Multi-source data fused early risk screening method and system for growth retardation of children

Abstract

The invention provides a multi-source data fused early risk screening method and system for growth retardation of children, and belongs to the technical field of medical information. The method comprises the following steps: acquiring a first predicted growth sequence and a second predicted growth sequence of a target child; performing compensation correction on the second predicted growth sequence based on the intergenerational compensation sequence to obtain a second corrected predicted growth sequence; performing fusion processing on the first predicted growth sequence and the second corrected predicted growth sequence, and establishing a target predicted growth sequence; and monitoring and acquiring a tracking and actually measured height sequence of the target child, performing comparative analysis on the tracking and actually measured height sequence and the target predicted growth sequence, and when the tracking and actually measured height sequence does not meet the target predicted growth sequence, generating a growth retardation early warning. The technical problem that early-stage growth retardation risks cannot be effectively identified due to insufficient early-warning accuracy caused by child growth prediction in the prior art is solved, and the technical effects of improving the child growth retardation risk prediction accuracy and realizing early-stage accurate early-warning are achieved.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, and in particular to a method and system for early risk screening of growth retardation in children that integrates multi-source data. Background Technology

[0002] Children's growth and development are important indicators for assessing their health. Growth retardation, as a significant manifestation of malnutrition and developmental abnormalities, seriously affects children's physical and mental well-being. Early identification and intervention of growth retardation are crucial for improving children's health.

[0003] Currently, existing methods for predicting child growth mainly rely on a single data source for analysis. For example, they may build predictive models based solely on an individual child's historical growth data and predict future growth by analyzing the changing trends of height, a growth indicator. However, these methods suffer from insufficient predictive accuracy in practical applications. Furthermore, existing technologies generally suffer from delayed warnings in identifying the risk of growth retardation. They often only detect abnormalities after significant deviations in growth have been observed, missing the optimal time for early intervention and failing to effectively identify early risks of growth retardation. Summary of the Invention

[0004] This invention addresses the technical problems of insufficient accuracy in early warning and inability to effectively identify early growth retardation risks in existing technologies by providing a method and system for early risk screening of growth retardation in children that integrates multi-source data.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides a method for early risk screening of growth retardation in children by integrating multi-source data, comprising: acquiring individual growth history data of a target child before a prediction start point; performing individual growth prediction on the target child based on the individual growth history data to obtain a first predicted growth sequence; acquiring the growth history data of the target child's parents; performing genetic growth prediction on the target child based on the parents' growth history data to obtain a second predicted growth sequence; determining the intergenerational compensation sequence of the target child; and performing compensation correction on the second predicted growth sequence based on the intergenerational compensation sequence to obtain a second corrected predicted growth sequence; fusing the first predicted growth sequence and the second corrected predicted growth sequence to establish a target predicted growth sequence; monitoring and acquiring the tracked measured height sequence of the target child; comparing and analyzing the tracked measured height sequence with the target predicted growth sequence; and generating a growth retardation warning when the tracked measured height sequence does not meet the target predicted growth sequence.

[0006] Secondly, this invention provides an early risk screening system for growth retardation in children that integrates multi-source data, comprising: an individual growth prediction module, used to acquire the individual growth history data of the target child before the prediction start point, and to perform individual growth prediction on the target child based on the individual growth history data to obtain a first predicted growth sequence; a genetic growth prediction module, used to acquire the growth history data of the target child's parents, and to perform genetic growth prediction on the target child based on the parents' growth history data to obtain a second predicted growth sequence; an intergenerational compensation correction module, used to determine the intergenerational compensation sequence of the target child, and to perform compensation correction on the second predicted growth sequence based on the intergenerational compensation sequence to obtain a second corrected predicted growth sequence; a sequence fusion processing module, used to fuse the first predicted growth sequence and the second corrected predicted growth sequence to establish a target predicted growth sequence; and a tracking monitoring and early warning module, used to monitor and acquire the tracking measured height sequence of the target child, compare and analyze the tracking measured height sequence with the target predicted growth sequence, and generate a growth retardation early warning when the tracking measured height sequence does not meet the target predicted growth sequence.

[0007] The beneficial effects of this invention are: The process involves: acquiring individual growth history data of the target child prior to the prediction start point; predicting individual growth based on this data to obtain a first predicted growth sequence, establishing an individualized prediction basis by analyzing the child's own growth patterns; acquiring the growth history data of the target child's parents; predicting genetic growth based on this data to obtain a second predicted growth sequence, thus introducing the influence of genetic factors on the child's growth and development; determining the intergenerational compensation sequence for the target child; and correcting the second predicted growth sequence based on this sequence to obtain a second corrected predicted growth sequence, thereby correcting biases in genetic prediction and improving its accuracy; fusing the first and second corrected predicted growth sequences to establish the target predicted growth sequence, achieving an organic combination of individual growth patterns and genetic factors for a more accurate comprehensive prediction result; and monitoring and acquiring the target child's measured height sequence, comparing and analyzing it with the target predicted growth sequence. If the measured height sequence does not meet the target predicted growth sequence, a growth retardation warning is generated, enabling early identification and timely warning of the risk of growth retardation.

[0008] This technical solution integrates two data sources: individual child growth history data and parental genetic data. Furthermore, by using intergenerational compensation sequences to correct genetic predictions, it effectively mitigates the impact of intergenerational growth differences, significantly improving the accuracy of child growth predictions. In addition, by real-time monitoring and comparative analysis of measured height sequences and target predicted growth sequences, abnormalities can be detected and early warnings generated at early stages of growth deviation. This enables early and accurate identification of growth retardation risks, providing strong support for early clinical intervention. Attached Figure Description

[0009] Figure 1 A flowchart illustrating the early risk screening method for childhood growth retardation that integrates multi-source data provided by the present invention; Figure 2 This is a schematic diagram of the structure of the early risk screening system for childhood growth retardation that integrates multi-source data provided by the present invention.

[0010] In the attached diagram, the components represented by each number are as follows: Individual growth prediction module 11, genetic growth prediction module 12, intergenerational compensation correction module 13, sequence fusion processing module 14, tracking monitoring and early warning module 15. Detailed Implementation

[0011] 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.

[0012] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0013] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0014] Example 1, as Figure 1 As shown, embodiments of the present invention provide an early risk screening method for childhood growth retardation that integrates multi-source data, including: S1. Obtain the individual growth history data of the target child before the prediction start point, and make individual growth predictions for the target child based on the individual growth history data to obtain the first predicted growth sequence.

[0015] Specifically, the prediction start point refers to the time point at which growth prediction begins, typically set at the current monitoring point of the target child. Before the prediction start point, the target child has already undergone a period of growth and development, forming an individual-specific growth trajectory. Individual growth history data is constructed by collecting height data from multiple historical monitoring points. Historical monitoring points are set monthly; for example, height measurements from 1 month and 2 months of age up to the month preceding the prediction start point are obtained, forming a monthly height data sequence as individual growth history data. This individual growth history data reflects the target child's own growth and development patterns and characteristics, providing an important data foundation for subsequent individual growth prediction.

[0016] Then, based on individual growth history data, a historical height sequence is generated. This historical height sequence is then processed by an individual growth predictor to obtain predicted height intervals for multiple future monitoring nodes, generating the first predicted growth sequence. The individual growth predictor can identify and learn the target child's individual growth pattern and infer the predicted height intervals for multiple future monitoring nodes after the initial prediction node, serving as the first predicted growth sequence. This first predicted growth sequence reflects future growth expectations derived from the target child's own growth history, laying the foundation for subsequent multi-source data fusion. Represented as predicted height intervals, this first predicted growth sequence reflects the uncertainty of prediction, providing a more reliable reference for clinical decision-making.

[0017] By acquiring the first predictive growth sequence, a foundation for growth prediction based on the target child's individual historical data was established. The first predictive growth sequence can capture the target child's own growth and development characteristics and trends, providing individualized predictive information for subsequent multi-source data fusion.

[0018] S2. Obtain the growth history data of the target child's parents, and perform genetic growth prediction on the target child based on the growth history data of the parents to obtain a second predicted growth sequence.

[0019] Specifically, a child's growth and development are not only influenced by their own historical growth patterns but also closely related to genetic factors. Parents' growth history data contains important genetic information and can provide a reference for predicting the growth and development trends of target children. Parents' growth history data is obtained by collecting height data from the target child's parents at various monitoring points throughout the entire lifecycle. Then, the parental growth history data is serialized to obtain the paternal and maternal historical height sequences. These sequences reflect the respective growth and development trajectories of the target child's parents, providing a data foundation for genetic growth prediction.

[0020] The paternal and maternal historical height sequences are input into a genetic growth predictor to obtain a full-cycle predicted height interval sequence for the target child. The genetic growth predictor is constructed based on a large set of paternal historical height sequences, maternal historical height sequences, and labeled height interval sequences of sample children. It can learn and identify the influence of genetic factors on children's growth and development. Subsequently, the full-cycle predicted height interval sequence is truncated based on the prediction start node to obtain a second predicted growth sequence. The second predicted growth sequence reflects the future growth trend of the target child predicted based on genetic factors, complementing the first predicted growth sequence and laying the foundation for subsequent multi-source data fusion.

[0021] By obtaining a second predicted growth sequence, a growth prediction dimension based on genetic factors was established, which helps to improve the accuracy of screening for growth retardation risk.

[0022] S3. Determine the intergenerational compensation sequence of the target child, and perform compensation correction on the second predicted growth sequence based on the intergenerational compensation sequence to obtain the second corrected predicted growth sequence.

[0023] Specifically, while genetic growth predictors can predict the growth trend of a target child based on parental growth history data, direct genetic predictions are subject to bias due to differences in growth and development environments, nutritional levels, and other factors between generations. To improve prediction accuracy, an intergenerational compensation mechanism is introduced to correct the genetic prediction results.

[0024] First, similarity matching is performed based on the target child's full-cycle predicted height interval sequence to obtain multiple matching full-cycle predicted height interval sequences, each corresponding to multiple sample identifiers. Through similarity matching, historical samples with genetically similar predictions to the target child can be found. Then, multiple matching full-cycle measured height sequences are extracted based on the multiple sample identifiers. These matching full-cycle measured height sequences reflect the actual growth of similar children. Next, based on the multiple matching full-cycle measured height sequences and multiple matching full-cycle predicted height interval sequences, an intergenerational compensation sequence is obtained. By calculating the median of each matching full-cycle predicted height interval sequence, multiple matching full-cycle predicted height median sequences are obtained. The difference between each matching full-cycle measured height sequence and its corresponding matching full-cycle predicted height median sequence is calculated to obtain multiple deviation sequences. These deviation sequences are statistically summarized by monitoring nodes, and the mean deviation of multiple deviation values ​​at each monitoring node is calculated. The mean deviations of each monitoring node are then combined sequentially to form the intergenerational compensation sequence. Then, the intergenerational compensation sequence was applied to the second predicted growth sequence to compensate and correct the predicted height interval in the second predicted growth sequence, thus obtaining the second corrected predicted growth sequence.

[0025] Through intergenerational compensation correction, the second-corrected predicted growth sequence can better reflect the actual growth patterns of contemporary children, eliminate the influence of intergenerational environmental differences on genetic prediction, and provide a more accurate and reliable prediction basis for subsequent multi-source data fusion.

[0026] S4. The first predicted growth sequence and the second corrected predicted growth sequence are fused together to establish the target predicted growth sequence.

[0027] Specifically, the first predicted growth sequence is based on the individual growth history data of the target child, reflecting the child's own growth and development characteristics and trends; the second corrected predicted growth sequence is based on the parents' growth history data and has undergone intergenerational compensation correction, reflecting the influence of genetic factors on the child's growth. A fusion processing strategy is used to process the first and second predicted growth sequences, utilizing the prediction information from these two different sources to improve the accuracy and reliability of the predictions.

[0028] First, a first preset weight and a second preset weight are obtained. These weights are used to balance the contributions of individual historical predictions and genetic predictions to the final prediction result and can be set based on expert experience. Then, the first predicted growth sequence and the second corrected predicted growth sequence are aligned according to the monitoring nodes to ensure accurate correspondence of prediction data for the same future monitoring nodes. Based on the alignment results, a weighted fusion interval is calculated for the predicted height intervals at each monitoring node. The lower bound of the weighted fusion interval is the weighted average of the lower bounds of the first predicted height interval and the second corrected predicted height interval, and the upper bound is the weighted average of the upper bounds of the first predicted height interval and the second corrected predicted height interval. Finally, the weighted fusion intervals from each monitoring node are combined sequentially to form the target predicted growth sequence.

[0029] Through fusion processing, the target predicted growth sequence integrates information from both the individual's historical growth pattern and genetic factors. It maintains sensitivity to the individual characteristics of the target child while incorporating the influence of genetic background, forming a more comprehensive and accurate basis for growth prediction and providing a reliable predictive reference standard for subsequent screening for growth retardation risks.

[0030] S5. Monitor and acquire the actual height sequence of the target child, compare and analyze the actual height sequence with the target predicted growth sequence, and generate a growth retardation warning when the actual height sequence does not meet the target predicted growth sequence.

[0031] Specifically, after establishing the target predictive growth sequence, the growth and development status of the target children is assessed by continuously monitoring their actual growth, and potential risks of growth retardation are identified in a timely manner.

[0032] First, the number of tracking and monitoring nodes is set to determine the number of future monitoring nodes that need to be continuously monitored. Then, according to the number of tracking and monitoring nodes, the actual height data of the target child at the corresponding monitoring nodes after the predicted starting node is collected periodically to form a tracked measured height sequence, reflecting the target child's true growth status at subsequent monitoring nodes. Subsequently, the tracked measured height sequence is compared and analyzed with the target predicted growth sequence to obtain the measured deviation from the lower limit. Specifically, the measured height value of each monitoring node in the tracked measured height sequence is compared one by one with the predicted height range of the corresponding monitoring node in the target predicted growth sequence; when the measured height value is lower than the lower limit of the predicted height range, it is determined to be a node deviating from the lower limit; when the measured height value is within the predicted height range, it is determined to be a normal node; the number of nodes deviating from the lower limit is counted to obtain the measured deviation from the lower limit. The ratio of the measured deviation from the lower limit to the number of tracking and monitoring nodes is calculated to obtain the deviation from the lower limit ratio. When the deviation from the lower limit ratio is greater than or equal to the risk ratio threshold, it is determined that the tracked measured height sequence does not meet the target predicted growth sequence, and a growth retardation warning is generated.

[0033] Through continuous monitoring and comparative analysis, abnormal growth patterns in target children can be identified in a timely manner. When the actual growth deviates significantly from the expected range, an early warning signal can be issued in a timely manner, providing clinicians and parents with an opportunity for early intervention and effectively preventing the occurrence of growth retardation in children.

[0034] Furthermore, the individual growth history data of the target child prior to the prediction start point is obtained, and individual growth prediction is performed on the target child based on the individual growth history data to obtain a first predicted growth sequence, including: S11. Collect the height data of the target child at multiple historical monitoring nodes before the prediction start node, as the individual growth history data of the target child; S12. Generate a historical node height sequence based on the individual's historical growth data, and call the individual growth predictor to process the historical node height sequence to obtain the predicted height range of multiple future monitoring nodes, and generate the first predicted growth sequence.

[0035] In a preferred embodiment, firstly, height data of the target child at multiple historical monitoring points prior to the prediction start point are collected as the target child's individual growth history data. Specifically, height measurement data of the target child is collected from each month of age after birth until the prediction start point, according to a preset monitoring frequency. Monitoring points are set monthly, meaning that a fixed date each month is used as a monitoring point for height measurement. For example, if the prediction start point is set at 24 months of age for the target child, then height data of the child at each historical monitoring point from 1 month, 2 months to 23 months of age needs to be collected. The height data arranged in time series constitutes the individual growth history data reflecting the target child's individual growth and development trajectory.

[0036] Subsequently, a historical node height sequence is generated based on individual growth history data. This historical node height sequence is then processed by an individual growth predictor to obtain predicted height intervals for multiple future monitoring nodes, generating the first predicted growth sequence. Specifically, the collected height data is arranged according to the chronological order of the monitoring nodes to form a historical node height sequence. This historical node height sequence serves as the input data for the individual growth predictor. Through processing and analysis of this sequence, the individual growth pattern and development trend of the target child can be identified. The individual growth predictor includes multiple growth prediction branches. Each branch performs prediction calculations based on the historical node height sequence, outputting a predicted height value for each future monitoring node, resulting in a predicted height sequence. For each future monitoring node, the predicted height values ​​from all growth prediction branches are collected. The maximum value is taken as the upper bound of the predicted height interval for that monitoring node, and the minimum value is taken as the lower bound, thus determining the predicted height interval for that monitoring node. Combining the predicted height intervals of each future monitoring node in chronological order constitutes the first predicted growth sequence.

[0037] Through the above steps, an individualized prediction foundation based on the target child's own growth history was established, providing individual growth information for subsequent multi-source data fusion analysis.

[0038] Furthermore, the construction steps of the individual growth predictor include: S121. Collect a set of historical children's growth records, which includes multiple historical children's growth records, and each historical children's growth record includes the height data of the historical children at each monitoring node; S122. Based on the predicted starting node, each of the historical children's growth records is divided into historical node height data and future node height data. S123. Construct a set of sample children's historical node height sequences based on historical node height data from multiple historical children's growth records, and construct a set of sample children's future node height sequences based on future node height data from multiple historical children's growth records. S124. Construct multiple growth prediction networks. Train multiple growth prediction networks until convergence based on the historical node height sequence set of the sample children and the future node height sequence set of the sample children, and obtain multiple growth prediction branches. S125. The multiple growth prediction branches are integrated to obtain the individual growth predictor.

[0039] In a preferred embodiment, firstly, a historical children's growth record set is collected. This set includes multiple historical children's growth records, each containing height data for a specific child at various monitoring points. Specifically, a large amount of complete growth and development data of children is collected from data sources such as medical institutions, child health centers, and health record systems to form the historical children's growth record set. Each historical children's growth record details a child's height measurements collected at monthly monitoring points, starting from one month after birth. For example, a child's growth record might include height data for 48cm at 1 month, 52cm at 2 months, 55cm at 3 months, etc., up to each monitoring point throughout the child's complete growth cycle, constituting the child's complete growth and development trajectory. The historical children's growth record set contains a large number of children's growth records, providing a sufficient data foundation for subsequent training of individual growth predictors.

[0040] Then, based on the prediction start point, each historical child's growth record is divided into historical node height data and future node height data. The prediction start point serves as the dividing line for each historical child's growth record. For example, if the prediction start point is set to 24 months of age, then the height data from 1 month to 23 months of age in each historical child's growth record is used as historical node height data, and the height data after 24 months of age is used as future node height data. This division method simulates a real-world prediction scenario, that is, predicting future growth trends based on a child's past growth history, ensuring complete consistency between the training data and the actual application scenario.

[0041] Subsequently, a sample set of historical height sequence data for children was constructed based on historical height data from multiple historical children's growth records, and a sample set of future height sequence data for children was constructed based on future height data from multiple historical children's growth records. All historical height data for all children were organized in time-series format, with each child's historical height data constituting a historical height sequence. The combined historical height sequence data of all children formed the sample set of historical height sequence data, which served as the input dataset for training. Simultaneously, all future height data for all children were organized in time-series format, with each child's future height data constituting a future height sequence. The combined future height sequence data of all children formed the sample set of future height sequence data, which served as the target output dataset for training.

[0042] Next, multiple growth prediction networks are constructed. These networks are trained until convergence using both historical height sequences and future height sequences of sample children, resulting in multiple growth prediction branches. Different neural network architectures, parameter settings, and training strategies are employed to construct these networks. The neural network architecture can be chosen from temporal prediction models such as Long Short-Term Memory (LSTM), gated recurrent units (GRU), and Transformers. Each growth prediction network is trained using the historical height sequences of sample children as input and the future height sequences as supervision signals. The network parameters are continuously adjusted using backpropagation to minimize the error between the predicted output and the true target until the network parameters converge. After training, multiple growth prediction branches with different learning characteristics and predictive capabilities are obtained. For example, a growth prediction network constructed using a Long Short-Term Memory (LSTM) network architecture, with 128 hidden layer nodes and a learning rate of 0.001, trained using the Adam optimizer, can effectively capture short-term growth patterns in children by learning the temporal characteristics of historical height data. Another example is a growth prediction network constructed using a gated recurrent unit (GRU) architecture, with 256 hidden layer nodes and a learning rate of 0.0005, while employing Dropout regularization to prevent overfitting. Compared to the LSM architecture, this network has stronger long-term memory capabilities and can better capture long-term trends and cyclical changes in children's growth. By constructing multiple growth prediction networks with different architectures and learning characteristics, it is possible to learn children's growth patterns from different perspectives, such as short-term fluctuations and long-term trends. In practical applications, the number of growth prediction networks can be set according to computational resources and prediction accuracy requirements. Generally, it is preferred to set 10 growth prediction networks to ensure sufficient diversity and accuracy in ensemble prediction; at least 2 growth prediction networks are required to achieve basic ensemble effects. Too few networks may lead to simplistic prediction results, while too many networks will increase computational complexity and training costs. Therefore, the number of prediction networks to grow needs to be determined by experts based on the actual application, balancing prediction performance and computational efficiency.

[0043] Subsequently, multiple growth prediction branches are integrated to obtain an individual growth predictor. An ensemble learning strategy is employed to combine multiple growth prediction branches into an individual growth predictor. Specifically, a fully connected layer is placed after the output layer of each growth prediction branch to fuse the prediction results. In actual prediction, the target child's historical height sequence is simultaneously input into all growth prediction branches. Each growth prediction branch performs prediction calculations based on the historical height sequence, outputting a predicted height value for each future monitoring node, resulting in a predicted height sequence. Then, the predicted height sequences from all growth prediction branches are used as input to the fully connected layer, which integrates the predicted height sequences from multiple growth prediction branches. For each future monitoring node, the fully connected layer collects the predicted height values ​​from all growth prediction branches for that future monitoring node, taking the maximum value as the upper bound of the predicted height interval for that monitoring node and the minimum value as the lower bound, thus determining the predicted height interval for that monitoring node. This integration method fully utilizes the predictive advantages of different branches, improving the overall prediction accuracy and robustness.

[0044] Through the detailed construction process described above, the individual growth predictor not only has the ability to predict future growth trends based on children's historical growth data, but also improves the reliability and stability of the prediction through multi-branch integration, providing technical support for subsequent individual growth prediction.

[0045] Furthermore, the growth history data of the target child's parents is obtained, and genetic growth prediction is performed on the target child based on the parental growth history data to obtain a second predicted growth sequence, including: S21. Collect the height data of the target child's parents at the full-cycle monitoring nodes to obtain the parents' growth history data; S22. The parental growth history data is serialized to obtain the father's historical height sequence and the mother's historical height sequence. S23. Input the paternal historical height sequence and maternal historical height sequence into the genetic growth predictor to obtain the full-cycle predicted height interval sequence of the target child. The genetic growth predictor is constructed based on the sample paternal historical height sequence set, the sample maternal historical height sequence set, and the sample child labeled height interval sequence set.

[0046] S24. Based on the predicted starting node, the full-cycle predicted height interval sequence is truncated to obtain the second predicted growth sequence.

[0047] In a preferred embodiment, firstly, height data of the target child's parents at various monitoring points throughout the entire growth cycle are collected to obtain parental growth history data. Specifically, complete height measurement data of the target child's father and mother are collected from each month of age after birth until adulthood. The entire growth cycle covers the entire growth and development period starting from one month of age, including infancy, childhood, adolescence, and adulthood. For example, the father's growth history data includes height data at each monitoring point from 1 month of age and 2 months of age until adulthood, and the mother's growth history data similarly includes height data at each monitoring point throughout her complete growth cycle. This complete parental growth history data provides an informational basis for subsequent genetic growth prediction.

[0048] Then, the parents' growth history data were serialized to obtain the father's historical height sequence and the mother's historical height sequence. The father's full-cycle height data was arranged according to the time sequence of the monitoring nodes to form the father's historical height sequence; the mother's full-cycle height data was arranged according to the time sequence of the monitoring nodes to form the mother's historical height sequence. These two sequences respectively reflect the complete growth and development characteristics of the target child's parents.

[0049] Subsequently, the paternal and maternal historical height sequences are input into the genetic growth predictor to obtain the target child's full-cycle predicted height interval sequence. The genetic growth predictor is a model specifically designed for genetic growth prediction, built upon a large amount of sample data. It is constructed from a set of sample paternal historical height sequences, a set of sample maternal historical height sequences, and a set of sample child labeled height interval sequences. Specifically, first, a large amount of full-cycle historical height sequence data from both parents is collected. The collected paternal full-cycle historical height sequences are organized in sequence form to form the sample paternal historical height sequence set; the collected maternal full-cycle historical height sequences are also organized in sequence form to form the sample maternal historical height sequence set. Simultaneously, genetic experts, based on genetic growth and development theories and extensive statistical analysis, label the expected height range for the corresponding child at each monitoring node for different combinations of parental growth patterns, forming a set of sample child labeled height interval sequences. These labeled height intervals reflect the reasonable height range for children at various growth stages under a specific genetic background; the labeling result for each monitoring node is a height interval rather than a single numerical value. Next, a genetic growth prediction network was constructed. This network employs a dual-input structure, receiving historical height sequences from both the father and mother. Internally, a multi-layer neural network is used to extract and fuse features from the paternal and maternal growth information. During training, the historical height sequences from both the father and mother are simultaneously input into the network, with the target output being the set of height interval sequences labeled by the children. Since the target output is in interval form, the network's output layer is designed with a dual-output structure, predicting the upper and lower bounds of the height intervals for each monitoring node. Backpropagation is used to optimize the network parameters, ensuring that the predicted upper and lower bounds of the height intervals are as close as possible to the intervals labeled by experts. After training, a genetic growth predictor is obtained, capable of receiving historical height sequences from any father and mother and directly outputting the corresponding full-cycle predicted height interval sequence. The prediction result for each monitoring node is a height interval, reflecting the uncertainty and reasonable range of genetic growth prediction.

[0050] Subsequently, the full-cycle predicted height interval sequence is truncated based on the predicted starting node to obtain the second predicted growth sequence. Since practical applications only require predicting the growth of the target child after the predicted starting node, it is necessary to extract the portion after the predicted starting node from the full-cycle predicted height interval sequence. Using the predicted starting node as the truncating point, the predicted height intervals of all monitoring nodes after that node are extracted and combined in chronological order to form the second predicted growth sequence.

[0051] Through the above steps, a growth prediction dimension based on genetic factors was established. The second predicted growth sequence reflects the influence of parental genetic factors on the growth and development of the target child, providing important genetic prediction information for subsequent multi-source data fusion.

[0052] Further, the intergenerational compensation sequence of the target child is determined, and the second predicted growth sequence is compensated and corrected based on the intergenerational compensation sequence to obtain a second corrected predicted growth sequence, including: S31. Perform similarity matching retrieval based on the full-cycle predicted height interval sequence of the target child to obtain multiple matching full-cycle predicted height interval sequences, and multiple matching full-cycle predicted height interval sequences correspond to multiple sample identifiers. S32. Extract multiple matching full-cycle measured height sequences based on the multiple sample identifiers; S33. Based on the multiple matched full-cycle measured height sequences and the multiple matched full-cycle predicted height interval sequences, the intergenerational compensation sequence is obtained. S34. Apply the intergenerational compensation sequence to the second predicted growth sequence, and compensate and correct the predicted height interval in the second predicted growth sequence to obtain the second corrected predicted growth sequence.

[0053] In a preferred embodiment, firstly, a similarity matching search is performed based on the target child's full-cycle predicted height interval sequence to obtain multiple matching full-cycle predicted height interval sequences, each corresponding to a unique sample identifier. Specifically, the target child's full-cycle predicted height interval sequence is compared with each historical full-cycle predicted height interval sequence stored in the historical sample database. For each historical sample, the predicted height intervals of the target child and the historical sample at the corresponding monitoring nodes are compared one by one, and the intersection of the two intervals is calculated. If the intersection is not empty, it is recorded as 1; if the intersection is empty, it is recorded as 0. The matching results of all monitoring nodes are summed to obtain the number of matching nodes, and then the ratio of the number of matching nodes to the total number of monitoring nodes is calculated as a similarity index. When the similarity index is greater than a preset threshold, the historical sample is determined to be similar to the target child, its full-cycle predicted height interval sequence is taken as a matching sequence, and the corresponding sample identifier is recorded. In this way, multiple historical samples with similar genetic prediction characteristics to the target child are selected, resulting in multiple matching full-cycle predicted height interval sequences. Each matching full-cycle predicted height interval sequence corresponds to a unique sample identifier for subsequent data extraction.

[0054] Then, using the obtained sample identifiers, corresponding full-cycle measured height sequences were extracted from the historical sample database, resulting in multiple matched full-cycle measured height sequences that reflect the actual growth records of the sample children. These matched full-cycle measured height sequences record the actual height measurements of children with similar predicted characteristics to the target children during their actual growth process, providing a realistic reference basis for intergenerational compensation analysis.

[0055] Subsequently, based on multiple matched full-cycle measured height sequences and multiple matched full-cycle predicted height interval sequences, an intergenerational compensation sequence was obtained. Specifically, the median of each predicted height interval in each matched full-cycle predicted height interval sequence was calculated, resulting in multiple matched full-cycle predicted height median sequences. The difference between each matched full-cycle measured height sequence and its corresponding matched full-cycle predicted height median sequence was calculated, resulting in multiple deviation sequences. These multiple deviation sequences were statistically summarized according to monitoring nodes, and the mean deviation of multiple deviation values ​​at each monitoring node was calculated. The mean deviations of each monitoring node were then combined sequentially to form the intergenerational compensation sequence. This intergenerational compensation sequence reflects the systematic adjustment of actual growth relative to genetic prediction under the current growth environment.

[0056] Next, the intergenerational compensation sequence was applied to the second predicted growth sequence to compensate and correct the predicted height intervals in the second predicted growth sequence, resulting in the second corrected predicted growth sequence. The compensation values ​​of each monitoring node in the intergenerational compensation sequence were added to the predicted height intervals of the corresponding monitoring nodes in the second predicted growth sequence. By adjusting the upper and lower bounds of the intervals, the influence of intergenerational environmental differences on genetic prediction was eliminated, making the prediction results closer to the actual growth patterns of children today.

[0057] Through the above-mentioned intergenerational compensation correction process, the second corrected predictive growth sequence can more accurately reflect the expected growth trend of the target child based on genetic factors in the current growth environment, providing a more reliable genetic prediction basis for subsequent multi-source data fusion.

[0058] Furthermore, based on the multiple matched full-cycle measured height sequences and the multiple matched full-cycle predicted height interval sequences, the intergenerational compensation sequence is obtained, including: S331. Calculate the median of each predicted height interval in each matched full-cycle predicted height interval sequence to obtain multiple matched full-cycle predicted height median sequences. S332. Calculate the difference between the measured height sequence of each matching full-cycle and the corresponding predicted height median sequence of the matching full-cycle to obtain multiple deviation sequences; S333. The multiple deviation sequences are statistically summarized according to the monitoring nodes, the mean deviation of multiple deviation values ​​at each monitoring node is calculated, and the mean deviation of each monitoring node is combined in sequence to form the intergenerational compensation sequence.

[0059] In a preferred embodiment, firstly, the median of each predicted height interval in each matched full-cycle predicted height interval sequence is calculated, resulting in multiple matched full-cycle predicted height median sequences. Specifically, for each matched full-cycle predicted height interval sequence, the median of its predicted height interval is calculated for each monitoring node. The median of a predicted height interval is the arithmetic mean of the upper and lower bounds of the interval, i.e., the median equals the sum of the upper and lower bounds of the interval divided by 2. The medians of each monitoring node are arranged in chronological order to form the matched full-cycle predicted height median sequence corresponding to that matched full-cycle predicted height interval sequence. By performing the same processing on all matched full-cycle predicted height interval sequences, multiple matched full-cycle predicted height median sequences are obtained.

[0060] Then, the difference between each matched full-cycle measured height sequence and its corresponding matched full-cycle predicted median height sequence is calculated to obtain multiple deviation sequences. Specifically, each matched full-cycle measured height sequence is compared with its corresponding full-cycle predicted median height sequence at each monitoring node, and the difference between the measured height value and the predicted median height at each monitoring node is calculated. The measured height value minus the predicted median height is the deviation value for that node; a positive value indicates that the measured height is higher than the genetic prediction, and a negative value indicates that the measured height is lower than the genetic prediction. The deviation values ​​of each monitoring node are combined in chronological order to form corresponding deviation sequences. By performing the same calculation on each matched full-cycle measured height sequence and its corresponding matched full-cycle predicted median height sequence, multiple deviation sequences are obtained, reflecting the deviation of each similar sample from the genetic prediction during actual growth.

[0061] Subsequently, multiple deviation sequences were statistically summarized according to monitoring nodes. The mean deviation of multiple deviation values ​​at each monitoring node was calculated, and the mean deviations of each monitoring node were sequentially combined to form an intergenerational compensation sequence. Specifically, all deviation sequences were categorized and statistically analyzed according to monitoring nodes. For each monitoring node, all deviation values ​​of that monitoring node were collected from all deviation sequences, and the arithmetic mean of these deviation values ​​was calculated as the mean deviation for that monitoring node. The mean deviation reflects the average degree of deviation of a group of children with similar predictive characteristics from the genetic prediction at that growth stage. By sequentially combining the mean deviations of each monitoring node in chronological order to form an intergenerational compensation sequence, the systematic impact of intergenerational changes in the growth environment on the genetic prediction results was demonstrated, providing a quantitative compensation benchmark for subsequent prediction correction.

[0062] Through the above calculation process, intergenerational compensation sequences can effectively capture the differences between the growth environment of contemporary children and the historical genetic prediction basis, providing a basis for eliminating intergenerational bias and improving the accuracy of genetic prediction.

[0063] Furthermore, the first predicted growth sequence and the second corrected predicted growth sequence are fused to establish the target predicted growth sequence, including: S41. Obtain the first preset weight and the second preset weight; S42. Align the first predicted growth sequence and the second corrected predicted growth sequence according to the monitoring nodes, and calculate the weighted fusion interval for the predicted height interval at each monitoring node based on the alignment matching results. S43. The weighted fusion intervals of each monitoring node are sequentially combined to form the target prediction growth sequence.

[0064] In a preferred embodiment, firstly, a first preset weight and a second preset weight are obtained. Specifically, weight parameters are set to balance the contributions of individual growth prediction and genetic growth prediction to the final fusion result. The first preset weight corresponds to the weight of the first predicted growth sequence, reflecting the importance of the individual's historical growth pattern in the prediction; the second preset weight corresponds to the weight of the second corrected predicted growth sequence, reflecting the importance of genetic factors in the prediction. The settings of these two weight values ​​can be adjusted according to actual application needs, the child's age stage, and prediction accuracy requirements, and are usually determined by child growth and development experts based on clinical experience and statistical analysis results. The sum of the first preset weight and the second preset weight equals 1 to ensure the rationality of the fusion result. For example, for infants under 2 years old, since there is less individual growth history data, the influence of genetic factors is relatively more significant, so the first preset weight can be set to 0.3 and the second preset weight to 0.7; for children over 3 years old, the individual growth pattern is relatively stable, so the first preset weight can be set to 0.6 and the second preset weight to 0.4, to rely more on individual historical growth characteristics for prediction.

[0065] Subsequently, the first predicted growth sequence and the second corrected predicted growth sequence are aligned and matched according to the monitoring nodes, and a weighted fusion interval is calculated for the predicted height intervals at each monitoring node based on the alignment and matching results. Specifically, the first predicted growth sequence and the second corrected predicted growth sequence are aligned according to the same monitoring node time order to ensure that the predicted data at the same time node in the two sequences can accurately correspond. For each monitoring node, the predicted height intervals in the first predicted growth sequence and the second corrected predicted growth sequence are extracted and denoted as the first predicted height interval and the second corrected predicted height interval, respectively. The weighted fusion interval for that monitoring node is calculated, where the lower bound of the weighted fusion interval is the weighted average of the lower bounds of the first predicted height interval and the second corrected predicted height interval, and the upper bound is the weighted average of the upper bounds of the first predicted height interval and the second corrected predicted height interval. The weighted average is calculated using the first preset weight and the second preset weight as weight coefficients, that is, the lower bound weighted average is equal to the lower bound of the first predicted height interval multiplied by the first preset weight plus the lower bound of the second corrected predicted height interval multiplied by the second preset weight, and the upper bound weighted average is calculated in a similar way.

[0066] Subsequently, the weighted fusion intervals of each monitoring node are sequentially combined to form the target predicted growth sequence. The calculated weighted fusion intervals of each monitoring node are then arranged and combined in chronological order to form a complete target predicted growth sequence. This target predicted growth sequence integrates individual historical growth characteristics and genetic prediction information, balancing the contributions of the two prediction sources through weighted fusion, resulting in a more comprehensive and accurate growth prediction result.

[0067] Through the above fusion process, the target predicted growth sequence can effectively integrate multi-source prediction information, maintaining sensitivity to the individual characteristics of the target child while incorporating the influence of genetic background, thus providing a reliable predictive benchmark for subsequent growth retardation risk screening.

[0068] Furthermore, the system monitors and acquires the actual height sequence of the target child, compares and analyzes the actual height sequence with the target predicted growth sequence, and generates a growth retardation warning when the actual height sequence does not meet the target predicted growth sequence, including: S51. Set the number of tracking and monitoring nodes; S52. According to the number of tracking and monitoring nodes, periodically collect the actual height data of the target child at the corresponding monitoring nodes after the prediction start node to form the tracking measured height sequence. S53. Compare and analyze the tracked measured height sequence with the target predicted growth sequence to obtain the measured deviation from the lower limit; S54. Calculate the ratio of the measured deviation from the lower limit to the number of tracking and monitoring nodes to obtain the deviation from the lower limit ratio; S55. When the deviation from the lower limit ratio is greater than or equal to the risk ratio threshold, it is determined that the tracked measured height sequence does not meet the target predicted growth sequence, and a growth retardation warning is generated.

[0069] In a preferred embodiment, firstly, the number of monitoring nodes is set, determining the number of future monitoring nodes requiring continuous monitoring. Specifically, based on the target child's age, growth stage, and clinical monitoring needs, the number of monitoring nodes requiring follow-up monitoring after the predicted starting point is determined. The setting of the number of monitoring nodes needs to balance comprehensive monitoring with practical feasibility, ensuring sufficient monitoring data to support risk assessment while avoiding excessively long monitoring periods that could hinder timely warnings. Setting too few monitoring nodes may lead to insufficient basis for judgment, while setting too many will prolong the risk identification time window, affecting the timeliness of early intervention. Based on setting monitoring nodes monthly, typically 6 or 12 monitoring nodes can be set, corresponding to 6 months and 12 months of follow-up monitoring time, respectively.

[0070] Subsequently, according to the number of monitoring nodes, the actual height data of the target children at the corresponding monitoring nodes after the predicted starting node are collected periodically to form a tracked measured height sequence. Specifically, starting from the predicted starting node, the height of the target children is measured regularly according to the preset monitoring frequency. Each monitoring node needs to accurately record the child's actual height value to ensure the standardization and consistency of the measurement. The actual height data of each monitoring node are arranged in chronological order to form a tracked measured height sequence, reflecting the true growth status of the target children and providing a practical data basis for subsequent comparative analysis.

[0071] Next, the actual height sequence was compared and analyzed with the target predicted growth sequence to obtain the number of deviations from the lower limit. Specifically, the actual height value of each monitoring node in the tracked actual height sequence was compared one by one with the predicted height interval of the corresponding monitoring node in the target predicted growth sequence. When the actual height value was lower than the lower limit of the predicted height interval, it was determined to be a node deviating from the lower limit; when the actual height value was within the predicted height interval, it was determined to be a normal node. The number of all nodes deviating from the lower limit was counted to obtain the number of deviations from the lower limit. This number of deviations from the lower limit reflects the severity of the deviation of the target child's actual growth from the expected lower limit.

[0072] Next, the ratio of the measured deviation from the lower limit to the number of monitoring nodes is calculated to obtain the deviation ratio. Dividing the measured deviation from the lower limit by the number of monitoring nodes yields the deviation ratio, which reflects the frequency with which the target child's growth deviates from the expected lower limit throughout the monitoring period. A higher deviation ratio indicates a greater risk of growth retardation. When the deviation ratio is greater than or equal to a risk ratio threshold, it is determined that the measured height sequence does not meet the target predicted growth sequence, and a growth retardation warning is generated. A risk ratio threshold is set as the critical standard for judging the risk of growth retardation. This risk ratio threshold is usually determined by pediatric growth and development experts based on clinical experience and statistical analysis. When the calculated deviation ratio reaches or exceeds the risk ratio threshold, the target child is determined to be at risk of growth retardation, and an early warning signal is immediately generated.

[0073] Through the aforementioned monitoring and early warning mechanisms, real-time tracking and risk identification of the growth status of target children can be achieved, providing clinicians with objective data analysis and risk assessment basis, assisting doctors in making professional diagnostic judgments and treatment decisions, and thus providing effective technical support for the early detection and prevention of growth retardation problems.

[0074] Furthermore, the measured height sequence is compared and analyzed with the target predicted growth sequence to obtain the measured deviation from the lower limit, including: S531. Compare the measured height values ​​of each monitoring node in the tracked measured height sequence with the predicted height range of the corresponding monitoring node in the target predicted growth sequence one by one. S532. When the measured height is lower than the lower limit of the predicted height range, it is determined to be a deviation from the lower limit node. S533. Count the number of nodes that deviate from the lower limit to obtain the measured number of deviations from the lower limit.

[0075] In a preferred embodiment, firstly, the measured height values ​​of each monitoring node in the tracked measured height sequence are compared one by one with the predicted height intervals of the corresponding monitoring nodes in the target predicted growth sequence. Specifically, according to the time sequence of the monitoring nodes, the measured height values ​​of each monitoring node in the tracked measured height sequence are extracted sequentially, while the predicted height intervals of the corresponding monitoring nodes in the target predicted growth sequence are extracted. This ensures accurate correspondence of time nodes during the comparison process and avoids comparison errors caused by time deviations. The comparison of each monitoring node involves determining the relationship between a specific measured height value and a predicted height interval.

[0076] When the measured height is lower than the lower limit of the predicted height range, it is identified as a deviation from the lower limit node. Specifically, for each monitoring node, the measured height is compared with the lower limit of the predicted height range. If the measured height is less than the lower limit of the predicted height range, it indicates that the child's actual growth at that monitoring node has not reached the expected minimum standard, and the node is marked as a deviation from the lower limit node. When the measured height is within the predicted height range, i.e., the measured height is greater than or equal to the lower limit and less than or equal to the upper limit, it is identified as a normal node, indicating that the child's growth at that time point meets expectations. The number of deviation nodes is counted to obtain the measured deviation from the lower limit number. Specifically, the judgment results of all monitoring nodes are iterated, and the total number of nodes marked as deviations from the lower limit is counted. This measured deviation from the lower limit number directly reflects the number of times the child deviates from the expected lower limit of growth during the entire tracking and monitoring period. The larger the value, the higher the risk of growth retardation. The measured deviation from the lower limit number serves as an important parameter for subsequent calculation of the deviation ratio, providing basic data for quantitatively assessing the growth risk of the child.

[0077] Through the above-mentioned process of comparison and statistical analysis, abnormalities in the growth and development of target children can be accurately identified, providing reliable data support for the assessment of the risk of growth retardation.

[0078] Example 2, as Figure 2 As shown, based on the same inventive concept as the early risk screening method for childhood growth retardation that integrates multi-source data provided in Embodiment 1, this embodiment of the invention also provides an early risk screening system for childhood growth retardation that integrates multi-source data, including: The individual growth prediction module 11 is used to acquire the individual growth history data of the target child before the prediction start point, and to perform individual growth prediction on the target child based on the individual growth history data to obtain a first predicted growth sequence. The genetic growth prediction module 12 is used to acquire the growth history data of the parents of the target child, and perform genetic growth prediction on the target child based on the growth history data of the parents to obtain a second predicted growth sequence. The intergenerational compensation correction module 13 is used to determine the intergenerational compensation sequence of the target child, and to compensate and correct the second predicted growth sequence based on the intergenerational compensation sequence to obtain the second corrected predicted growth sequence. The sequence fusion processing module 14 is used to fuse the first predicted growth sequence and the second corrected predicted growth sequence to establish the target predicted growth sequence; The tracking monitoring and early warning module 15 is used to monitor and acquire the actual height sequence of the target child, compare and analyze the actual height sequence with the target predicted growth sequence, and generate a growth retardation warning when the actual height sequence does not meet the target predicted growth sequence.

[0079] Furthermore, the execution steps of the individual growth prediction module 11 include: Obtain the individual growth history data of the target child before the prediction start point, and perform individual growth prediction on the target child based on the individual growth history data to obtain a first predicted growth sequence, including: Height data of the target child at multiple historical monitoring points before the prediction start point are collected as the individual growth history data of the target child; Based on the individual's historical growth data, a historical node height sequence is generated, and the individual growth predictor is invoked to process the historical node height sequence to obtain the predicted height range of multiple future monitoring nodes, thereby generating the first predicted growth sequence.

[0080] Furthermore, the construction steps of the individual growth predictor include: Collect a set of historical children's growth records, which includes multiple historical children's growth records, and each historical children's growth record includes the height data of the historical children at each monitoring node; Based on the predicted starting node, each of the historical children's growth records is divided into historical node height data and future node height data; A set of sample children's historical node height sequences is constructed based on historical node height data from multiple historical children's growth records, and a set of sample children's future node height sequences is constructed based on future node height data from multiple historical children's growth records. Multiple growth prediction networks are constructed. Based on the historical node height sequence set and the future node height sequence set of the sample children, multiple growth prediction networks are trained until convergence, resulting in multiple growth prediction branches. The multiple growth prediction branches are integrated to obtain the individual growth predictor.

[0081] Furthermore, the execution steps of the genetic growth prediction module 12 include: Height data of the parents of the target child were collected at all monitoring points throughout the cycle to obtain the parents' growth history data; The parents' growth history data are serialized to obtain the father's historical height sequence and the mother's historical height sequence. The paternal and maternal historical height sequences are input into a genetic growth predictor to obtain the full-cycle predicted height interval sequence of the target child. The genetic growth predictor is constructed based on the sample paternal historical height sequence set, the sample maternal historical height sequence set, and the sample child labeled height interval sequence set.

[0082] Based on the predicted starting node, the full-cycle predicted height interval sequence is truncated to obtain the second predicted growth sequence.

[0083] Furthermore, the execution steps of the intergenerational compensation correction module 13 include: Similarity matching retrieval is performed based on the full-cycle predicted height interval sequence of the target child to obtain multiple matching full-cycle predicted height interval sequences, and multiple matching full-cycle predicted height interval sequences correspond to multiple sample identifiers. Based on the multiple sample identifiers, extract multiple matching full-cycle measured height sequences; Based on the multiple matched full-cycle measured height sequences and the multiple matched full-cycle predicted height interval sequences, the intergenerational compensation sequence is obtained. The intergenerational compensation sequence is applied to the second predicted growth sequence to compensate and correct the predicted height interval in the second predicted growth sequence, thereby obtaining the second corrected predicted growth sequence.

[0084] Furthermore, the execution steps of the intergenerational compensation correction module 13 also include: Calculate the median of each predicted height interval in each matched full-cycle predicted height interval sequence to obtain multiple matched full-cycle predicted height median sequences. Calculate the difference between the measured height sequence and the corresponding predicted height median sequence for each matching full-cycle period to obtain multiple deviation sequences; The multiple deviation sequences are statistically summarized according to the monitoring nodes, the mean deviation of multiple deviation values ​​at each monitoring node is calculated, and the mean deviation of each monitoring node is combined in sequence to form the intergenerational compensation sequence.

[0085] Furthermore, the execution steps of the sequence fusion processing module 14 include: Obtain the first preset weight and the second preset weight; The first predicted growth sequence and the second corrected predicted growth sequence are aligned and matched according to the monitoring nodes, and the weighted fusion interval is calculated for the predicted height interval at each monitoring node based on the alignment and matching results. The weighted fusion intervals of each monitoring node are combined sequentially to form the target prediction growth sequence.

[0086] Furthermore, the execution steps of the tracking, monitoring, and early warning module 15 include: Set the number of tracking and monitoring nodes; According to the number of tracking and monitoring nodes, the actual height data of the target child at the corresponding monitoring node after the predicted starting node is collected periodically to form the tracking measured height sequence. By comparing and analyzing the tracked measured height sequence with the target predicted growth sequence, the measured deviation from the lower limit is obtained; Calculate the ratio of the measured deviation from the lower limit to the number of tracking and monitoring nodes to obtain the deviation from the lower limit ratio; When the deviation from the lower limit ratio is greater than or equal to the risk ratio threshold, it is determined that the tracked measured height sequence does not meet the target predicted growth sequence, and a growth retardation warning is generated.

[0087] Furthermore, the execution steps of the tracking, monitoring, and early warning module 15 also include: The measured height values ​​of each monitoring node in the tracked measured height sequence are compared one by one with the predicted height range of the corresponding monitoring node in the target predicted growth sequence. When the measured height is lower than the lower limit of the predicted height range, it is determined to be a deviation from the lower limit node; The number of nodes that deviate from the lower limit is counted to obtain the measured number of deviations from the lower limit.

[0088] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0089] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0090] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.

[0091] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0092] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0093] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0094] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for early risk screening of growth retardation in children that integrates multi-source data, characterized in that, The method includes: Obtain the individual growth history data of the target child before the prediction start point, and make individual growth prediction of the target child based on the individual growth history data to obtain the first predicted growth sequence. Obtain the growth history data of the parents of the target child, and perform genetic growth prediction on the target child based on the growth history data to obtain a second predicted growth sequence; The intergenerational compensation sequence of the target child is determined, and the second predicted growth sequence is compensated and corrected based on the intergenerational compensation sequence to obtain the second corrected predicted growth sequence. The first predicted growth sequence and the second corrected predicted growth sequence are fused together to establish the target predicted growth sequence; The system monitors and acquires the actual height sequence of the target child, compares and analyzes the actual height sequence with the target predicted growth sequence, and generates a growth retardation warning when the actual height sequence does not meet the target predicted growth sequence.

2. The method according to claim 1, characterized in that, Obtain the individual growth history data of the target child before the prediction start point, and perform individual growth prediction on the target child based on the individual growth history data to obtain a first predicted growth sequence, including: Height data of the target child at multiple historical monitoring points before the prediction start point are collected as the individual growth history data of the target child; Based on the individual's historical growth data, a historical node height sequence is generated, and the individual growth predictor is invoked to process the historical node height sequence to obtain the predicted height range of multiple future monitoring nodes, thereby generating the first predicted growth sequence.

3. The method according to claim 2, characterized in that, The steps for constructing the individual growth predictor include: Collect a set of historical children's growth records, which includes multiple historical children's growth records, and each historical children's growth record includes the height data of the historical children at each monitoring node; Based on the predicted starting node, each of the historical children's growth records is divided into historical node height data and future node height data; A set of sample children's historical node height sequences is constructed based on historical node height data from multiple historical children's growth records, and a set of sample children's future node height sequences is constructed based on future node height data from multiple historical children's growth records. Multiple growth prediction networks are constructed. Based on the historical node height sequence set and the future node height sequence set of the sample children, multiple growth prediction networks are trained until convergence, resulting in multiple growth prediction branches. The multiple growth prediction branches are integrated to obtain the individual growth predictor.

4. The method according to claim 1, characterized in that, Obtain the growth history data of the target child's parents, and perform genetic growth prediction on the target child based on the parental growth history data to obtain a second predicted growth sequence, including: Height data of the parents of the target child were collected at all monitoring points throughout the cycle to obtain the parents' growth history data; The parents' growth history data are serialized to obtain the father's historical height sequence and the mother's historical height sequence. The paternal and maternal historical height sequences are input into the genetic growth predictor to obtain the full-cycle predicted height interval sequence of the target child. The genetic growth predictor is constructed based on the sample paternal historical height sequence set, the sample maternal historical height sequence set, and the sample child labeled height interval sequence set. Based on the predicted starting node, the full-cycle predicted height interval sequence is truncated to obtain the second predicted growth sequence.

5. The method according to claim 4, characterized in that, Determine the intergenerational compensation sequence of the target child, and perform compensation correction on the second predicted growth sequence based on the intergenerational compensation sequence to obtain the second corrected predicted growth sequence, including: Similarity matching retrieval is performed based on the full-cycle predicted height interval sequence of the target child to obtain multiple matching full-cycle predicted height interval sequences, and multiple matching full-cycle predicted height interval sequences correspond to multiple sample identifiers. Based on the multiple sample identifiers, extract multiple matching full-cycle measured height sequences; Based on the multiple matched full-cycle measured height sequences and the multiple matched full-cycle predicted height interval sequences, the intergenerational compensation sequence is obtained. The intergenerational compensation sequence is applied to the second predicted growth sequence to compensate and correct the predicted height interval in the second predicted growth sequence, thereby obtaining the second corrected predicted growth sequence.

6. The method according to claim 5, characterized in that, Based on the multiple matched full-cycle measured height sequences and the multiple matched full-cycle predicted height interval sequences, the intergenerational compensation sequence is obtained, including: Calculate the median of each predicted height interval in each matched full-cycle predicted height interval sequence to obtain multiple matched full-cycle predicted height median sequences. Calculate the difference between the measured height sequence and the corresponding predicted height median sequence for each matching full-cycle period to obtain multiple deviation sequences; The multiple deviation sequences are statistically summarized according to the monitoring nodes, the mean deviation of multiple deviation values ​​at each monitoring node is calculated, and the mean deviation of each monitoring node is combined in sequence to form the intergenerational compensation sequence.

7. The method according to claim 1, characterized in that, The first predicted growth sequence and the second corrected predicted growth sequence are fused to establish the target predicted growth sequence, including: Obtain the first preset weight and the second preset weight; The first predicted growth sequence and the second corrected predicted growth sequence are aligned and matched according to the monitoring nodes, and the weighted fusion interval is calculated for the predicted height interval at each monitoring node based on the alignment and matching results. The weighted fusion intervals of each monitoring node are combined sequentially to form the target prediction growth sequence.

8. The method according to claim 1, characterized in that, The system monitors and acquires the actual height sequence of the target child, compares and analyzes the actual height sequence with the target predicted growth sequence, and generates a growth retardation warning when the actual height sequence does not meet the target predicted growth sequence, including: Set the number of tracking and monitoring nodes; According to the number of tracking and monitoring nodes, the actual height data of the target child at the corresponding monitoring node after the predicted starting node is collected periodically to form the tracking measured height sequence. By comparing and analyzing the tracked measured height sequence with the target predicted growth sequence, the measured deviation from the lower limit is obtained; Calculate the ratio of the measured deviation from the lower limit to the number of tracking and monitoring nodes to obtain the deviation from the lower limit ratio; When the deviation from the lower limit ratio is greater than or equal to the risk ratio threshold, it is determined that the tracked measured height sequence does not meet the target predicted growth sequence, and a growth retardation warning is generated.

9. The method according to claim 8, characterized in that, The measured height sequence is compared and analyzed with the target predicted growth sequence to obtain the measured deviation from the lower limit, including: The measured height values ​​of each monitoring node in the tracked measured height sequence are compared one by one with the predicted height range of the corresponding monitoring node in the target predicted growth sequence. When the measured height is lower than the lower limit of the predicted height range, it is determined to be a deviation from the lower limit node; The number of nodes that deviate from the lower limit is counted to obtain the measured number of deviations from the lower limit.

10. A system for early risk screening of childhood growth retardation that integrates multi-source data, characterized in that: The system for implementing the method as described in any one of claims 1 to 9, the system comprising: The individual growth prediction module is used to acquire the individual growth history data of the target child before the prediction start point, and to perform individual growth prediction on the target child based on the individual growth history data to obtain a first predicted growth sequence. The genetic growth prediction module is used to acquire the growth history data of the parents of the target child, and to perform genetic growth prediction on the target child based on the growth history data of the parents to obtain a second predicted growth sequence. The intergenerational compensation correction module is used to determine the intergenerational compensation sequence of the target child, and to compensate and correct the second predicted growth sequence based on the intergenerational compensation sequence to obtain the second corrected predicted growth sequence. The sequence fusion processing module is used to fuse the first predicted growth sequence and the second corrected predicted growth sequence to establish the target predicted growth sequence; The tracking and monitoring early warning module is used to monitor and acquire the actual height sequence of the target child, compare and analyze the actual height sequence with the target predicted growth sequence, and generate a growth retardation early warning when the actual height sequence does not meet the target predicted growth sequence.

Images