Intelligent processing method of obstetric information

Abstract

The present application relates to the technical field of obstetric data processing, in particular to an intelligent processing method of obstetric information. The method collects fetal heart rate time series data, uterine contraction pressure time series data and time stamp; cycle segmentation is carried out based on uterine contraction pressure shape characteristics, so that each analysis unit corresponds to a complete uterine contraction cycle; the uterine contraction peak time, fetal heart minimum point time and cycle length of each cycle are extracted, and a deceleration index of fusion amplitude and phase coupling relationship is calculated; the adaptive time decay weight and data quality weight are used to weight and aggregate the deceleration index of each cycle at the session level, so as to obtain a risk score and a credible factor; the number of processable sessions is dynamically determined based on the gateway buffer queue depth, and the scheduling priority is constructed in combination with the risk score and the credible factor. The present application avoids the cross-window truncation problem caused by fixed time window segmentation, and preserves the complete phase coupling information of deceleration events and uterine contraction.

Description

Technical Field

[0001] This invention relates to the field of obstetric data processing technology, and more specifically to an intelligent processing method for obstetric information. Background Technology

[0002] In obstetric clinical monitoring, fetal heart rate monitoring is a routine method for assessing the intrauterine condition of the fetus. Bedside fetal monitors continuously acquire fetal heart rate signals via an ultrasonic Doppler transducer, while simultaneously recording uterine contraction pressure signals via a pressure sensor. Both are transmitted as time-series data to the central monitoring system via the ward gateway. In clinical practice, physicians identify the risk of fetal hypoxia by interpreting the patterns of fetal heart rate changes relative to uterine contraction pressure. Late decelerations (fetal heart rate troughs lagging behind uterine contraction peaks) indicate insufficient placental perfusion reserve, while early decelerations (fetal heart rate troughs preceding uterine contraction peaks) suggest potential fetal head compression. These assessments heavily rely on the relationship between fetal heart rate deceleration events and the uterine contraction cycle.

[0003] However, existing technologies for automating the analysis of continuous monitoring data typically employ fixed time windows to segment signals at equal intervals. Because uterine contraction signals have their own physiological rhythm structure, this fixed-time-window segmentation results in the same contraction cycle being divided into two adjacent time windows. This truncates the temporal characteristics of deceleration events across windows at the window boundaries, making the feature integrity of deceleration events in the boundary region lower than that in the central region of the window. This affects the accuracy of fetal heart rate-uterine contraction coupling relationship analysis. Furthermore, during periods of sudden increases in the number of concurrent monitoring sessions, the gateway buffer queue depth increases synchronously, and the receiving end's computing resources have practical limitations. If all sessions are processed indiscriminately, the warning timeliness of high-risk sessions will generally degrade. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide an intelligent processing method for obstetric information, the specific technical solution of which is as follows:

[0005] A method for intelligent processing of obstetric information, the method comprising:

[0006] Collect fetal heart rate time-series data, uterine contraction pressure time-series data and corresponding timestamps from the ward gateway port; record the gateway buffer queue depth and the total number of active monitoring sessions;

[0007] Based on the morphological characteristics of the uterine contraction pressure time series data, the boundary of the uterine contraction cycle is obtained; based on the boundary of the uterine contraction cycle, all fetal heart rate time series data and uterine contraction pressure time series data are divided into different uterine contraction cycles.

[0008] Obtain the peak time of contraction, the lowest point of fetal heart rate, and the length of each contraction cycle; calculate the deceleration index for each contraction cycle based on the number of fetal heart rate time-series data, the peak time of contraction, the lowest point of fetal heart rate, and the length of the contraction cycle; in each session, weighted aggregate the deceleration index based on the time interval between each contraction cycle and the current time, and the proportion of effective sampling points within each contraction cycle, to obtain the risk score and credibility factor for each session.

[0009] Based on the gateway buffer queue depth and historical queue depth baselines, the number of sessions that can be processed immediately is obtained; based on the risk score and trust factor, the scheduling priority of each session is obtained; and the sessions are processed immediately according to the scheduling priority.

[0010] Furthermore, the boundaries of the uterine contraction cycle are obtained, including:

[0011] The fetal heart rate time-series data and uterine contraction pressure time-series data are respectively composed of the fetal heart rate values ​​and uterine contraction pressure values ​​corresponding to different sampling points arranged in chronological order; let the first... The uterine contraction pressure value at each sampling point The timestamp is ;

[0012] Calculate the rolling median for all sample points using a causal window; when When a contraction crosses from below its rolling median to above its rolling median, the corresponding moment is marked as the starting boundary of a cycle. The set of samples between two adjacent starting boundaries of cycles constitutes a contraction cycle.

[0013] Neighborhood length The initial value is set to a preset fixed time length, and then it is divided after each cycle is completed. Updated to the median of the set of completed contraction cycles for this session, after two consecutive updates. When the relative change is less than a preset threshold, The value converges and updates cease.

[0014] Once the system completes a detection of all cycle start boundaries in the currently accumulated uterine contraction pressure time series data, it is counted as a detection attempt. If no new cycle start boundary can be detected in two consecutive detection attempts, the session is marked as a cycle indivisible state.

[0015] Furthermore, methods for obtaining deceleration indicators include:

[0016] Let the first The fetal heart rate value at each sampling point was Validity is marked as , express efficient, express Fill in empty values;

[0017] The deceleration index is obtained according to the deceleration index calculation formula, which is shown below:

[0018]

[0019] In the formula, Indicates period The deceleration index; Indicates period All of The set that constitutes; express the median; express The minimum value; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Inside of The timestamp when the minimum value was taken; Indicates period The difference between the end timestamp and the start timestamp.

[0020] Furthermore, methods for obtaining risk scores include:

[0021] Let the total number of contraction cycles that have been segmented in the session be... The latest timestamp of the system clock is The set consisting of the lengths of all completed contraction cycles in this session is ; Set period The total number of internal sampling points is ;

[0022] The risk score is obtained based on the risk scoring calculation formula, which is shown below:

[0023]

[0024] In the formula, This indicates the risk score for each session; Indicates period The deceleration index; express the median; Indicates validity marker, express efficient, express Fill in empty values; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Total number of internal sampling points; Indicates period Distance from the current time The weights are determined by the time intervals; Indicates the proportion of valid sampling points; This represents an exponential function with the natural constant as its base.

[0025] Furthermore, methods for obtaining the credibility factor include:

[0026] The credibility factor is obtained according to the credibility factor calculation formula, which is shown below:

[0027]

[0028] In the formula, The trust factor represents each session; express the median; Indicates validity marker, express efficient, express Fill in empty values; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Total number of internal sampling points; Indicates period Distance from the current time The weights are determined by the time intervals; Indicates the proportion of valid sampling points; This represents an exponential function with the natural constant as its base.

[0029] Furthermore, methods for obtaining the number of sessions that can be processed in real time include:

[0030] Let the total number of active monitoring sessions be The gateway buffer queue depth is The set of gateway buffer queue depths recorded in historical data is as follows: The number of sessions that can be processed instantly. Calculate using the following formula:

[0031]

[0032] In the formula, express The median, i.e., the baseline value of the historical queue depth; This represents the floor function; This represents the floor function;

[0033] Set up a session Risk score Credibility factor is , conversation scheduling priority Determine by the following formula:

[0034]

[0035] Sessions are processed immediately based on scheduling priority, including: placing sessions with a periodically indivisible state of true at the head of the sorting sequence, and processing the remaining sessions according to priority. Values ​​sorted in descending order. When values ​​are equal, Sort the secondary sorting keys in descending order; then sort the first... One session is marked as being processed immediately, and the remaining sessions are marked as being processed later.

[0036] Furthermore, it also includes: maintaining the latest Calculate the first quartile of the risk score set of sessions processed within the hour. and the third and fourth quartiles ;Will conversation Mark as high risk, conversation Marked as medium risk, conversation Marked as low risk; for sessions where the periodic indivisible state is true, its risk score is set to the median of the risk scores of all processed historical sessions, and its credibility factor is set to... .

[0037] The present invention has the following beneficial effects:

[0038] This invention uses the morphological characteristics of uterine contraction pressure signals as a basis for periodic segmentation, ensuring that each analysis unit corresponds to a complete physiological uterine contraction cycle, thus avoiding the window truncation problem introduced by fixed time window segmentation. The deceleration index is coupled through the multiplication of amplitude and phase terms, simultaneously encoding the relative temporal relationship between deceleration depth and troughs / peaks. The amplitude term is normalized using the median fetal heart rate within the cycle as a ratio, eliminating the influence of individual baseline heart rate differences. The phase term smoothly maps the temporal relationship to a finite interval through sine mapping, ensuring that late deceleration requires both depth and phase conditions to be met simultaneously to obtain a higher index value, consistent with clinical interpretation logic. At the session level, the decay constant of the time decay weight is taken as the median of the set of session cycle lengths, allowing the decay rate to automatically adapt to the labor stage; that is, active phase contractions are frequent and decay rapidly, while latent phase contractions are sparse and decay slowly, avoiding mismatch of fixed decay constants at different labor stages. Data quality weights quantify signal integrity by the proportion of effective sampling within a cycle, automatically suppressing the contribution of low-quality cycles caused by probe detachment to the risk score. Simultaneously, a confidence factor explicitly indicates the data support level for the current risk estimate, providing a reliable reference for clinical decision-making. Regarding resource scheduling, the number of sessions that can be processed immediately is determined based on the deviation of the queue depth from historical benchmarks. Rounding up ensures that at least one session receives immediate processing even under extreme backlog conditions. Furthermore, a scheduling priority is jointly constructed using risk scores and confidence factors, prioritizing the timely warning of high-risk and high-uncertainty sessions with limited computing resources. Attached Figure Description

[0039] To more clearly illustrate the technical solutions and advantages in the embodiments of the present 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of an intelligent processing method for obstetric information provided in one embodiment of the present invention. Detailed Implementation

[0041] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent processing method for obstetric information proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0043] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent processing method for obstetric information provided by this invention.

[0044] Please see Figure 1 This illustrates an intelligent processing method for obstetric information provided by an embodiment of the present invention, the method comprising:

[0045] Step S1: Collect fetal heart rate time-series data, uterine contraction pressure time-series data and corresponding timestamps from the ward gateway port; record the gateway buffer queue depth and the total number of active monitoring sessions.

[0046] In this embodiment of the invention, the data source is the ward gateway port. After being sent from the gateway port, the data is continuously collected by the processing system interface. The collected data is data packets from the bedside fetal monitor. The data packets are in binary frame format, and each frame encapsulates fetal heart rate time-series data, uterine contraction pressure time-series data, and corresponding timestamps for the same bed and the same time interval. Simultaneously, the system continuously records the buffer queue depth in the sending gateway and the total number of currently active monitoring sessions.

[0047] The basic parameters included in the collected data are defined as follows. Definition of the first... The fetal heart rate values ​​at each sampling point were used. It indicates that the unit is This is obtained by parsing the fetal heart rate data field in the data frame. Based on The field parsing results show that during probe displacement, there are no valid physiological measurement values ​​at the corresponding sampling position within the device frame. The frame protocol uses a pre-defined null value filler bit to indicate this state. This null value filler can be directly observed through frame field parsing. Therefore, by... Further derive validity markers, using express, This indicates that the fetal heart rate is valid. This indicates that the value is a null filler value. Define the first... The uterine contraction pressure values ​​at each sampling point were used. This indicates that the unit is a dimensionless relative pressure unit, obtained by parsing the contraction data field in the data frame. Define the first... The timestamp of each sampling point is used It indicates that the unit is seconds.

[0048] Define the current sender gateway's buffer queue depth using This indicates the number of data packets. The gateway automatically pushes data packets to the data receiving port through a callback mechanism. The receiving end does not repeatedly query the gateway status through polling; instead, it passively responds and completes the reception of the sent data when the callback is triggered. The receiving end maintains an auto-incrementing integer counter. And the initial value is Each time the gateway pushes a batch of data, triggering a callback on the receiving end, Self-incrementing The server system clock synchronizes and records the trigger timestamp of this callback. Simultaneously, the current queue depth snapshot is read through the gateway runtime counter interface. Thus, one callback corresponds to one data triple. The ward gateway simultaneously receives data streams from monitors at multiple beds. During the interval between two push notifications, newly generated data packets continuously enter the gateway's sending buffer and queue up. The gateway executes a push notification after each data packet aggregation, but after the push is completed, there may already be new data packets piling up in the buffer. That is, the corresponding accumulation amount.

[0049] Define the total number of currently active monitoring sessions, using This indicates that the unit is the number of sessions, with one session corresponding to the monitoring data stream of one bed. When the relevant bedside device completes session registration at the receiving end and begins to continuously transmit data, the session is marked as active. When the device goes offline or the data stream is interrupted and triggers session closure, the session is removed from the active set.

[0050] Step S2: Obtain the contraction cycle boundary based on the morphological characteristics of the contraction pressure time series data; divide all fetal heart rate time series data and contraction pressure time series data into different contraction cycles based on the contraction cycle boundary.

[0051] At the physiological level, the relative temporal differences between deceleration troughs and contraction peaks reflect different pathophysiological bases: late deceleration patterns where the trough lags behind the peak suggest insufficient placental perfusion reserve, while early decelerations where the trough precedes the peak are usually associated with fetal head compression. These characteristics are only fully preserved when a single contraction cycle is used as the basic unit of analysis. If the inherent rhythmic structure of contraction signals is not considered, and continuous signals are divided into equal intervals using fixed time windows, the same contraction cycle may be divided into two adjacent time windows, causing the temporal characteristics of cross-window deceleration events to be truncated at the window boundaries. This truncation results in lower feature integrity of deceleration events at the edge boundaries compared to the central region of the window. Therefore, in this embodiment of the invention, the contraction cycle boundary is obtained based on the morphological characteristics of the contraction pressure time series data; and all fetal heart rate time series data and contraction pressure time series data are divided into different contraction cycles based on the contraction cycle boundary.

[0052] Preferably, in one embodiment of the present invention, obtaining the boundary of the uterine contraction cycle includes:

[0053] The fetal heart rate time-series data and uterine contraction pressure time-series data are respectively composed of the fetal heart rate values ​​and uterine contraction pressure values ​​corresponding to different sampling points arranged in chronological order; let the first... The uterine contraction pressure value at each sampling point The timestamp is ;

[0054] Calculate the rolling median for all sampling points. The rolling median is calculated using a causal window, i.e., for the _th ... The sampling point has a neighborhood range of the nth sampling point. The first to the second Each sampling point; when When a contraction crosses from below its rolling median to above its rolling median, the corresponding moment is marked as the starting boundary of a cycle. The set of samples between two adjacent starting boundaries of cycles constitutes a contraction cycle.

[0055] The neighborhood length The initial value is 30 seconds, and then it is divided after each cycle is completed. Updated to the median of the set of completed contraction cycles for this session; after two consecutive updates When the relative change is less than 5%, When the value converges and updates stop; If the cycle start boundary still cannot be detected after the update, Restore to the initial value and retry the detection. If the detection fails after two consecutive attempts, mark the session as a periodically indivisible state.

[0056] Step S3: Obtain the peak time of contraction, the lowest point of fetal heart rate, and the length of each contraction cycle; calculate the deceleration index for each contraction cycle based on the number of fetal heart rate time-series data, the peak time of contraction, the lowest point of fetal heart rate, and the length of the contraction cycle; in each session, weight and aggregate the deceleration index based on the time interval between each contraction cycle and the current time, and the proportion of effective sampling points in each contraction cycle, to obtain the risk score and credibility factor for each session.

[0057] After segmenting the uterine contraction cycle, each cycle contains complete fetal heart rate-uterine contraction coupling information. It is necessary to encode the depth and temporal characteristics of deceleration events into comparable values ​​using quantitative indicators. Therefore, in this embodiment of the invention, the deceleration index for each uterine contraction cycle is calculated based on the number of fetal heart rate time-series data, the peak time of contraction, the lowest point time of fetal heart rate, and the length of the uterine contraction cycle.

[0058] Preferably, in one embodiment of the present invention, the method for obtaining the deceleration index includes:

[0059] The deceleration index is obtained according to the deceleration index calculation formula, which is shown below:

[0060]

[0061] In the formula, Indicates period The deceleration index; Indicates period All of The set that constitutes; express the median; express The minimum value; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Inside of The timestamp when the minimum value was taken; Indicates period The difference between the end timestamp and the start timestamp.

[0062] In the formula, This reflects the relative magnitude of the decrease in fetal heart rate during a contraction cycle. During a contraction, increased intrauterine pressure compresses the uterine spiral arteries, causing a transient reduction in blood flow to the intervillous space on the maternal side. If placental reserve is sufficient, the fetus can maintain relatively stable blood oxygen saturation through its own oxygen reserves, and the fetal heart rate will not decrease significantly. However, when the exchange area shrinks due to placental aging or other reasons, the same contraction intensity will cause a greater decrease in fetal blood oxygen saturation, leading to a significant deceleration of the fetal heart rate via aortic chemoreceptor reflex. Therefore, The result value can reflect the remaining level of placental reserve function, and The reason for using the ratio is that there are individual differences in the baseline heart rate of different fetuses. These differences are affected by multiple factors such as gestational age, fetal behavior, and maternal medication. After the ratio is normalized, the result reflects the proportion of the deceleration amplitude relative to the individual's baseline level, ensuring comparability between different sessions.

[0063] This reflects the temporal relationship between the fetal heart rate trough and the peak of uterine contractions. There is a finite conduction delay between the interruption of maternal-fetal oxygen exchange caused by uterine contractions and the hypoxic response of the fetal cardiovascular system. When placental reserves are sufficient, fetal oxygen reserves are abundant, this conduction delay is relatively long, and the trough tends to appear after the peak of uterine contractions. When placental reserves are nearly depleted, fetal basal oxygen reserves are already low, and the same interruption of oxygen supply will trigger the chemoreceptor reflex more quickly, resulting in a relatively earlier trough. In addition, there is also early deceleration corresponding to mechanical deceleration, caused by increased intracranial pressure due to pressure on the fetal head in the birth canal. This leads to a decrease in heart rate via a direct mechanical nerve reflex through the vagus nerve. This reflex arc does not pass through chemoreceptors, therefore the trough appears before or simultaneously with the peak of uterine contractions and is not directly related to placental function. Based on these physiological differences, The positive and negative signs distinguish the pathophysiological mechanisms of the two types of deceleration: the larger the positive value, the farther the fetal heart rate trough lags behind the uterine contraction peak, indicating a more typical late deceleration characteristic and a higher likelihood of placental insufficiency, suggesting a higher clinical risk; the result value tends to be close to... The trough of the fetal heart rate wave is almost synchronous with the peak of the uterine contraction, and the distinction between the types of deceleration is in the transitional and ambiguous zone between early and late onset. The larger the absolute value of the negative value, the further the fetal heart rate wave trough leads the peak of the uterine contraction, that is, the more typical the early deceleration characteristics are. This is usually related to the benign mechanical mechanism of fetal head compression, indicating a low clinical risk.

[0064] Since a single deceleration event is insufficient to independently support the risk assessment of the entire session, the indicative value of cycles farther from the current time for the current risk will decrease, and signal loss due to factors such as probe displacement will reduce the reliability of the index estimation within the cycle. Therefore, in this embodiment of the invention, in each session, the deceleration index is weighted and aggregated based on the time interval between each contraction cycle and the current time, and the proportion of effective sampling points within each contraction cycle, to obtain the risk score and confidence factor for each session.

[0065] Preferably, in one embodiment of the present invention, the method for obtaining the risk score includes:

[0066] For a session, define the total number of contraction cycles, using express, This represents the total number of contraction cycles that have been segmented in this session and increments as processing progresses. Define a cycle index, using... Indicates and Define the period. Total number of internal samples, using express, For period The number of sampling points included within the time span.

[0067] The risk score is obtained based on the risk scoring calculation formula, which is shown below:

[0068]

[0069] In the formula, This indicates the risk score for each session; Indicates period The deceleration index; express the median; Indicates validity marker, express efficient, express Fill in empty values; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Total number of internal sampling points; Indicates period Distance from the current time The weights are determined by the time intervals; Indicates the proportion of valid sampling points; This represents an exponential function with the natural constant as its base.

[0070] In the formula, This is a time decay weight, ensuring that contraction cycles further removed from the current moment contribute less to the risk score. The decay rate of this weight is determined by... The denominator value is the median of the set of cycle durations in the session. In practice, during the active labor phase, contractions are frequent and the cycle duration is short. The fetal oxygenation status may change substantially within minutes between two contractions, and the indicative value of deceleration events in more distant cycles should decay rapidly. During the latent phase, contractions are infrequent and the cycle duration is relatively long, and the fetal status evolves slowly, so the indicative value of deceleration events in more distant cycles should be preserved to some extent. This represents the proportion of effective sampling points within the cycle, reflecting the signal integrity of the fetal heart rate channel within that cycle. When the probe shifts relative to the maternal abdominal wall, the ultrasound Doppler transducer cannot continuously capture the fetal cardiac echo, and the corresponding sampling position within the device frame is filled with a null value. This is a record of the event's occurrence. The lower the value within a period, the higher the proportion of time the probe detaches during that period. From a signal estimation perspective, insufficient sample size indicates this. The reliability of the values ​​decreases, and the tendency to deviate from the true physiological state becomes more pronounced. Risk Score It uses time decay weight and data quality weight as the joint weight. The weighted mean of the two dimensions is coupled through multiplication, and the contribution of the corresponding period is reduced when either dimension is insufficient.

[0071] Preferably, in one embodiment of the present invention, the method for obtaining the credibility factor includes:

[0072] The credibility factor is obtained according to the credibility factor calculation formula, which is shown below:

[0073]

[0074] In the formula, The trust factor represents each session; express the median; Indicates validity marker, express efficient, express Fill in empty values; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Total number of internal sampling points; Indicates period Distance from the current time The weights are determined by the time intervals; Indicates the proportion of valid sampling points; This represents an exponential function with the natural constant as its base.

[0075] In the formula, the confidence factor is the ratio of the weighted effective sampling quantity in the numerator to the weighted total number of periods in the denominator, representing how much valid, recent data supports the current calculation. The estimate is as follows: When all periods are close to full effective sampling and are relatively close to the current timestamp. Approaching ,illustrate The results are estimated to be supported by sufficient data; when the molecular fraction is low due to continuous shedding of the recent periodic probe. The value will drop significantly, indicating a lack of recent valid data, and the reliability will also decrease.

[0076] For a given session, when the period indivisibility flag is false, the session corresponding to the specified time period is determined through the aforementioned process. Value and Value. When the indivisible cycle flag is true, it indicates that no cycle segmentation points can be detected in the continuously acquired data for this session's uterine contraction pressure signals. This session lacks meaningful cycle-level feature records, and the corresponding values ​​cannot be obtained through direct calculation. Value and Value, at this time The value is directly set to the total number of sessions in the history of this processing system. The median of the values, Set directly to The median is chosen to reflect the central trend of the overall session risk level under the current operating environment, and tends to be a conservative estimate. Set as Then explicitly mark the data content of this session as unavailable.

[0077] Step S4: Based on the gateway buffer queue depth and the historical queue depth baseline, obtain the number of sessions that can be processed immediately; based on the risk score and trust factor, obtain the scheduling priority of each session; based on the scheduling priority, process the sessions immediately.

[0078] After calculating the risk score and trust factor for each session, only risk information at the single session level is obtained. There is a lack of a mechanism to convert this risk information into processing priorities, leading to a uniform degradation in the timeliness of warnings for each session when computing resources are strained. During peak concurrency periods, the number of active sessions surges, and the gateway queue depth increases synchronously, limiting the actual computing resources available to the receiving end within a single callback. Therefore, in this embodiment of the invention, the number of sessions that can be processed immediately is obtained based on the gateway buffer queue depth and historical queue depth benchmarks; the scheduling priority for each session is obtained based on the risk score and trust factor; and sessions are processed immediately according to the scheduling priority.

[0079] Preferably, in one embodiment of the present invention, the method for obtaining the number of sessions that can be processed in real time includes:

[0080] Let the total number of active monitoring sessions be The gateway buffer queue depth is The set of gateway buffer queue depths recorded in historical data is as follows: The number of sessions that can be processed instantly. Calculate using the following formula:

[0081]

[0082] In the formula, express The median, i.e., the baseline value of the historical queue depth; This represents the floor function; This represents the function for rounding up.

[0083] In the formula, The current buffer queue depth is measured based on the historical rolling median of the queue depth itself. The degree of relative offset. When When the value is lower than the baseline value reflected in historical data, The function's truncation effect makes the excess amount... The result of the exponential function is ,correspond All sessions are scheduled with the same priority. When the excess amount increases significantly, the exponential function part monotonically decreases and approaches zero. Corresponding The value decreases accordingly. By using the rounding up operator, at least one session is guaranteed to receive immediate processing, preventing the system from completely stopping responding under extreme backlog conditions.

[0084] Set up a session Risk score Credibility factor is , conversation scheduling priority Determine by the following formula:

[0085]

[0086] In the formula, The higher the value, the more clinically relevant the session needs to be. This value reflects the degree of data insufficiency; a higher value corresponds to a higher scheduling priority. Addition is chosen for coupling because either of these two factors necessitates a higher priority.

[0087] Sessions are processed in real time according to scheduling priority, including:

[0088] During the sorting phase, For each session, the session with a periodically indivisible state that is true is placed at the beginning of the sorted sequence. For the remaining sessions, they are sorted according to... Values ​​sorted in descending order. When values ​​are equal, As a secondary sorting key, the key values ​​are arranged in descending order, meaning that sessions with less data are processed first, given the same level of risk.

[0089] After sorting, through The value determines the risk level label. The processing system maintains a recent... Hours of processed sessions The set of values, and the label mapping used Hour The set of values ​​ends at the end of the previous callback, and is the current beat. After mapping and output, the data is appended to the set for later use; therefore, the risk level label for each session is determined by the distribution of historical data preceding it. The most recent... The hourly window is used because the monitoring workload in obstetric wards exhibits diurnal rhythm fluctuations. Specifically, the initiation of natural labor is related to the diurnal rhythm of maternal melatonin and oxytocin secretion. The proportion of women entering active labor during the night and early morning hours is higher than during the day, leading to differences in the number of monitoring sessions and the incidence of abnormal fetal heart rate events at night compared to the daytime. If the window length is shorter than one 24-hour cycle, then... The risk distribution captured by the set of values ​​only reflects the local characteristics of the current time period and lacks complete coverage of the load rhythm throughout the day. The first and third quartiles of this set are calculated and defined as follows: and : It is then marked as high risk. It is then marked as medium risk. It is then marked as low risk.

[0090] After sorting and determining the rank labels, the truncation phase begins: the first... The first session is marked as being in an immediate processing state, and data processing is performed immediately in the current callback; the second... To the A session is marked as deferred, and its data is retained in the processing system session buffer, waiting to be re-sorted and truncated in the next callback.

[0091] In summary, fetal heart rate time-series data, uterine contraction pressure time-series data, and corresponding timestamps are collected from the ward gateway port; the gateway buffer queue depth and the total number of active monitoring sessions are recorded; the uterine contraction cycle boundary is obtained based on the morphological characteristics of the uterine contraction pressure time-series data; all fetal heart rate time-series data and uterine contraction pressure time-series data are segmented into different uterine contraction cycles based on the uterine contraction cycle boundary; the peak time of contraction, the lowest point of fetal heart rate, and the length of each uterine contraction cycle are obtained; the deceleration index for each uterine contraction cycle is calculated based on the number of fetal heart rate time-series data, the peak time of contraction, the lowest point of fetal heart rate, and the length of the uterine contraction cycle; in each session, the deceleration index is weighted and aggregated based on the time interval between each uterine contraction cycle and the current time, and the proportion of effective sampling points within each uterine contraction cycle, to obtain the risk score and credibility factor for each session. The number of sessions that can be processed immediately is obtained based on the gateway buffer queue depth and the historical queue depth benchmark; the scheduling priority for each session is obtained based on the risk score and credibility factor; and sessions are processed immediately based on the scheduling priority.

[0092] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0093] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for intelligent processing of obstetric information, characterized in that, The method includes: Collect fetal heart rate time-series data, uterine contraction pressure time-series data and corresponding timestamps from the ward gateway port; record the gateway buffer queue depth and the total number of active monitoring sessions; Based on the morphological characteristics of the contraction pressure time series data, the contraction cycle boundary is obtained; based on the contraction cycle boundary, all the fetal heart rate time series data and the contraction pressure time series data are divided into different contraction cycles. The peak time of contraction, the lowest point of fetal heart rate, and the length of each contraction cycle are obtained. Based on the number of fetal heart rate time-series data, the peak time of contraction, the lowest point of fetal heart rate, and the length of the contraction cycle, a deceleration index for each contraction cycle is calculated. In each session, the deceleration index is weighted and aggregated based on the time interval between each contraction cycle and the current time, and the proportion of effective sampling points in each contraction cycle, to obtain the risk score and credibility factor for each session. Based on the gateway buffer queue depth and the historical queue depth baseline, the number of sessions that can be processed immediately is obtained; based on the risk score and the trust factor, the scheduling priority of each session is obtained; and the sessions are processed immediately according to the scheduling priority.

2. The intelligent processing method for obstetric information according to claim 1, characterized in that, Obtain the boundaries of the uterine contraction cycle, including: The fetal heart rate time-series data and uterine contraction pressure time-series data are respectively composed of the fetal heart rate values ​​and uterine contraction pressure values ​​corresponding to different sampling points arranged in chronological order; let the first... The uterine contraction pressure value at each sampling point The timestamp is ; Calculate the rolling median for all sample points using a causal window; when When a contraction crosses from below its rolling median to above its rolling median, the corresponding moment is marked as the starting boundary of a cycle. The set of samples between two adjacent starting boundaries of cycles constitutes a contraction cycle. Neighborhood length The initial value is set to a preset fixed time length, and then it is divided after each cycle is completed. Updated to the median of the set of completed contraction cycles for this session, after two consecutive updates. When the relative change is less than a preset threshold, The value converges and updates cease. Once the system completes a detection of all cycle start boundaries in the currently accumulated uterine contraction pressure time series data, it is counted as a detection attempt. If no new cycle start boundary can be detected in two consecutive detection attempts, the session is marked as a cycle indivisible state.

3. The intelligent processing method for obstetric information according to claim 2, characterized in that, The method for obtaining the deceleration index includes: Let the first The fetal heart rate value at each sampling point was Validity is marked as , express efficient, express Fill in empty values; The deceleration index is obtained according to the deceleration index calculation formula, which is as follows: In the formula, Indicates period The deceleration index; Indicates period All of The set that constitutes; express the median; express The minimum value; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Inside of The timestamp when the minimum value was taken; Indicates period The difference between the end timestamp and the start timestamp.

4. The intelligent processing method for obstetric information according to claim 3, characterized in that, The methods for obtaining the risk score include: Let the total number of contraction cycles that have been segmented in the session be... The latest timestamp of the system clock is The set consisting of the lengths of all completed contraction cycles in this session is ; Set period The total number of internal sampling points is ; The risk score is obtained according to the risk scoring calculation formula, which is shown below: In the formula, This indicates the risk score for each session; Indicates period The deceleration index; express the median; Indicates validity marker, express efficient, express Fill in empty values; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Total number of internal sampling points; Indicates period Distance from the current time The weights are determined by the time intervals; This indicates the proportion of effective sampling points; This represents an exponential function with the natural constant as its base.

5. The intelligent processing method for obstetric information according to claim 4, characterized in that, The method for obtaining the credibility factor includes: The credibility factor is obtained according to the credibility factor calculation formula, which is as follows: In the formula, The trust factor represents each session; express the median; Indicates validity marker, express efficient, express Fill in empty values; Indicates period Inside The timestamp when the maximum value is retrieved; Indicates period Total number of internal sampling points; Indicates period Distance from the current time The weights are determined by the time intervals; This indicates the proportion of effective sampling points; This represents an exponential function with the natural constant as its base.

6. The intelligent processing method for obstetric information according to claim 1, characterized in that, The method for obtaining the number of sessions that can be processed in real time includes: Let the total number of active monitoring sessions be The gateway buffer queue depth is The set of gateway buffer queue depths recorded in historical data is as follows: The number of sessions that can be processed in real time; Calculate using the following formula: In the formula, express The median, i.e., the baseline value of the historical queue depth; This represents the floor function; Set up a session The risk score is The credibility factor is , conversation The scheduling priority Determine by the following formula: Sessions are processed immediately based on scheduling priority, including: placing sessions with a periodically indivisible state of true at the head of the sorting sequence, and processing the remaining sessions according to priority. Values ​​sorted in descending order. When values ​​are equal, Sort the secondary sorting keys in descending order; then sort the first... One session is marked as being processed immediately, and the remaining sessions are marked as being processed later.

7. The intelligent processing method for obstetric information according to claim 6, characterized in that, Also includes: maintaining the latest Calculate the first quartile of the risk score set of sessions processed within the hour. and the third and fourth quartiles ;Will conversation Mark as high risk, conversation Marked as medium risk, conversation Marked as low risk; for sessions where the periodic indivisible state is true, its risk score is set to the median of the risk scores of all processed historical sessions, and its credibility factor is set to... .

Images