Intelligent monitoring method and system for perfusion curative effect of thin endometrial rich microplate

Abstract

The invention relates to the technical field of medical artificial intelligence monitoring, and discloses an intelligent monitoring method and system for a perfusion curative effect of a thin endometrial rich microplate. The method comprises the following steps: standardizing and fusing multi-source monitoring data, constructing a derivative feature set reflecting a treatment coordination rule through cross-cycle feature derivation, and establishing a curative effect behavior coding library according to the derivative feature set. After the real-time data is subjected to standardization processing, frame-by-frame mode matching and similarity calculation are carried out on the real-time data and a coding library, a corresponding analysis path is activated according to a result, and a curative effect behavior fragment with a timestamp is generated. The segments are associated and aggregated to form a complete curative effect event chain, key turning points are marked, and finally a structured monitoring report and a decision signal are generated through multi-dimensional influence evaluation. According to the invention, the conversion from static parameter judgment to dynamic curative effect behavior intelligent analysis is realized, and the continuity, accuracy and early warning capability of treatment process monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of medical artificial intelligence monitoring technology, specifically to a method and system for intelligent monitoring of the therapeutic effects of platelet-rich perfusion in thin endometrium. Background Technology

[0002] Current monitoring of the efficacy of platelet-rich infusion therapy for thin endometrium primarily relies on the assessment of independent parameters within a single session or cycle. Conventional techniques include measuring morphological indicators such as endometrial thickness and type at specific time points, or recording isolated parameters such as pressure and flow rate during infusion, with alarms triggered by set thresholds for abnormalities. This monitoring data is fragmented, and analysis typically lags behind the treatment procedure, resulting in static and piecemeal assessment reports. Physicians must rely on personal experience to subjectively correlate results from different time points to infer the overall trend of treatment response.

[0003] The fundamental flaw in existing technologies lies in their inability to capture the dynamic and synergistic changes between physiological parameters and treatment progress from continuous, multi-cycle treatment data. Important information such as the temporal correlation between morphological improvement and perfusion dynamics parameters, and the evolutionary trends of patient response patterns across different treatment cycles, are overlooked. This often leads to delayed clinical intervention, hindering early warning and pathway analysis at the onset of therapeutic shifts or abnormal behaviors, thus affecting the precision and timeliness of treatment.

[0004] The problem this invention aims to solve is to realize the transformation from static parameter monitoring to dynamic efficacy behavior analysis. By integrating multi-source data and analyzing their cross-cycle synergistic patterns, an intelligent framework that can match and analyze different response modes is constructed, thereby providing continuous, structured, and interpretable intelligent monitoring and decision support for the entire process of thin endometrial perfusion therapy. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent monitoring method and system for the therapeutic effect of platelet-rich perfusion in thin endometrium, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a method for intelligent monitoring of the therapeutic effect of platelet-rich perfusion in thin endometrium, the method comprising: Patient data from multiple independent monitoring sources were standardized to form a fused data stream that includes endometrial morphological parameters, perfusion hydrodynamic parameters, and treatment stage labels; Based on the fused data stream, cross-cycle feature derivation is performed to generate a derived feature set that reflects the synergistic change pattern between the treatment process and physiological parameters; Using the derived feature set, a therapeutic behavior coding library containing normal response patterns and multiple predefined abnormal response patterns is constructed; After converting the real-time collected medical monitoring data into a standard format, it is matched frame by frame with the patterns in the therapeutic behavior coding library and similarity is calculated. The matching results and confidence scores are then output. Based on the matching results and confidence scores, the corresponding pattern analysis path is activated to perform deep structural analysis on real-time medical monitoring data and generate a sequence of therapeutic behavior fragments with timestamps. The therapeutic behavior segments are context-linked and aggregated to form a complete therapeutic event chain, and key turning points are marked. Based on the aforementioned efficacy event chain, a multi-dimensional impact assessment is conducted to quantify the potential impact of specific efficacy behaviors on the overall treatment process, generating structured efficacy monitoring reports and decision support signals.

[0007] Preferably, the standardization of patient data from multiple independent monitoring sources to form a fused data stream including endometrial morphological parameters, perfusion hydrodynamic parameters, and treatment stage labels specifically includes: Receives raw, heterogeneous data from ultrasound imaging equipment, perfusion recording systems, and clinical information systems; The original heterogeneous data is time-stamp aligned and units are standardized to resolve the differences in time series and units between different monitoring sources; Extract the core metrics from the aligned data and attach treatment stage labels containing treatment cycles and interventions; The core indicators labeled with the treatment stage are encapsulated and concatenated in chronological order to generate the fused data stream.

[0008] Preferably, based on the fused data stream, cross-cycle feature derivation is performed to generate a derived feature set reflecting the synergistic changes in treatment progress and physiological parameters, specifically including: From the fused data stream, a data window containing several consecutive treatment cycles is slidably extracted; Within each data window, the delayed response, rate of change, and trend consistency index of a specific physiological parameter relative to the treatment phase label are calculated. The calculation results of all data windows are aggregated to form the derived feature set that characterizes the dynamic evolution of physiological parameters driven by the treatment process.

[0009] Preferably, the step of constructing a therapeutic behavior coding library containing normal response patterns and multiple predefined abnormal response patterns using the derived feature set specifically includes: Based on successful treatment cases in historical data, common patterns are summarized from the derived feature set and encoded as the normal response pattern; Based on clinical medical knowledge, typical adverse reaction or inadequate efficacy scenarios are defined and mapped to the space of the derived feature set to construct a variety of predefined abnormal response patterns. The normal response pattern and various predefined abnormal response patterns are structured, stored, and indexed to form the therapeutic behavior coding library.

[0010] Preferably, the step of converting the real-time collected medical monitoring data into a standard format and then performing frame-by-frame matching and similarity calculation with the patterns in the therapeutic behavior coding library, and outputting the matching results and confidence scores, specifically includes: Perform the same standardization process on the real-time collected medical monitoring data to generate real-time standard data fragments; Calculate the distance or similarity measure between the real-time standard data fragment and each pattern in the therapeutic behavior coding library in the feature space; The pattern with the highest similarity is selected as the matching result, and the corresponding similarity metric is normalized and used as the confidence score.

[0011] Preferably, the step of activating the corresponding pattern analysis path based on the matching results and confidence scores to perform deep structural analysis on real-time medical monitoring data and generate a sequence of therapeutic behavior fragments with timestamps specifically includes: Based on the pattern type of the matching result, invoke the dedicated parsing rule and feature extractor bound to that pattern type; Using the dedicated parsing rules and feature extractor, the real-time medical monitoring data that produced the matching result is decomposed into fine-grained parts to identify sub-behaviors and state transitions. Each identified sub-behavior is labeled with its precise start and end timestamps, and arranged in chronological order to form the sequence of therapeutic behavior segments; The process utilizes specialized parsing rules and feature extractors to perform fine-grained decomposition of the real-time medical monitoring data that generated the matching result, identifying sub-behaviors and state transitions, specifically including: Based on the pattern type of the matching result, load the parsing template that is pre-associated with the pattern type; the parsing template defines the sub-behavior types and their logical order that are expected to be included for this type of pattern data; A dedicated feature extractor bound to the parsing template is used to calculate a set of fine-grained indicators for the real-time medical monitoring data, including the slope, numerical level, and fluctuation frequency of key physiological parameters within a specific time window. Based on the logical order and threshold rules defined in the parsing template, the calculated fine-grained index sequence is divided into continuous intervals, each interval representing a specific sub-behavior; Identify the moments when significant changes occur between adjacent sub-behavior intervals and mark these moments as state transition points; All identified sub-behaviors, their corresponding start and end time intervals, and the timestamps of state transition points are combined and output in a structured manner as the identification results of the sub-behaviors and state transitions.

[0012] Preferably, the step of contextually associating and aggregating the sequence of therapeutic behavior fragments to assemble them into a complete therapeutic event chain, and marking key turning points, specifically includes: Analyze the causal logic and temporal proximity between adjacent segments in the therapeutic behavior segment sequence; Based on the causal logic and temporal proximity, multiple closely related therapeutic behavior segments are aggregated into a larger therapeutic event unit. All therapeutic event units are linked in chronological order to form the complete therapeutic event chain, and the key turning points in the chain where the treatment effect changes significantly are identified.

[0013] Preferably, the multi-dimensional impact assessment based on the therapeutic event chain, quantifying the potential impact of specific therapeutic behaviors on the overall treatment process, specifically includes: Define multiple dimensions for impact assessment, including time dimension, resource dimension, and clinical endpoint achievement dimension; Within the chain of therapeutic events, the specific therapeutic behavior to be evaluated is located, and the chain changes in its subsequent events are analyzed across multiple dimensions. Based on the magnitude and range of the aforementioned chain changes, the potential impact of the specific therapeutic behavior is calculated using a pre-defined quantitative model.

[0014] Preferably, the step of comprehensively considering the magnitude and range of the chain changes and calculating the potential impact of the specific therapeutic behavior using a preset quantitative model specifically includes: Quantifiable impact indicators are defined for the time dimension, resource dimension, and clinical endpoint achievement dimension, respectively; the time dimension impact indicator is the delay time in subsequent treatment stages, the resource dimension impact indicator is the additional medical resource cost, and the clinical endpoint achievement dimension impact indicator is the attenuation value of the expected improvement in key physiological parameters. Extract the actual measured values ​​of the influencing indicators of each dimension within a preset observation window after the occurrence of the specific therapeutic behavior from the therapeutic event chain; The actual measured values ​​of the influencing indicators of each dimension are input into the pre-trained quantitative model. The quantitative model is trained with historical efficacy event chain data and integrates multi-dimensional indicators through weighted summation. The weight coefficients reflect the importance of each dimension to the overall treatment process. The quantitative model outputs a comprehensive score as a quantitative value of the potential impact of the specific therapeutic behavior.

[0015] Preferably, when the processor executes the computer program, it implements the steps of the intelligent monitoring method for the efficacy of platelet-rich perfusion in thin endometrium as described in any of the above-mentioned methods.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By deriving features across treatment cycles, time-series modeling is performed on multi-source data streams integrating endometrial morphology, perfusion dynamics, and treatment stages. Derived features reflecting the synergistic changes in treatment progress and physiological parameters are extracted. This allows for the characterization of the dynamic interaction and evolution patterns of multi-dimensional indicators such as endometrial growth and blood perfusion across multiple treatment cycles. This identifies potential patterns and gradual trend shifts that cannot be detected in a single examination, providing quantitative evidence based on data evolution patterns for dynamic efficacy assessment and early risk warning.

[0017] By utilizing a pre-built therapeutic behavior coding library, standardized monitoring data collected in real time is matched frame-by-frame with various patterns in the library, and similarity calculations are performed. Based on the matching results, corresponding deep structural parsing paths are activated. This achieves automated semantic segmentation and pattern recognition of continuous data streams, enabling real-time parsing of specific therapeutic behavior segments such as delayed response and fluctuating response. The original monitoring waveforms and images are transformed into structured event sequences with clear clinical significance and time stamps.

[0018] By contextualizing and aggregating sequences of therapeutic behavior fragments, a complete chain of therapeutic events is assembled and key turning points are marked, enabling multi-dimensional quantitative assessment of their impact. This process integrates discrete behavioral identifications into a logically coherent narrative of the treatment process, elucidating the relationships between different events and their potential weighting in relation to the final treatment goal. The resulting structured report provides clear and interpretable decision support signals for the timing of clinical interventions and the adjustment of strategies. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the working principle of the intelligent monitoring method for the efficacy of thin-film platelet-rich endometrial perfusion as described in this invention. Figure 2 A flowchart for data standardization and fusion data flow generation; Figure 3 Flowchart for building a coding library for therapeutic behaviors; Figure 4 Dynamic monitoring graphs of endometrial thickness and perfusion efficiency in different cycles of platelet-rich perfusion therapy for thin endometrium; Figure 5 Dynamic trend graph of behavioral fragment sequence and key physiological parameters for the therapeutic effect of platelet-rich perfusion in thin endometrium. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 This invention provides an intelligent monitoring method for the efficacy of platelet-rich endometrial perfusion therapy. The method includes: standardizing patient data from multiple independent monitoring sources to form a fused data stream containing endometrial morphological indicators, perfusion hydrodynamic parameters, and treatment stage labels. Based on the fused data stream, cross-cycle feature derivation is performed to generate a derived feature set reflecting the synergistic changes in treatment progress and physiological parameters. Using the derived feature set, an efficacy behavior coding library containing normal response patterns and multiple predefined abnormal response patterns is constructed. Real-time acquired medical monitoring data is converted into a standard format and then matched frame-by-frame with patterns in the efficacy behavior coding library, with similarity inclusion calculations performed to output matching results and confidence scores. Based on the matching results and confidence scores, corresponding pattern analysis paths are activated to perform deep structural analysis of the real-time medical monitoring data, generating a time-stamped sequence of efficacy behavior segments. Contextual association and aggregation are performed on the efficacy behavior segment sequences to assemble them into a complete efficacy event chain, and key turning points are marked. Based on the efficacy event chain, a multi-dimensional impact assessment is performed to quantify the potential impact of specific efficacy behaviors on the overall treatment process, generating a structured efficacy monitoring report and decision support signals.

[0022] Example 1: See Figure 2 The system receives raw, heterogeneous data from ultrasound imaging equipment, perfusion recording systems, and clinical information systems. It performs timestamp alignment and unit standardization on the raw, heterogeneous data to resolve time-series and dimensional differences between different monitoring sources. Core indicators are extracted from the aligned data and labeled with treatment stage tags based on treatment cycles and interventions. These core indicators with treatment stage tags are then encapsulated and concatenated chronologically to generate a fused data stream. From the fused data stream, data windows containing several consecutive treatment cycles are slidably extracted. Within each data window, the delayed response, rate of change, and trend consistency index of specific physiological parameters relative to the treatment stage tag are calculated. The calculation results from all data windows are aggregated to form a derived feature set characterizing the dynamic evolution of physiological parameters driven by the treatment process.

[0023] In practice, the system receives raw heterogeneous data from ultrasound imaging equipment, including time-series measurements of endometrial thickness, echo type, and perfusion index. It also receives raw heterogeneous data from the perfusion recording system, including instantaneous records of perfusion fluid flow rate, pressure, and total perfusion volume. Furthermore, it receives raw heterogeneous data from the clinical information system, including treatment cycle dates, platelet-rich plasma concentrations used, and records of adjuvant medications. The system aligns these raw heterogeneous data with timestamps and standardizes units. The ultrasound imaging equipment uses the examination completion time as the timestamp, the perfusion recording system uses the milliseconds of the operation log as the timestamp, and the clinical information system uses the time of prescription as the timestamp. All these timestamps are then adjusted to a relative timeline based on the actual start time of the treatment procedure. The units for endometrial thickness are standardized to millimeters, perfusion index to dimensionless ratios, flow rate to milliliters per minute, and pressure to kilopascals. Core indicators were extracted from the aligned and unified data, including maximum endometrial thickness, mean perfusion index, peak perfusion velocity, and mean perfusion pressure. Each core indicator record was labeled with a treatment phase tag, including the treatment cycle number and specific intervention, for example, "Cycle 2 - Monitoring on day 5 after platelet-rich plasma perfusion". The core indicators with treatment phase tags were then encapsulated and concatenated strictly according to their corresponding timestamps to generate a fused data stream containing multi-dimensional parameters and phase semantics.

[0024] In some embodiments, timestamp alignment addresses temporal differences between different monitoring sources. For example, an ultrasound examination may be performed two hours after a perfusion procedure, while a medication order may have been recorded by the clinical information system the day before the perfusion procedure. The alignment process associates the ultrasound and perfusion data with the same treatment event sequence based on absolute time and corrects the medication order timestamp to match the perfusion procedure start timestamp. Unit normalization addresses dimensional differences between different monitoring sources. For instance, one ultrasound device outputs a blood flow signal intensity in grayscale values ​​of 0-255, while another device outputs a normalized value of 0-1. The normalization process divides all grayscale values ​​by 255, converting them into a uniform 0-1 range dimensionless ratio.

[0025] Optionally, the addition of treatment phase labels depends on the structured chemotherapy regimen plan parsed from the clinical information system. The regimen plan defines consecutive cycle numbers and key intervention points within each cycle. The system automatically matches and labels the corresponding cycle number and intervention phase description based on the absolute time interval of the data timestamp.

[0026] In practice, from the generated fused data stream, a data window containing three consecutive treatment cycles is slidably extracted, with a complete treatment cycle as the step size. Within each extracted data window, the delayed response, rate of change, and trend consistency index of specific physiological parameters relative to the treatment phase label are calculated. The difference between the endometrial thickness measured on the fifth day after each "platelet-rich plasma perfusion" phase label and the baseline thickness value on the day before that perfusion is calculated as the thickness delayed response index. The rate of change of the mean perfusion index sequence within the data window between the times corresponding to two adjacent "monitoring" phase labels is calculated. The covariance of the peak perfusion velocity sequence and the mean perfusion pressure sequence within the data window at the same time is calculated, and then divided by the product of the standard deviations of the two sequences to obtain the trend consistency index characterizing the coordinated change of perfusion hydrodynamic parameters. The calculation results of all sliding data windows are aggregated to form a derived feature set that characterizes the dynamic evolution of physiological parameters driven by the treatment process. The derived feature set includes a series of cross-cycle delayed response value sequences, rate of change sequences, and trend consistency index sequences.

[0027] It is understandable that the calculation of cross-cycle features relies on the continuous truncation of a sliding window. After the calculation of a data window containing cycle 1, cycle 2, and cycle 3 is completed, the window slides forward one cycle, and the next window will contain data from cycle 2, cycle 3, and cycle 4 and perform the same feature calculation, thereby generating derived features for each treatment cycle in multiple overlapping context windows. The formula for calculating the trend consistency index is shown below:

[0028] in: Indicators representing trend consistency Represents peak perfusion velocity sequence With mean perfusion pressure sequence covariance, Represents peak perfusion velocity sequence standard deviation Mean perfusion pressure sequence The standard deviation.

[0029] Example 2: See Figure 3Based on successful treatment cases from historical data, common patterns are summarized from the derived feature set and encoded as normal response patterns. Typical adverse reactions or inadequate efficacy scenarios are defined according to clinical medical knowledge and mapped to the space of the derived feature set, constructing multiple predefined abnormal response patterns. The normal response patterns and multiple predefined abnormal response patterns are structured, stored, and indexed to form an efficacy behavior coding library. Real-time collected medical monitoring data undergoes the same operations as standardization to generate real-time standard data fragments. The distance or similarity metric between the real-time standard data fragments and each pattern in the efficacy behavior coding library in the feature space is calculated. The pattern with the highest similarity is selected as the matching result, and the corresponding similarity metric value is normalized and used as a confidence score.

[0030] In practice, based on successful treatment cases from historical data, common patterns are summarized from the derived feature set and encoded as normal response patterns. Successful treatment cases in the historical data refer to records of cases where endometrial thickness reached clinical standards and pregnancy outcomes were good after completing the preset treatment course. Patterns are extracted from the derived feature set corresponding to these cases. For example, the delayed response index of endometrial thickness shows a monotonically increasing trend over three consecutive treatment cycles, with the increase falling within a specific range; the rate of change of the mean blood perfusion index shows a trend of rapid increase followed by stabilization within the cycle; and the trend consistency index between peak perfusion velocity and mean perfusion pressure remains positively high. These patterns are parameterized as feature thresholds and change trajectory templates and encoded into a structured normal response pattern record stored in the database.

[0031] In some embodiments, typical adverse reactions or inadequate treatment scenarios are defined based on clinical medical knowledge. For example, the "delayed response" scenario refers to no significant increase in endometrial thickness on the fifth day after perfusion; the "overreaction" scenario refers to a rapid increase in the mean perfusion index accompanied by severe fluctuations during the monitoring period; and the "poor coordination" scenario refers to a negative or low value in the trend consistency index between peak perfusion velocity and mean perfusion pressure. These scenario descriptions are mapped to the space of a derived feature set, transforming them into specific feature constraints. The "delayed response" scenario is mapped to an endometrial thickness delay response index below a threshold for two consecutive cycles; the "overreaction" scenario is mapped to an absolute value of the rate of change of the mean perfusion index exceeding a threshold and a sequence fluctuation variance exceeding a threshold; and the "poor coordination" scenario is mapped to a trend consistency index below a threshold.

[0032] Optionally, normal response patterns and multiple predefined abnormal response patterns can be structured, stored, and indexed to form a therapeutic behavior coding library. The structured storage is a database table, with each pattern record containing a pattern number, pattern type, pattern description, and a series of feature constraint fields. Indexing is based on key fields of the feature constraints; for example, inverted indexes are created for endometrial thickness delay response indicators, mean blood perfusion index change rate, and trend consistency indicators to support subsequent efficient retrieval and matching operations.

[0033] In practice, the real-time medical monitoring data undergoes the same standardization process as the data acquisition process to generate a real-time standard data segment. The real-time medical monitoring data comes from a patient undergoing treatment. The system receives the patient's latest ultrasound measurement data, perfusion instrument output data, and the operation record of the current treatment. This data is then time-stamped and standardized in units. The maximum endometrial thickness, mean perfusion index, peak perfusion velocity, and mean perfusion pressure are extracted as core indicators, and a label indicating the current treatment stage is added to form a uniformly formatted real-time standard data segment.

[0034] In some embodiments, the distance or similarity metric between the real-time standard data fragment and each pattern in the therapeutic behavior coding library is calculated in the feature space. For each pattern record in the therapeutic behavior coding library, the feature dimension corresponding to its defined feature constraints is extracted, and feature values ​​of the same dimension are calculated or extracted from the real-time standard data fragment. A distance metric is used to calculate the Euclidean distance between the real-time feature vector and the pattern feature constraint center vector. Alternatively, a similarity metric is used to calculate the cosine similarity between the real-time feature vector and the pattern feature vector. A smaller distance value or a larger cosine similarity value indicates a higher degree of matching.

[0035] It is understandable that distance calculation in the feature space requires normalization of features with different dimensions. For example, all feature values ​​such as the delayed response index of endometrial thickness and the rate of change of the mean blood perfusion index are standardized using the mean and standard deviation of historical data, converting them into standard scores with a mean of 0 and a variance of 1, to ensure consistent weighting of each dimension in the distance calculation. Optionally, the pattern with the highest similarity is selected as the matching result, and the corresponding similarity metric is normalized and used as the confidence score. All patterns in the therapeutic behavior coding library are traversed to obtain a set of similarity values, from which the pattern record corresponding to the maximum value is found. This maximum value is mapped to the interval of 0 to 100 using a preset normalization function, such as linear scaling. The maximum and minimum values ​​of historical similarity values ​​are used as the range boundaries, and the current maximum value is mapped to a specific score within this interval. This score is the confidence score output to the user. The matching result and the confidence score together constitute the system's pattern recognition conclusion for the current real-time treatment status.

[0036] In practice, the similarity metric uses the cosine similarity method based on normalized feature vectors, and its calculation formula is as follows:

[0037] in: This represents the calculated cosine similarity. This represents the total number of feature dimensions involved in the matching. Indicates the real-time standard data segment in the first Normalized feature values ​​on each feature dimension This indicates that a predefined pattern in the therapeutic behavior coding library is in the first... Normalized center values ​​or expected values ​​on each feature dimension.

[0038] Example 3: Based on the pattern type of the matching result, the dedicated parsing rules and feature extractor bound to that pattern type are invoked. Using the invoked dedicated parsing rules and feature extractor, the real-time medical monitoring data that generated the matching result is decomposed into fine-grained segments, identifying sub-behaviors and state transitions. Each identified sub-behavior is labeled with its precise start and end timestamps, arranged chronologically to form a sequence of therapeutic behavior segments. Based on the pattern type of the matching result, a parsing template pre-associated with that pattern type is loaded. The parsing template defines the expected sub-behavior types and their logical order for this type of pattern data. The dedicated feature extractor bound to the parsing template is applied to calculate a set of fine-grained indicators for the real-time medical monitoring data, including the slope, numerical level, and fluctuation frequency of key physiological parameters within a specific time window. According to the logical order and threshold rules defined in the parsing template, the calculated fine-grained indicator sequence is divided into continuous intervals, each interval representing a sub-behavior. The moment when a significant change occurs between adjacent sub-behavior intervals is identified and marked as a state transition point. All identified sub-behaviors, their corresponding start and end time intervals, and the timestamps of state transition points are combined and output in a structured manner as the identification results of sub-behaviors and state transitions.

[0039] In practice, based on the matching results and confidence scores, the corresponding pattern analysis path is activated. The pattern type of the matching results is a predefined abnormal response pattern called "overreaction." The system calls a dedicated parsing rule and feature extractor bound to the "overreaction" pattern type. Using the called dedicated parsing rule and feature extractor, the real-time medical monitoring data that generated the matching results is decomposed into fine-grained segments to identify sub-behaviors and state transitions. Each identified sub-behavior is labeled with its precise start and end timestamps and arranged in chronological order to form a sequence of therapeutic behavior fragments.

[0040] In some embodiments, based on the pattern type of the matching result, a parsing template pre-associated with that pattern type is loaded. The parsing template defines the expected sub-behavioral types and their logical order for the "overreaction" pattern data. For example, the three sub-behavioral types, "rapid rise in blood flow," "high-value fluctuation," and "decline adjustment," are expected to occur in this order. A dedicated feature extractor bound to the parsing template is applied to calculate a set of fine-grained indicators for the real-time medical monitoring data. The real-time medical monitoring data is a sequence of average blood perfusion index samples taken at high frequencies over the past 72 hours. The dedicated feature extractor uses a ten-minute time window to calculate the slope, numerical level, and fluctuation frequency of the average blood perfusion index within each window. The slope is obtained through linear fitting, the numerical level is the mean of the measured values ​​within the window, and the fluctuation frequency is statistically analyzed by detecting the number of times the measured value crosses a preset median line.

[0041] In practice, based on the logical order and threshold rules defined in the parsing template, the calculated fine-grained index sequence is divided into continuous intervals. The definition rule for the "rapid rise in blood flow" in the parsing template is that the slope of change is continuously greater than the positive threshold. And the numerical level is below the threshold. The definition rule for a "high-value fluctuation period" is that the numerical level is consistently higher than a threshold. And the fluctuation frequency is higher than the threshold. The definition rule for a "pullback and adjustment period" is that the slope of change is less than or equal to a negative threshold. And the numerical level is below the threshold. The system compares the real-time calculated change slope sequence, numerical level sequence, and fluctuation frequency sequence with these rules. The continuous time points that simultaneously satisfy all the rules of the "rapid rise in blood flow" are divided into the first interval, the continuous time points that subsequently satisfy all the rules of the "high value fluctuation period" are divided into the second interval, and the continuous time points that subsequently satisfy all the rules of the "fallback and adjustment period" are divided into the third interval. Each interval represents a sub-behavior.

[0042] Optionally, the system identifies moments of significant change between adjacent sub-behavior intervals and marks these moments as state transition points. For example, the state transition point from "rapid blood flow rise phase" to "high value fluctuation phase" is the moment after the first time point that satisfies the "high value fluctuation phase" rule and the moment after the last time point that satisfies the "rapid blood flow rise phase" rule. The system records the absolute timestamp of this state transition point. All identified sub-behaviors and their corresponding start and end time intervals, along with the timestamps of the state transition points, are combined and structured into a structured output as the identification results of sub-behaviors and state transitions. The output format is a list containing the sub-behavior type, start time, end time, and transition point time.

[0043] Optionally, each identified sub-behavior is labeled with its precise start and end timestamps. The start timestamp of the sub-behavior "rapid blood flow rise phase" is marked with the acquisition time of the first data point within its time interval, and the end timestamp is marked with the acquisition time of the last data point within the interval. All sub-behavior descriptions with labeled timestamps are arranged in chronological order to form a therapeutic behavior segment sequence. The therapeutic behavior segment sequence is a list sorted by time, and each element in the list records the sub-behavior type and its start and end times.

[0044] In some embodiments, the formula for calculating the slope of change is as follows:

[0045] in: Indicates the first The slope of change calculated within a sliding time window This indicates the total number of data points contained within the time window. Indicates the sequential index of the data points within the window. This represents the mean of the sequential indices. Indicates the first Within the first time window Average perfusion index measurement at each index location Indicates the first The average of all mean blood perfusion index measurements within a sliding time window. This formula is used to quantify the trend strength and direction of physiological parameter changes within each sliding time window.

[0046] Example 4: Define multiple dimensions for impact assessment, including time, resource, and clinical endpoint achievement dimensions. In the efficacy event chain, locate the specific efficacy behavior to be assessed and analyze the chain reactions of its subsequent events across multiple dimensions. Based on the magnitude and scope of these chain reactions, calculate the potential impact of the specific efficacy behavior using a pre-defined quantitative model. Define quantifiable impact indicators for the time, resource, and clinical endpoint achievement dimensions. The time dimension impact indicator is the delay time in subsequent treatment stages; the resource dimension impact indicator is the additional medical resource cost; and the clinical endpoint achievement dimension impact indicator is the attenuation value of the expected improvement in key physiological parameters. Extract the actual measured values ​​of the impact indicators for each dimension within a pre-defined observation window after the specific efficacy behavior occurs from the efficacy event chain. Input the actual measured values ​​of the impact indicators for each dimension into a pre-trained quantitative model. The quantitative model is trained using historical efficacy event chain data and internally integrates multi-dimensional indicators through a weighted summation method. The weight coefficients reflect the importance of each dimension to the overall treatment process. The quantitative model outputs a comprehensive score as a quantified value of the potential impact of the specific efficacy behavior.

[0047] In practice, multiple dimensions of impact assessment were defined, including time, resources, and clinical endpoint achievement. A specific therapeutic behavior to be assessed was located within the complete therapeutic event chain. One therapeutic behavior unit, identified as an "overreaction," occurred during the monitoring phase of treatment cycle 3. The chain of events following the "overreaction" therapeutic behavior was analyzed across multiple dimensions. In the time dimension, the subsequent "endometrial preparation" event was delayed by five days compared to the original plan. In the resource dimension, two additional ultrasound monitoring sessions and one consultation were added. In the clinical endpoint achievement dimension, the endometrial thickness increase in the next cycle's "endometrial thickness increase" event was less than 30% of the increase in the previous cycle. Based on the magnitude and scope of the chain of changes, a pre-defined quantitative model was used to calculate the potential impact of the specific therapeutic behavior, the "overreaction." The specific implementation includes defining quantifiable impact indicators for the time dimension, resource dimension, and clinical endpoint achievement dimension. The time dimension impact indicator is the delay time occurring in subsequent treatment stages; the resource dimension impact indicator is the additional medical resource cost consumed; and the clinical endpoint achievement dimension impact indicator is the attenuation value of the expected improvement in key physiological parameters. The actual measured values ​​of each dimension impact indicator are extracted from the efficacy event chain after the occurrence of a specific efficacy behavior of "overreaction" within a preset observation window. The actual measured values ​​of each dimension impact indicator are normalized to eliminate dimensional differences. The normalized indicator values ​​are input into a pre-trained quantitative model, which is trained on historical efficacy event chain data. Internally, it integrates multi-dimensional indicators through a weighted summation method. The weight coefficients are determined by analyzing the correlation strength between changes in different dimension indicators and the final treatment outcome in historical data, reflecting the importance of each dimension to the overall treatment process. The quantitative model outputs a comprehensive score as a quantitative value of the potential impact of a specific efficacy behavior.

[0048] In some embodiments, quantifiable impact indicators are defined for the time dimension, resource dimension, and clinical endpoint achievement dimension. The time dimension impact indicator is the delay time occurring in subsequent treatment stages; the resource dimension impact indicator is the additional medical resource cost incurred; and the clinical endpoint achievement dimension impact indicator is the attenuation value of the expected improvement in key physiological parameters. The delay time is calculated in days, the additional medical resource cost is calculated in monetary units, and the attenuation value of the expected improvement in key physiological parameters is expressed as a percentage. From the complete therapeutic event chain, the actual measured values ​​of each dimension's impact indicators are extracted within a preset observation window after the occurrence of the specific therapeutic behavior of "overreaction." The preset observation window covers the time period from the start of the "overreaction" to the end of the next complete treatment cycle. In specific implementations, refer to Table 1 to extract the actual measured values ​​of each dimension's impact indicators.

[0049] Table 1: Measurement values ​​of multidimensional influencing indicators of specific therapeutic behaviors

[0050] Optionally, the actual measured values ​​of the influencing indicators of each dimension are input into a pre-trained quantitative model. This model is trained using historical efficacy event chain data and internally integrates multi-dimensional indicators through a weighted summation method. The weight coefficients reflect the importance of each dimension to the overall treatment process. The weight coefficients are determined by analyzing the correlation strength between changes in different dimension indicators and the final treatment outcome in historical data. The weight coefficient for the time dimension is set to 0.4, the weight coefficient for the resource dimension is set to 0.2, and the weight coefficient for the clinical endpoint achievement dimension is set to 0.4. Before inputting into the quantitative model, the measured values ​​of each dimension need to be normalized by dividing the delay time by the preset maximum allowable delay value, dividing the resource cost by the preset average resource cost per cycle, and inputting the decay value directly as a percentage decimal.

[0051] In some embodiments, the quantification model outputs a comprehensive score as a quantitative value of the potential impact of a specific therapeutic behavior. The formula for calculating the comprehensive score is shown below:

[0052] in: This represents the overall impact score output by the quantification model. This represents the weighting coefficient for the time dimension. This indicates that the normalized time dimension affects the indicator value. This represents the weight coefficient for the resource dimension. This indicates that the normalized resource dimension affects the indicator value. This represents the weighting coefficient of the clinical endpoint achievement dimension. This indicates the normalized clinical endpoint achievement dimension impact index value.

[0053] It is understandable that the weighting coefficients of the quantitative model are not fixed. In specific applications, they can be adjusted according to different treatment center strategies or clinical priorities. For example, centers that focus more on treatment efficiency may increase the weighting coefficients for the time dimension, while centers that focus more on resource management may increase the weighting coefficients for the resource dimension. After deployment, the quantitative model can still use newly generated efficacy event chain data to iteratively optimize the weighting coefficients.

[0054] Optional, overall rating This will be mapped to a qualitative impact level, for example, 0-30 points for "minor impact", 31-70 points for "moderate impact", and 71-100 points for "serious impact". Based on the calculated score of 58, the potential impact of the specific therapeutic behavior of "overreaction" is assessed as "moderate impact".

[0055] See Figure 4In the monitoring of platelet-rich endometrial perfusion efficacy in thin endometrium, the synergistic trend of endometrial thickness (mm) and perfusion efficiency (%) was visually presented within treatment cycles 1 to 5, and the "overreaction point" was identified. Specifically, in treatment cycles 1 to 2, endometrial thickness increased from 6.2mm to 7.5mm, and perfusion efficiency simultaneously increased from 85% to 88% (corresponding to 7.8mm), showing a positive synergistic response in the early stages of treatment. In treatment cycle 3 (overreaction stage), endometrial thickness dropped back to 6.8mm, and perfusion efficiency plummeted to 72%, consistent with the key turning point characteristics of the "overreaction" efficacy behavior. In treatment cycles 4 to 5, endometrial thickness recovered to 7.8mm (returning to the level of cycle 2), and perfusion efficiency gradually recovered to 83%, reflecting the restorative effect of subsequent intervention on efficacy. Through the dynamic correlation of the two indicators, the short-term disturbance and recovery process of the "overreaction" event on the treatment process was clearly depicted, providing visualized physiological parameters for abnormal pattern matching in the efficacy behavior coding database.

[0056] Example 5: Analyzing the causal relationships and temporal proximity between adjacent segments in a sequence of therapeutic behavior segments. Based on these relationships and temporal proximity, multiple closely related therapeutic behavior segments are aggregated into a larger therapeutic event unit. All therapeutic event units are linked chronologically to form a complete therapeutic event chain, and key turning points where significant changes in treatment effects occur are identified within this chain.

[0057] In practice, the causal relationship and temporal proximity between adjacent segments in the therapeutic behavior sequence were analyzed. The sequence consisted of three segments arranged chronologically: segment A was labeled "rapid rise in blood flow," segment B was labeled "high-value fluctuation," and segment C was labeled "coordinated decrease in flow velocity and pressure." The analysis determined that the ending event of segment A, "peak blood flow velocity," was the direct cause of the beginning event of segment B, "high-value fluctuation in blood flow." Segments A and B were closely connected temporally, with an interval less than a preset threshold. The analysis also determined that there was no direct causal relationship between segment C and segment B, but the start time of segment C fell within a reasonable physiological response time window after the end of segment B, demonstrating temporal proximity.

[0058] In some embodiments, based on causal logic and temporal proximity, multiple closely related therapeutic behavior segments are aggregated into a larger therapeutic event unit. The system, according to preset aggregation rules, merges the first segment and the second segment, which have a direct causal relationship and a time interval below a threshold. Segment A and segment B satisfy the above conditions and are therefore aggregated into a new event unit, named the "overreaction period." Segment C only has temporal proximity with the "overreaction period" event unit and does not satisfy a strong causal relationship; therefore, segment C is not aggregated into this event unit but remains independent or is correlated with other subsequent segments.

[0059] Optionally, all therapeutic event units can be linked chronologically to form a complete therapeutic event chain. The "overreaction period" event unit serves as a node in the chain, with its start time being the start time of segment A and its end time being the end time of segment B. An independent segment C serves as the next node in the chain. Key inflection points where significant changes in treatment efficacy occur are identified within the chain. These inflection points are identified based on the extreme values ​​or sign changes of the rate of change of key physiological parameters within or between adjacent event units. Within the "overreaction period" event unit, the moment when the rate of change of the mean perfusion index changes from consistently positive to violently alternating between positive and negative is identified as a key inflection point, named the "loss of stability point." At the boundary between the end of the "overreaction period" event unit and the beginning of segment C, the moment when the trend of peak perfusion velocity changes from stable to significantly decreasing is identified as another key inflection point, named the "perfusion dynamic decay point."

[0060] It is understandable that the identification of key inflection points relies on real-time analysis of specific parameter sequences. The analysis algorithm continuously monitors the first derivative of sequences such as mean perfusion index and peak perfusion velocity. When the absolute value of the derivative exceeds a set threshold or the sign of the derivative changes and persists for a certain period of time, the algorithm determines that a significant change has occurred at that moment, and then marks it as a candidate for key inflection point. Finally, it is confirmed after being verified in relation to the context of the event chain.

[0061] In practice, whether two therapeutic behavior segments meet the condition of temporal proximity is determined by a quantified proximity score, calculated as follows:

[0062] in: This represents the calculated time proximity score. It is a preset attenuation coefficient. This represents the end timestamp of the previous therapeutic behavior segment m. This represents the start timestamp of the next therapeutic behavior segment n. This represents the absolute value of the difference between two timestamps. This formula quantifies the temporal proximity of two consecutive therapeutic behavior segments; the closer the score is to 1, the stronger the temporal proximity. When the score is higher than the preset aggregation threshold, the system considers that the two segments meet the time proximity condition, which can be used as the basis for considering whether to aggregate them together with the causal logical relationship.

[0063] See Figure 5 Using time as the horizontal axis, the changes in mean perfusion index (left vertical axis) and peak perfusion velocity (right vertical axis) are simultaneously displayed, and the therapeutic behavior segments and key turning points are combined to divide the treatment into stages. Specifically, the three therapeutic behavior segments are distinguished by background color: Segment A (light blue background) is the "rapid rise in blood flow," in which both mean perfusion index and peak perfusion velocity show a significant upward trend, reflecting the enhanced perfusion dynamics in the early stage of treatment; Segment B (light yellow background) is the "high-value fluctuation period," in which both parameters enter the high-value range and are accompanied by violent fluctuations, reflecting the dynamic instability of the perfusion state; Segment C (light pink background) is the "coordinated decline of flow velocity and pressure period," in which both parameters show a simultaneous downward trend, indicating the decay of perfusion efficacy. Two key turning points are marked with dashed lines: the purple dashed line corresponds to the "loss of stability point," located within segment B, marking the moment when the rate of change of the mean perfusion index changes from a sustained positive to a violent alternation between positive and negative, signifying the transition of the perfusion state from an orderly increase to a fluctuating and unstable phase; the green dashed line corresponds to the "perfusion dynamics decay point," located at the boundary between segments B and C, marking the moment when the peak perfusion velocity trend changes from stable to a significant decrease, indicating a core turning point in perfusion dynamics. The dynamic correlation characteristics of the curves clearly present the synergistic evolution of physiological parameters during the treatment process: the parameter changes from segment A to segment B reflect the aggregation logic (causal driving + temporal proximity) of the "overreaction period" event unit, while the independent decrease in segment C reflects the subsequent phased decay of perfusion efficacy, providing an intuitive basis for the construction of the therapeutic behavior event chain.

[0064] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent monitoring of the therapeutic effect of platelet-rich perfusion in thin endometrium, characterized in that, The method includes: Patient data from multiple independent monitoring sources were standardized to form a fused data stream that includes endometrial morphological parameters, perfusion hydrodynamic parameters, and treatment stage labels; Based on the fused data stream, cross-cycle feature derivation is performed to generate a derived feature set that reflects the synergistic change pattern between the treatment process and physiological parameters; Using the derived feature set, a therapeutic behavior coding library containing normal response patterns and multiple predefined abnormal response patterns is constructed; After converting the real-time collected medical monitoring data into a standard format, it is matched frame by frame with the patterns in the therapeutic behavior coding library and similarity is calculated. The matching results and confidence scores are then output. Based on the matching results and confidence scores, the corresponding pattern analysis path is activated to perform deep structural analysis on real-time medical monitoring data and generate a sequence of therapeutic behavior fragments with timestamps. The therapeutic behavior segments are context-linked and aggregated to form a complete therapeutic event chain, and key turning points are marked. Based on the aforementioned efficacy event chain, a multi-dimensional impact assessment is conducted to quantify the potential impact of specific efficacy behaviors on the overall treatment process, generating structured efficacy monitoring reports and decision support signals.

2. The intelligent monitoring method for platelet-rich perfusion therapy in thin endometrium according to claim 1, characterized in that, The standardization of patient data from multiple independent monitoring sources forms a fused data stream containing endometrial morphological parameters, perfusion hydrodynamic parameters, and treatment stage labels, specifically including: Receives raw, heterogeneous data from ultrasound imaging equipment, perfusion recording systems, and clinical information systems; The original heterogeneous data is time-stamp aligned and units are standardized to resolve the differences in time series and units between different monitoring sources; Extract the core metrics from the aligned data and attach treatment stage labels containing treatment cycles and interventions; The core indicators labeled with the treatment stage are encapsulated and concatenated in chronological order to generate the fused data stream.

3. The intelligent monitoring method for platelet-rich endometrial perfusion therapy according to claim 1, characterized in that, Based on the fused data stream, cross-cycle feature derivation is performed to generate a derived feature set reflecting the synergistic changes in treatment progress and physiological parameters, specifically including: From the fused data stream, a data window containing several consecutive treatment cycles is slidably extracted; Within each data window, the delayed response, rate of change, and trend consistency index of a specific physiological parameter relative to the treatment phase label are calculated. The calculation results of all data windows are aggregated to form the derived feature set that characterizes the dynamic evolution of physiological parameters driven by the treatment process.

4. The intelligent monitoring method for platelet-rich perfusion therapy in thin endometrium according to claim 3, characterized in that, The step of constructing a therapeutic behavior coding library containing normal response patterns and multiple predefined abnormal response patterns using the derived feature set specifically includes: Based on successful treatment cases in historical data, common patterns are summarized from the derived feature set and encoded as the normal response pattern; Based on clinical medical knowledge, typical adverse reaction or inadequate efficacy scenarios are defined and mapped to the space of the derived feature set to construct a variety of predefined abnormal response patterns. The normal response pattern and various predefined abnormal response patterns are structured, stored, and indexed to form the therapeutic behavior coding library.

5. The intelligent monitoring method for platelet-rich perfusion therapy in thin endometrium according to claim 4, characterized in that, The process of converting real-time collected medical monitoring data into a standard format and then performing frame-by-frame matching and similarity calculation with patterns in the therapeutic behavior coding library, outputting matching results and confidence scores, specifically includes: Perform the same standardization process on the real-time collected medical monitoring data to generate real-time standard data fragments; Calculate the distance or similarity measure between the real-time standard data fragment and each pattern in the therapeutic behavior coding library in the feature space; The pattern with the highest similarity is selected as the matching result, and the corresponding similarity metric is normalized and used as the confidence score.

6. The intelligent monitoring method for platelet-rich perfusion therapy in thin endometrium according to claim 5, characterized in that, Based on the matching results and confidence scores, the corresponding pattern analysis path is activated to perform deep structural analysis on the real-time medical monitoring data, generating a sequence of therapeutic behavior fragments with timestamps, specifically including: Based on the pattern type of the matching result, invoke the dedicated parsing rule and feature extractor bound to that pattern type; Using the dedicated parsing rules and feature extractor, the real-time medical monitoring data that produced the matching result is decomposed into fine-grained parts to identify sub-behaviors and state transitions. Each identified sub-behavior is labeled with its precise start and end timestamps, and arranged in chronological order to form the sequence of therapeutic behavior segments; The process utilizes specialized parsing rules and feature extractors to perform fine-grained decomposition of the real-time medical monitoring data that generated the matching result, identifying sub-behaviors and state transitions, specifically including: Based on the pattern type of the matching result, load the parsing template that is pre-associated with the pattern type; the parsing template defines the sub-behavior types and their logical order that are expected to be included for this type of pattern data; A dedicated feature extractor bound to the parsing template is used to calculate a set of fine-grained indicators for the real-time medical monitoring data, including the slope, numerical level, and fluctuation frequency of key physiological parameters within a specific time window; Based on the logical order and threshold rules defined in the parsing template, the calculated fine-grained index sequence is divided into continuous intervals, each interval representing a specific sub-behavior; Identify the moments when significant changes occur between adjacent sub-behavior intervals and mark these moments as state transition points; All identified sub-behaviors, their corresponding start and end time intervals, and the timestamps of state transition points are combined and output in a structured manner as the identification results of the sub-behaviors and state transitions.

7. The intelligent monitoring method for platelet-rich perfusion therapy in thin endometrium according to claim 6, characterized in that, The process of contextually associating and aggregating the therapeutic behavior fragment sequence to assemble it into a complete therapeutic event chain, and marking key turning points, specifically includes: Analyze the causal logic and temporal proximity between adjacent segments in the therapeutic behavior segment sequence; Based on the causal logic and temporal proximity, multiple closely related therapeutic behavior segments are aggregated into a larger therapeutic event unit. All therapeutic event units are linked in chronological order to form the complete therapeutic event chain, and the key turning points in the chain where the treatment effect changes significantly are identified.

8. The intelligent monitoring method for platelet-rich perfusion therapy in thin endometrium according to claim 1, characterized in that, The multi-dimensional impact assessment based on the therapeutic event chain quantifies the potential impact of specific therapeutic behaviors on the overall treatment process, specifically including: Define multiple dimensions for impact assessment, including time dimension, resource dimension, and clinical endpoint achievement dimension; Within the chain of therapeutic events, the specific therapeutic behavior to be evaluated is located, and the chain changes in its subsequent events are analyzed across multiple dimensions. Based on the magnitude and range of the aforementioned chain changes, the potential impact of the specific therapeutic behavior is calculated using a pre-defined quantitative model.

9. The intelligent monitoring method for platelet-rich perfusion therapy of thin endometrium according to claim 8, characterized in that, The process of comprehensively considering the magnitude and range of the chain changes, and calculating the potential impact of the specific therapeutic behavior using a pre-defined quantitative model, specifically includes: Quantifiable impact indicators are defined for the time dimension, resource dimension, and clinical endpoint achievement dimension, respectively; the time dimension impact indicator is the delay time in subsequent treatment stages, the resource dimension impact indicator is the additional medical resource cost, and the clinical endpoint achievement dimension impact indicator is the attenuation value of the expected improvement in key physiological parameters. Extract the actual measured values ​​of the influencing indicators of each dimension within a preset observation window after the occurrence of the specific therapeutic behavior from the therapeutic event chain; The actual measured values ​​of the influencing indicators of each dimension are input into the pre-trained quantitative model. The quantitative model is trained with historical efficacy event chain data and integrates multi-dimensional indicators through weighted summation. The weight coefficients reflect the importance of each dimension to the overall treatment process. The quantitative model outputs a comprehensive score as a quantitative value of the potential impact of the specific therapeutic behavior.

10. A thin-film endometrial platelet-rich perfusion therapy intelligent monitoring system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent monitoring method for the efficacy of platelet-rich perfusion in thin endometrium as described in any one of claims 1 to 9.

Images