a grouped network.
[0093] 3. Starting with each unclassified fracture mode unit, expand and merge all fracture mode units that are connected to it, gradually merging all interconnected units. Each group of interconnected fracture mode units forms a fracture mode cluster, and the fracture mode cluster set is output.
[0094] For each fracture mode cluster, the dominant fracture mode feature is extracted and represented as:
[0095] ;in, Indicates the first The dominant fracture mode feature of a fracture mode cluster is defined by `mode()`, which represents the most frequent feature combination in the cluster. Specifically, for all fracture mode units in a fracture mode cluster, their corresponding feature vectors are extracted, namely fracture location labels and policy substitution operation type labels; the occurrence frequency of each policy substitution operation type in the cluster is counted; the most frequent operation type is found; the temporal location distribution of each fracture mode unit is calculated to determine the time interval with the most concentrated occurrence; the most frequent policy substitution operation type and the fracture location interval with the most concentrated occurrence are combined to form the dominant fracture mode feature of the fracture mode cluster.
[0096] Mapping the dominant fracture pattern features to the corresponding bladder function classification results: ;in, Indicates the first Bladder function classification results corresponding to each fracture pattern cluster. This represents the mapping function from fracture pattern characteristics to bladder function classification. The specific mapping scheme is as follows:
[0097] 1. Input the dominant fracture mode features of a fracture mode cluster. The dominant fracture mode features include fracture location features and strategy substitution operation type.
[0098] 2. Based on the fracture location characteristics, determine which of the following urination stages the fracture mode mainly occurs in, including the urination initiation stage, the urination stabilization stage, and the urination decay stage, and output the stage attribute label.
[0099] 3. Input the stage attributes and strategy substitution operation type into the preset bladder function classification mapping rules, and match the corresponding bladder function classification according to the rules. An example of the mapping rules is as follows:
[0100] If it is in the initiation stage and the alternative operation type is abdominal pressure compensation, it is mapped to urination initiation difficulty type;
[0101] If it is in a stable phase and the type of alternative operation is synergistic mismatch, then it is mapped to sphincter synergistic dysfunction.
[0102] If it is in a stable phase and the alternative operation type is outlet block response, then it is mapped to bladder outlet obstruction type.
[0103] If it is in the decay phase and the replacement operation type is low-drive decay type, it is mapped to detrusor weakness type.
[0104] If intermittent urination occurs and spans multiple phases, it is classified as unstable urination.
[0105] If a pseudo-recovery compensatory type occurs, it is mapped to a compensatory anomalous type.
[0106] 4. The matched functional category is output as the bladder function classification result for that fracture pattern cluster. If multiple fracture pattern clusters exist for the same subject, the main classification or combined classification result is output according to the degree of dominance or frequency of occurrence.
[0107] S3. Input the bladder function classification result and the fracture pattern unit, identify abnormal fractures that do not conform to the bladder function classification result, and output the abnormal identification result based on the continuity relationship of the abnormal fractures before and after replacement.
[0108] S31, the matching degree between the feature vector of each fracture mode unit and the typical pattern features in the bladder function classification results is calculated, and expressed as: ;in, Indicates the first The matching degree of each fracture mode unit, The function for calculating feature similarity takes the current fracture mode unit feature vector and the typical pattern features of the corresponding subtype as input. It calculates positional consistency by determining whether the fracture location falls within the typical position interval. If it does, the consistency score is high; otherwise, it is low. Next, it calculates operation type consistency by determining whether the strategy substitution operation type is consistent. Consistency results in a high score, while inconsistency results in a low score. The positional consistency score and operation type consistency score are combined to obtain an overall matching degree, which is then output as the matching degree value. Indicates the first Feature vectors of each fracture mode unit This represents the typical pattern characteristics corresponding to the bladder function classification results.
[0109] When the following conditions are met: When this fracture mode element is identified, it is marked as an anomalous fracture. ;
[0110] in, This represents the preset fractal threshold, ranging from 0.6 to 0.75, with 0.7 being the preferred value. This represents the fracture mode element that is marked as an anomalous fracture.
[0111] Typical pattern features corresponding to bladder function classification results are obtained through the following steps:
[0112] 1. Input the multiple fracture pattern clusters that have been obtained, classify them according to the bladder function classification results, and aggregate all fracture pattern clusters belonging to the same classification category to obtain a set of samples corresponding to each functional classification.
[0113] 2. For each fracture mode cluster, extract its dominant fracture mode features, i.e., the results obtained in S2.
[0114] The dominant features include fracture location features and strategy substitution operation type, resulting in a set of candidate representative features for this category.
[0115] 3. Statistically analyze the frequency of each strategy substitution operation type in the category, identify the most frequent operation type, and simultaneously analyze the distribution range of the fracture location to identify the time period in which the fracture occurs most frequently, thereby obtaining the most typical mechanism type and the most typical occurrence stage.
[0116] 4. Combine the most common operation types and the most concentrated locations to form the typical pattern features corresponding to this bladder function classification, and output the standard reference features of this classification. .
[0117] S32, for each fracture mode unit marked as an anomalous fracture, extract its pre-substitution segment and post-substitution segment in the strategy substitution point sequence, represented as follows:
[0118] ;
[0119] ;
[0120] in, Indicates the first The pre-substitution segment corresponding to each fracture mode unit Indicates the first The replaced fragments corresponding to each fracture mode unit This indicates the strategy substitution point corresponding to the fracture mode element. Indicates the start time of the segment before replacement. Indicates the end time of the replaced segment. Indicates time interval Extraction of urodynamic sequences from the above. That is to say , .
[0121] The continuity index of urodynamic parameters between the end of the pre-replacement segment and the beginning of the post-replacement segment is calculated and expressed as: ;in, Indicates the first The continuity index of each fracture mode unit. Indicates the rate of change of pressure. Indicates the amplitude of the urine flow rate jump. Indicates the degree of abrupt change in electromyographic signals. This represents the weighting coefficient of each indicator, with values of 0.4, 0.4, and 0.2.
[0122] Each component is specifically represented as follows:
[0123] ;in, This represents the bladder pressure signal at the policy substitution point. Indicates time interval, This represents the bladder pressure signal value at the time interval preceding the strategy substitution point. Indicates the first The time position of the strategy substitution point corresponding to each fracture mode unit;
[0124] ;in, This represents the urine flow rate value at the strategy substitution point. This represents the urine flow rate value at the time interval preceding the strategy substitution point.
[0125] ;in, This represents the electromyographic signal value at the strategy substitution point. This represents the electromyographic signal value at the time interval preceding the strategy substitution point.
[0126] S33, when the following conditions are met: When this abnormal fracture is determined to be a true abnormality in a pathological or physiological sense: ;
[0127] in, This represents a preset threshold for abnormal continuation, ranging from 0.5 to 0.7, preferably 0.6. It is used to determine whether there is a genuine physiological break before and after the strategy substitution point, rather than noise or normal fluctuations. This represents a fracture mode element that has real anomaly significance.
[0128] Based on the location of the abnormal break in the original sequence and the continuity relationship between the segments before and after the replacement, the anomaly identification result is output, represented as: ;in, Indicates the first Anomaly identification results for individual fracture mode units. Indicates the exception type. Indicates the time interval in which the anomaly occurred. Indicates the level of severity of the abnormality.
[0129] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.
[0130] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for bladder function classification and abnormality identification based on urodynamic pattern clustering, characterized in that, Includes the following steps: S1. Input the urodynamic examination sequence of the target subject, identify the location where the original urination mechanism is interrupted during urination, generate the corresponding strategy substitution point sequence, and output the fracture mode unit with the strategy substitution point sequence as a constraint. S2. Input the fracture pattern unit, and based on the existing pattern clustering algorithm, introduce the substitution consistency constraint so that only fracture pattern units with the same substitution direction and fracture position participate in the same type merging, and output the fracture pattern cluster and the corresponding bladder function classification result. S3. Input the bladder function classification result and the fracture pattern unit, identify abnormal fractures that do not conform to the bladder function classification result, and output the abnormal identification result based on the continuity relationship of the abnormal fractures before and after replacement.
2. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 1, characterized in that, The urodynamic examination sequence is decomposed into time-series data of multi-channel signals, which include urinary flow rate signal, bladder pressure signal, abdominal pressure signal and electromyography signal.
3. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 2, characterized in that, S1 further includes a preset interruption detection threshold. Based on the preset interruption detection threshold, a mutation point is detected for each channel signal. Any location where a mutation occurs and causes the urination mechanism to deviate from the normal physiological curve is marked as the original interruption location. For each original interruption location, a strategy substitution point is generated by interpolation or pattern matching using adjacent normal urination segments. All strategy substitution points are arranged in chronological order to form a strategy substitution point sequence.
4. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 3, characterized in that, Each strategy substitution point in the strategy substitution point sequence serves as a segmentation boundary, dividing the original urodynamic examination sequence into multiple sub-segments, with each sub-segment being output as a break mode unit.
5. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 1, characterized in that, The S2 further includes parsing each fracture pattern unit into a feature vector including fracture location labels and strategy substitution point direction labels, transforming the original complex signal fragments in the fracture pattern unit into a structured representation, so that the pattern clustering algorithm can effectively merge based on location and mechanism consistency.
6. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 5, characterized in that, The substitution consistency constraint includes: two fracture pattern units are allowed to participate in the same category of merging calculation only when the interruption time windows corresponding to the fracture location labels of the two fracture pattern units overlap or are adjacent and the policy substitution point direction labels of the two are the same; clustering is performed on all fracture pattern units that satisfy the substitution consistency constraint, one or more fracture pattern clusters are output, and the dominant fracture pattern feature in each fracture pattern cluster is mapped to the corresponding bladder function classification result.
7. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 6, characterized in that, The break location label indicates the start and end break positions of the break mode unit in the original urodynamic examination sequence, and the strategy substitution point direction label indicates the substitution operation type corresponding to the strategy substitution point within the break mode unit.
8. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 1, characterized in that, S3 includes calculating the matching degree between the feature vector of each fracture mode unit and the typical pattern features in the bladder function classification result. If the matching degree is lower than the preset classification threshold, the current fracture mode unit is marked as an abnormal fracture.
9. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 8, characterized in that, For each fracture pattern unit marked as the abnormal fracture, the pre-substitution segment and post-substitution segment corresponding to it in the strategy substitution point sequence are extracted. The continuity index of urodynamic parameters between the end of the pre-substitution segment and the beginning of the post-substitution segment is calculated. When the continuity index exceeds the preset abnormal continuity threshold, the abnormal fracture is determined to have a real abnormality in a pathological or physiological sense. Based on the position of the abnormal fracture in the original sequence and the continuity relationship between the pre-substitution and post-substitution segments, the abnormal identification result is output.
10. The bladder function classification and abnormality identification method based on urodynamic pattern clustering according to claim 9, characterized in that, The continuity indicators include at least one of the following: pressure change rate, urine flow rate jump amplitude, and electromyographic signal mutation degree; the abnormality identification results include at least the abnormality type, the time interval of the abnormality, and the severity level of the abnormality.