A child spine shape intelligent evaluation method and system based on big data analysis

Abstract

The present application relates to the technical field of dynamic evaluation of spinal morphology, and particularly relates to a child spinal morphology intelligent evaluation method and system based on big data analysis. The method comprises the following steps: S1: obtaining static data and dynamic data, dividing the static data into a high-priority evaluation data set and a basic reference data set, and dividing the dynamic data into a key change data set and an auxiliary monitoring data set; taking the complete time period between the time point when the child spinal morphology first receives medical evaluation and the time point when the system confirms that the child spinal function and morphology reach a steady state as an evaluation and diagnosis cycle, and distributing the static data and the dynamic data to the evaluation and diagnosis cycle to obtain a data-stage distribution result. The present application constructs an intelligent evaluation system, accurately tracks child spinal development, dynamically optimizes data collection and stage distribution through error analysis and similarity comparison, and realizes individualized, efficient and self-adaptive closed-loop diagnosis management.

Description

Technical Field

[0001] This invention relates to the field of dynamic assessment technology of spinal morphology, and in particular to an intelligent assessment method and system for children's spinal morphology based on big data analysis. Background Technology

[0002] With the development of medical imaging technology and wearable sensors, significant progress has been made in intelligent assessment methods for pediatric spinal morphology. Existing technologies, by integrating static imaging data and dynamic functional data, have achieved preliminary quantitative analysis of spinal structure and function, providing crucial support for early diagnosis and intervention. These methods demonstrate clear advantages in improving assessment objectivity and reducing human error.

[0003] However, existing assessment systems generally suffer from insufficient adaptability. Traditional methods often employ fixed thresholds and static analysis procedures, failing to dynamically adjust assessment strategies based on individual differences. This leads to shortcomings such as assessment lag, irrational resource allocation, and a high risk of misjudgment when dealing with complex scenarios like children's growth spurts and pathological changes. Especially for obese children or those in rapid growth phases, the error between surface assessment and in-depth physician evaluation is often overlooked.

[0004] Therefore, there is an urgent need to study an intelligent assessment method that can adapt to individual changing needs in order to solve the problem of insufficient flexibility in existing systems and improve the accuracy and efficiency of spinal morphology assessment in children. Summary of the Invention

[0005] To overcome the shortcomings of existing pediatric spinal morphology assessments, such as insufficient dynamic adjustment capabilities leading to low assessment accuracy and efficiency, this invention provides a method and system for intelligent assessment of pediatric spinal morphology based on big data analysis.

[0006] The technical implementation scheme of the present invention is: an intelligent assessment method for children's spinal morphology based on big data analysis, comprising the following steps:

[0007] S1: Acquire static and dynamic data, divide the static data into a high-priority assessment dataset and a basic reference dataset, and divide the dynamic data into a key change dataset and an auxiliary monitoring dataset; take the complete time period from the time point when the child's spinal morphology is first medically assessed to the time point when the system confirms that the child's spinal function and morphology have reached a steady state as the assessment and diagnosis cycle, and allocate the static data and the dynamic data to the assessment and diagnosis cycle to obtain the data-stage allocation result;

[0008] S2: Obtain the comprehensive error calculation result based on the data-stage allocation result; construct a similarity matrix based on the comprehensive error calculation result;

[0009] S3: Analyze the stage error distribution and evaluation process of the current assessment and diagnosis cycle based on the similarity matrix, and adjust the stage sequence of the current assessment and diagnosis cycle.

[0010] Preferably, the acquisition of static and dynamic data, including dividing the static data into a high-priority evaluation dataset and a basic reference dataset, and dividing the dynamic data into a key change dataset and an auxiliary monitoring dataset, includes:

[0011] The static data refers to measurement data that characterizes the morphological and structural attributes of a child's spine at a single point in time;

[0012] The dynamic data refers to the data recording the continuous changes in the functional activity patterns of a child's spine over time;

[0013] If it is static data, then according to the morphological structure attributes and evaluation priority of the static data, the static data is divided into a high-priority evaluation dataset, and the rest of the static data is divided into a basic reference dataset.

[0014] The high-priority evaluation dataset refers to the subset of data that needs to be prioritized for calculation, based on morphological indicators that have core value for spinal pathology diagnosis in static data.

[0015] The basic reference dataset refers to the subset of data used as an analytical reference, which is divided from the static data based on auxiliary morphological indicators used to provide anatomical context and validation benchmarks.

[0016] If it is dynamic data, then based on the functional activity pattern and the significance of change of the dynamic data, the dynamic data is divided into a key change dataset, and the rest of the dynamic data is divided into an auxiliary monitoring dataset.

[0017] The key change dataset refers to a subset of data used for key decision-making, which is divided according to the functional change patterns in dynamic data that exceed a preset clinical significance threshold.

[0018] The auxiliary monitoring dataset refers to a subset of data used for long-term trend observation, which is divided according to background functional indicators that are within the normal fluctuation range in dynamic data.

[0019] Preferably, the assessment and diagnostic cycle is defined as the complete time period between the first medical evaluation of the child's spinal morphology and the point at which the system confirms that the child's spinal function and morphology have reached a steady state. The static data and the dynamic data are then allocated to the assessment and diagnostic cycle to obtain data-stage allocation results, including:

[0020] Based on the collection time points of the core assessment data and clinical events within the assessment and diagnosis cycle, the assessment and diagnosis cycle is divided into N stages;

[0021] The core assessment data includes the high-priority assessment dataset and the key change dataset;

[0022] Within each stage, the static data and dynamic data are allocated to the corresponding stage based on the data's timestamp to obtain the data-stage allocation result.

[0023] Preferably, obtaining the comprehensive error calculation result based on the data-stage allocation result includes:

[0024] Based on the data-stage allocation results, calculations are performed separately for each stage.

[0025] The average error of the high-priority evaluation dataset is used as the first error calculation result;

[0026] The average absolute deviation of the base reference dataset is used as the result of the second error calculation.

[0027] The mean square error between the dynamic parameter evaluation values ​​and the true values ​​of the key change dataset is used as the third error calculation result;

[0028] The reciprocal of the signal-to-noise ratio of the auxiliary monitoring dataset is used as the fourth error calculation result;

[0029] The weighted summation of the four error calculation results yields the comprehensive error calculation result for the aforementioned stage.

[0030] Preferably, constructing the similarity matrix based on the comprehensive error calculation result includes:

[0031] Based on the comprehensive error calculation results of each stage, the periodic error calculation results of each assessment and diagnosis cycle are obtained by weighted summation;

[0032] Based on the recorded diagnosis and treatment activities and timestamp information within the assessment and diagnosis period, a clinical event sequence is constructed according to the event type and the order of occurrence.

[0033] Based on the patient characteristics, data characteristics, and clinical event sequence of the assessment and diagnosis cycle, feature encoding and fusion processing are performed to obtain a cycle feature vector;

[0034] Based on the periodic feature vector, cosine similarity is used to calculate the similarity value between all the assessment and diagnosis periods, and a similarity matrix is ​​constructed based on the similarity value.

[0035] Preferably, the step of analyzing the stage error distribution and evaluation process of the current assessment and diagnosis cycle based on the similarity matrix, and adjusting the stage sequence of the current assessment and diagnosis cycle, includes:

[0036] Based on the matrix vector in the similarity matrix that corresponds to the current assessment and diagnosis cycle, a screening process is performed according to a preset similarity threshold to obtain other assessment and diagnosis cycles with similarity values ​​greater than the preset similarity threshold, and these other assessment and diagnosis cycles are defined as comparative assessment and diagnosis cycles.

[0037] Compare the calculated period error between the current assessment and diagnosis cycle and the compared assessment and diagnosis cycle;

[0038] If the calculated period error of the comparative evaluation and diagnosis cycle is less than the calculated period error of the current evaluation and diagnosis cycle, then the stage error distribution and evaluation process of the current evaluation and diagnosis cycle are analyzed, and the stage sequence of the current evaluation and diagnosis cycle is adjusted.

[0039] Preferably, the step of analyzing the stage error distribution and evaluation process of the current assessment and diagnosis cycle, and adjusting the stage sequence of the current assessment and diagnosis cycle, includes:

[0040] Calculate the proportion of the comprehensive error calculation result of each stage within the current assessment and diagnosis cycle to the total cycle error calculation result, and obtain the stage error contribution set;

[0041] The set of stage error contribution values ​​is sorted in descending order to obtain a sorting result. Based on the sorting result, one or more stages corresponding to the maximum value in the sorting result are selected and defined as key issue stages.

[0042] For the critical issue stage, an error source attribution judgment is performed: if the first error calculation result of the stage is greater than the preset first error threshold corresponding to the stage and data type, it is determined that the evaluation of the high-priority evaluation dataset of the stage is abnormal, and the collection frequency of the high-priority evaluation dataset is increased.

[0043] If the third error calculation result of the stage is greater than the preset third error threshold corresponding to the stage and data type, it is determined that the evaluation of the key change dataset of the stage is abnormal, and the collection cycle of the key change dataset is shortened.

[0044] If the second error calculation result or the fourth error calculation result of the stage is greater than the corresponding preset error threshold, it is determined that the noise level of the baseline or background data of the stage exceeds the allowable range, and a control command is generated to reduce or suspend the collection of the auxiliary monitoring dataset or the basic reference dataset.

[0045] The optimization method for the current assessment and diagnosis cycle is determined based on the increase in the collection frequency of the high-priority assessment dataset, the reduction in the collection cycle of the key change dataset, and the reduction in the collection of the basic reference dataset or auxiliary monitoring dataset.

[0046] Preferably, the method for determining the optimization of the current assessment and diagnosis cycle based on the increase in the collection frequency of the high-priority assessment dataset, the reduction in the collection cycle of the key change dataset, and the reduction in the collection of the basic reference dataset or auxiliary monitoring dataset includes:

[0047] The total time reduction due to the increased acquisition frequency during the remaining phase is calculated as the acquisition cycle shortening amount and defined as the core evaluation data acquisition time compression amount.

[0048] Calculate the total time saved due to the reduction in collection points during the remaining phase and define it as the auxiliary data collection time saving amount;

[0049] The sum of the time reduction in the core evaluation data acquisition and the time saving in the auxiliary data acquisition is taken as the total time reduction of the cycle.

[0050] The total cycle duration compression is preferentially allocated to the stages corresponding to the increased collection frequency of the high-priority evaluation dataset, thus obtaining a priority allocation result.

[0051] Based on the priority allocation results, the phase sequence and duration of the current assessment and diagnosis cycle are replanned.

[0052] Preferably, the step of replanning the phase sequence and duration of the current assessment and diagnosis cycle based on the priority allocation result includes:

[0053] If the duration of any optimized stage is greater than the average duration of the remaining stages, then that stage is split into two or more consecutive sub-stages.

[0054] If the duration of two consecutive stages after optimization is less than the average duration of the remaining stages, then the two stages are merged into a new stage.

[0055] Generate and output a new assessment and diagnostic cycle scheme optimized in terms of the total number of stages, stage order, duration of each stage, and data collection frequency within each stage;

[0056] The new assessment and diagnosis cycle scheme will be applied to the subsequent assessment process of the current child, and the comprehensive error calculation results will be continuously monitored. If the comprehensive error calculation result of this stage or the cycle error calculation result of the entire assessment and diagnosis cycle is lower than the corresponding value before optimization, the optimization will be confirmed to be effective. Otherwise, a new round of similarity matrix analysis and optimization process should be started.

[0057] Preferably, a children's spinal morphology intelligent assessment system based on big data analysis includes:

[0058] The data acquisition and classification module is used to acquire static and dynamic data, and divide high-priority assessment datasets, basic reference datasets, key change datasets, and auxiliary monitoring datasets. It determines the assessment and diagnosis cycle and data-stage allocation scheme based on the data type.

[0059] The error calculation and analysis module is used to calculate the comprehensive error of each stage based on the data-stage allocation results, construct periodic feature vectors and generate a similarity matrix, and output the stage error contribution set and the key issue stage judgment results.

[0060] The evaluation and optimization module is used to perform error source attribution judgment for key issue stages, adjust the collection frequency of high-priority evaluation datasets, the collection cycle of key change datasets and auxiliary monitoring datasets, calculate the total cycle time compression, and reallocate time resources.

[0061] The cycle reconstruction module is used to perform phase splitting and merging operations on the assessment and diagnosis cycle based on the time resource allocation results, generate a new assessment and diagnosis cycle plan and verify the optimization effect, trigger a new round of similarity matrix analysis until the optimization is effective.

[0062] Beneficial Effects: This invention constructs a multi-dimensional assessment and diagnosis cycle through intelligent classification and phased allocation of static and dynamic data, enabling precise temporal management and individualized tracking of children's spinal morphology. It identifies key problem stages based on error contribution and dynamically optimizes data acquisition strategies through error source attribution mechanisms, significantly improving the utilization efficiency and diagnostic reliability of high-value data. By comparing historical assessment cycles with a similarity matrix, it adaptively adjusts the stage sequence and time resource allocation, ensuring the assessment process is highly synchronized with the pace of children's growth and development. Through cycle reconstruction and continuous verification, a closed-loop optimized intelligent assessment system is formed, ultimately improving the accuracy of spinal morphology assessment while enhancing the system's adaptability to individual differences and optimizing clinical diagnostic efficiency and resource scheduling effectiveness. Attached Figure Description

[0063] Figure 1 This is a flowchart of the intelligent assessment method for children's spinal morphology based on big data analysis, as described in this invention.

[0064] Figure 2 This is a structural diagram of the intelligent assessment system for children's spinal morphology based on big data analysis, as described in this invention. Detailed Implementation

[0065] 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.

[0066] Example 1: An intelligent assessment method for children's spinal morphology based on big data analysis, such as... Figure 1 As shown, it includes the following steps:

[0067] S1-1: Obtain static and dynamic data, divide the static data into a high-priority evaluation dataset and a basic reference dataset, and divide the dynamic data into a key change dataset and an auxiliary monitoring dataset.

[0068] The static data refers to measurement data that characterizes the morphological and structural attributes of a child's spine at a single point in time;

[0069] The dynamic data refers to the data recording the continuous changes in the functional activity patterns of a child's spine over time;

[0070] If it is static data, then according to the morphological structure attributes and evaluation priority of the static data, the static data is divided into a high-priority evaluation dataset, and the rest of the static data is divided into a basic reference dataset.

[0071] The high-priority evaluation dataset refers to the subset of data that needs to be prioritized for calculation, based on morphological indicators that have core value for spinal pathology diagnosis in static data.

[0072] The basic reference dataset refers to the subset of data used as an analytical reference, which is divided from the static data based on auxiliary morphological indicators used to provide anatomical context and validation benchmarks.

[0073] If it is dynamic data, then based on the functional activity pattern and the significance of change of the dynamic data, the dynamic data is divided into a key change dataset, and the rest of the dynamic data is divided into an auxiliary monitoring dataset.

[0074] The key change dataset refers to a subset of data used for key decision-making, which is divided according to the functional change patterns in dynamic data that exceed a preset clinical significance threshold.

[0075] The auxiliary monitoring dataset refers to a subset of data used for long-term trend observation, which is divided according to background functional indicators that are within the normal fluctuation range in dynamic data.

[0076] Furthermore, a standardized morphological and functional parameter database based on a large sample of spinal development data from healthy children is pre-built and continuously updated. This database covers the statistical distribution characteristics of key indicators such as normal spinal curvature, range of motion, and electromyographic signal intensity for different age groups, sexes, and developmental stages. By comparing real-time collected static and dynamic data of children with this standardized database, abnormal patterns deviating from the normal range can be preliminarily identified, thus forming a big data-driven early screening and assessment system for pediatric spinal morphology. This provides an initial benchmark and early warning of abnormalities for subsequent optimization of accurate assessment and diagnosis cycles.

[0077] It should be noted that in pediatric spinal morphology assessment, static data is acquired through medical imaging equipment such as X-rays or CT scans. This data captures the morphological and structural attributes of the spine at a specific moment, such as vertebral alignment and skeletal geometry. Dynamic data, on the other hand, is continuously recorded through wearable sensors such as surface electromyography (EMG) or motion capture systems, recording patterns of spinal functional activity, such as changes in muscle electrical signals and range of motion. Existing technologies often use fixed thresholds to analyze this data, which cannot adapt to individual differences and growth changes, leading to assessment delays and the risk of misjudgment.

[0078] To address this issue, this approach divides static data into a high-priority assessment dataset and a baseline reference dataset. The high-priority assessment dataset includes morphological indicators with core value for pathological diagnosis, such as COBB angle measurements; these data are prioritized for calculation to support critical decisions. The baseline reference dataset includes auxiliary morphological indicators that provide anatomical context, such as vertebral body dimensions or bone mineral density measurements; these data serve as analytical references to ensure assessment accuracy. Similarly, dynamic data is divided into a critical change dataset and an auxiliary monitoring dataset. The critical change dataset contains functional change patterns exceeding clinical significance thresholds, such as abnormal muscle activity peaks or rapid postural deviations; these data are used to trigger timely interventions. The auxiliary monitoring dataset includes background functional indicators within normal fluctuation ranges, such as minor changes in daily posture; these data are used for long-term trend observation. Through this division, this step achieves optimized allocation of data resources, enhances the personalization and dynamic adaptability of assessments, effectively overcomes the shortcomings of static analysis in existing technologies, and improves diagnostic accuracy and efficiency.

[0079] The static data partitioning example is as follows: In a single spinal X-ray image analysis, the Cobb angle was measured to be 28°, the vertebral body rotation angle to be 18°, the sagittal diameter of the vertebral body to be 32mm, and the interpedicle distance to be 26mm. Based on their clinical importance, the Cobb angle (28°) and the vertebral body rotation angle (18°), two morphological indicators with core value for scoliosis diagnosis and Cobb angle classification, were assigned to the high-priority evaluation dataset; while the two auxiliary morphological indicators, the sagittal diameter of the vertebral body (32mm) and the interpedicle distance to be 26mm, which provide anatomical context and measurement benchmarks, were assigned to the basic reference dataset.

[0080] The dynamic data partitioning example is as follows: A child's back muscle activity was continuously monitored using surface electromyography (sEMG), and a set of electromyographic signal amplitude values ​​(unit: μV) were collected: [5,8,6,125,7,9,5,110,6,8] (the preset normal fluctuation range is the mean ± 2 times the standard deviation). The calculated signal mean is 10 μV, the standard deviation is 5 μV, and the normal range is approximately [0,20] μV. Therefore, the two abnormal peak signals with amplitudes of 125 μV and 110 μV are classified as the key change dataset; while the remaining data points with amplitudes fluctuating within the 0-20 μV range (such as 5,8,6,7,9,5,6,8 μV) are classified as the auxiliary monitoring dataset.

[0081] S1-2: The complete time period between the first medical assessment of the child's spinal morphology and the time point when the system confirms that the child's spinal function and morphology have reached a steady state is taken as the assessment and diagnosis cycle, and the static data and the dynamic data are allocated to the assessment and diagnosis cycle to obtain the data-stage allocation result;

[0082] Based on the collection time points of the core assessment data and clinical events within the assessment and diagnosis cycle, the assessment and diagnosis cycle is divided into N stages;

[0083] The core assessment data includes the high-priority assessment dataset and the key change dataset;

[0084] Within each stage, the static data and dynamic data are allocated to the corresponding stage based on the data's timestamp to obtain the data-stage allocation result.

[0085] It is important to note that traditional methods for assessing spinal morphology in children often rely on data collection and analysis at isolated time points, lacking dynamic tracking of the entire individual's development. This results in fragmented assessment results that fail to reflect trends. To address this issue, this step clearly defines an assessment and diagnostic cycle that begins with the child's first medical assessment and continues until spinal function and morphology are systematically determined to have reached a stable state. This comprehensive timeframe helps integrate static and dynamic data across different time periods, creating a continuous assessment view.

[0086] Based on the data collection time points and clinical events within the assessment and diagnosis cycle, the cycle is divided into multiple phases. Data collection time points refer to pre-planned standardized measurement times, such as imaging follow-ups every three months or monthly sensor data collection nodes. Clinical events include the occurrence of growth peaks, the implementation of specific treatment measures, or key milestones such as significant changes in symptoms. For example, the cycle can be divided into a diagnostic phase, an intensive monitoring phase, and a stable observation phase, each corresponding to different clinical goals and data requirements.

[0087] Based on this, static and dynamic data are automatically assigned to corresponding stages according to the timestamp information of the data, forming a data-stage allocation result. This allocation mechanism ensures a strict correspondence between data and clinical progress, avoiding the problem of data being out of sync with stages in traditional methods, thereby supporting more accurate time-series analysis and individualized assessment strategy adjustments.

[0088] S2-1: Obtain the comprehensive error calculation result based on the data-stage allocation result;

[0089] Based on the data-stage allocation results, calculations are performed separately for each stage.

[0090] The average error of the high-priority evaluation dataset is used as the first error calculation result;

[0091] The average absolute deviation of the base reference dataset is used as the result of the second error calculation.

[0092] The mean square error between the dynamic parameter evaluation values ​​and the true values ​​of the key change dataset is used as the third error calculation result;

[0093] The reciprocal of the signal-to-noise ratio of the auxiliary monitoring dataset is used as the fourth error calculation result;

[0094] The weighted summation of the four error calculation results yields the comprehensive error calculation result for the aforementioned stage.

[0095] It should be noted that traditional methods for assessing pediatric spinal morphology often suffer from systematic bias due to neglecting the error characteristics and clinical significance of different data types. To overcome this limitation, this step achieves accurate assessment through multi-dimensional error quantification. For high-priority assessment datasets, the average error is used to calculate the first error result, reflecting the overall trend of systematic measurement bias, such as the average difference between Cobb angle measurements and the gold standard. For the baseline reference dataset, the mean absolute deviation is used to calculate the second error result, assessing the dispersion of anatomical reference data, such as the fluctuation range of vertebral rotation angles. For the critical change dataset, the mean squared error is calculated using the mean squared error between dynamic parameter assessment values ​​(such as predicted muscle activity intensity values) and true values ​​(sensor measurements), enhancing sensitivity to abnormal dynamic changes. For the auxiliary monitoring dataset, the reciprocal of the signal-to-noise ratio is used to calculate the fourth error result, quantifying the signal-to-noise level of background functional data, such as the degree of noise interference in daily posture monitoring.

[0096] The above four errors are weighted and summed to form the overall stage error calculation result, and the formula is as follows: ,in, The results of the stage-based comprehensive error calculation characterize the overall error level of all datasets within a single stage. For high-priority evaluation of the first in the dataset The absolute error value of each data point Based on the reference dataset, the first The absolute deviation value of each data point For the key change dataset, the first The squared error between the evaluated value and the true value of each dynamic parameter To assist in monitoring the reciprocal of the signal-to-noise ratio of the dataset, For signal-to-noise ratio, For high-priority data errors, Based on the deviation of the reference data, For dynamic parameter error, To assist in monitoring the signal-to-noise ratio of data, , , , These are weighting coefficients set according to clinical importance. , , , The historical standard deviation (or maximum permissible error) of each error term is used to eliminate the influence of dimensions. This comprehensive error not only characterizes the overall level of data quality and model reliability at the current stage, but also provides a quantitative basis for subsequent optimization, thereby significantly improving the clinical credibility of the evaluation results.

[0097] S2-2: Construct a similarity matrix based on the comprehensive error calculation results;

[0098] Based on the comprehensive error calculation results of each stage, the periodic error calculation results of each assessment and diagnosis cycle are obtained by weighted summation;

[0099] Based on the recorded diagnosis and treatment activities and timestamp information within the assessment and diagnosis period, a clinical event sequence is constructed according to the event type and the order of occurrence.

[0100] Based on the patient characteristics, data characteristics, and clinical event sequence of the assessment and diagnosis cycle, feature encoding and fusion processing are performed to obtain a cycle feature vector;

[0101] Based on the periodic feature vector, cosine similarity is used to calculate the similarity value between all the assessment and diagnosis periods, and a similarity matrix is ​​constructed based on the similarity value.

[0102] It should be noted that in pediatric spinal morphology assessment, traditional methods lack the ability to compare historical assessment cycles, making it difficult to achieve personalized dynamic adjustments. This step establishes a quantitative correlation between the current assessment cycle and historical cycles by constructing a similarity matrix, providing data support for optimizing diagnostic strategies. First, based on the comprehensive error calculation results of each stage, a weighted summation method is used to calculate the cycle error result, using the following formula: ,in, To evaluate the calculation results of the periodic error in the diagnostic cycle, These are the stage weighting coefficients. For the first Phase-based composite error. This result reflects the overall error level throughout the entire evaluation cycle and is used to measure cycle reliability.

[0103] Clinical event sequences are constructed by integrating diagnostic and treatment activities (such as X-ray imaging and orthodontic adjustment) and their timestamps (such as 2023-05-10T09:30:00), and sorted by event type (diagnostic examination, treatment intervention) and occurrence time. For example, the sequence is represented as ["X-ray imaging, orthodontic fitting, rehabilitation training"]. Combining patient characteristics (age, Cobb angle), data characteristics (error distribution), and clinical event sequences, feature encoding is performed through one-hot encoding and numerical normalization. Then, feature splicing and fusion are used to generate a periodic feature vector. For example, the vector [0.75, 1.2, 0.3, 0.9] represents the key features of a certain period. The values ​​of each dimension of this vector correspond to the specific feature encoding results after fusion, for example, represented as: normalized patient Cobb angle deviation value (0.75), average error level of the period (1.2), weighted encoding of major clinical event types (0.3), and time-series influence factor of growth peak (0.9).

[0104] The similarity between periodic feature vectors is calculated using cosine similarity: ,in, periodic eigenvectors and Cosine similarity value between them periodic eigenvectors The modulus (Euclidean norm). periodic eigenvectors The modulus (Euclidean norm). , The features are periodic feature vectors. Cosine similarity effectively measures the directional consistency of high-dimensional feature vectors, ignoring absolute numerical deviations. A similarity matrix is ​​constructed based on the pairwise similarity values ​​of all periods. For example, in a 3×3 matrix, the i-th row and j-th column represents the similarity between periods i and j. This matrix quickly identifies similar historical periods, providing a reference for optimizing the current period, ultimately improving the adaptability and accuracy of the evaluation strategy.

[0105] S3-1: Analyze the stage error distribution and evaluation process of the current assessment and diagnosis cycle based on the similarity matrix, and adjust the stage sequence of the current assessment and diagnosis cycle.

[0106] Based on the matrix vector in the similarity matrix that corresponds to the current assessment and diagnosis cycle, a screening process is performed according to a preset similarity threshold to obtain other assessment and diagnosis cycles with similarity values ​​greater than the preset similarity threshold, and these other assessment and diagnosis cycles are defined as comparative assessment and diagnosis cycles.

[0107] Compare the calculated period error between the current assessment and diagnosis cycle and the compared assessment and diagnosis cycle;

[0108] If the calculated period error of the comparative evaluation and diagnosis cycle is less than the calculated period error of the current evaluation and diagnosis cycle, then the stage error distribution and evaluation process of the current evaluation and diagnosis cycle are analyzed, and the stage sequence of the current evaluation and diagnosis cycle is adjusted.

[0109] It should be noted that traditional methods for assessing spinal morphology in children lack an effective mechanism for referencing historical assessment cycles, making it difficult to dynamically optimize the assessment process based on individual characteristics. This step uses a similarity matrix to achieve accurate comparison between the current cycle and historical cycles. The preset similarity threshold is usually taken as the top 20 percentile of the similarity values ​​of historical cycles, for example, a threshold of 0.85. The row vector corresponding to the current assessment and diagnosis cycle is extracted from the similarity matrix. This vector contains the similarity values ​​between the current cycle and all historical cycles. Taking cycle P5 as an example, the matrix vector corresponding to the current assessment and diagnosis cycle is [0.92, 0.76, 0.88, 0.93], representing the similarity with cycles P1 to P4, respectively.

[0110] Historical periods with a similarity greater than a threshold of 0.85 are selected as comparative assessment diagnostic periods, such as periods P1, P3, and P4. The period error calculation results are compared between the current period and these comparative periods. If the error value of the comparative period is smaller, it indicates that the comparative assessment diagnostic period evaluation process is more reliable. The stage error distribution refers to the proportion of the comprehensive error of each stage within the period. The assessment process includes the arrangement of data collection frequency and stage duration. By analyzing the stage sequence optimization scheme of the comparative periods, the stage sequence of the current period is adjusted. For example, the static data collection frequency of high-error stages is increased from once a month to once a week, while the dynamic data collection duration is extended. This adaptive adjustment mechanism based on similar periods effectively improves the individualization of the assessment process and the diagnostic accuracy.

[0111] S3-2: The analysis of the stage error distribution and assessment process of the current assessment and diagnosis cycle, and the adjustment of the stage sequence of the current assessment and diagnosis cycle, includes:

[0112] Calculate the proportion of the comprehensive error calculation result of each stage within the current assessment and diagnosis cycle to the total cycle error calculation result, and obtain the stage error contribution set;

[0113] The set of stage error contribution values ​​is sorted in descending order to obtain a sorting result. Based on the sorting result, one or more stages corresponding to the maximum value in the sorting result are selected and defined as key issue stages.

[0114] For the critical issue stage, an error source attribution judgment is performed: if the first error calculation result of the stage is greater than the preset first error threshold corresponding to the stage and data type, it is determined that the evaluation of the high-priority evaluation dataset of the stage is abnormal, and the collection frequency of the high-priority evaluation dataset is increased.

[0115] If the third error calculation result of the stage is greater than the preset third error threshold corresponding to the stage and data type, it is determined that the evaluation of the key change dataset of the stage is abnormal, and the collection cycle of the key change dataset is shortened.

[0116] If the second error calculation result or the fourth error calculation result of the stage is greater than the corresponding preset error threshold, it is determined that the noise level of the baseline or background data of the stage exceeds the allowable range, and a control command is generated to reduce or suspend the collection of the auxiliary monitoring dataset or the basic reference dataset.

[0117] It should be noted that in the process of assessing pediatric spinal morphology, traditional methods often lead to unreasonable allocation of assessment resources because they cannot accurately locate the source of error. This step calculates the proportion of the comprehensive error of each stage to the total error of the cycle, forming a set of stage error contribution (for example, if an assessment and diagnosis cycle includes three stages, their comprehensive error calculation results are as follows). =0.8, =0.5, =1.1, the calculated period error for this period. =2.4. The set of stage error contribution percentages is formed by calculating the error proportion of each stage: Stage 1 contribution = 0.8 / 2.4 ≈ 33.3%, Stage 2 contribution = 0.5 / 2.4 ≈ 20.8%, Stage 3 contribution = 1.1 / 2.4 ≈ 45.8%. This set is {33.3%, 20.8%, 45.8%}. After sorting in descending order, Stage 3 is identified as the critical issue stage with the largest error contribution. This set quantitatively characterizes the degree of influence of each stage on the overall error. Sort the set in descending order to prioritize identifying the critical issue stages with the largest error contribution; if sorted in ascending order, major error sources would be ignored. The stages corresponding to the maximum values ​​are chosen because these stages have the most significant impact on the overall error; the stages corresponding to the minimum values ​​have a smaller impact on the error and do not need to be prioritized. The critical issue stages refer to the evaluation periods that have a major negative impact on the overall evaluation quality; the role of the critical issue stages is to serve as the key optimization targets.

[0118] The preset first, second, third, and fourth error thresholds are all set based on the distribution of error calculation results of their corresponding datasets (high-priority assessment dataset, basic reference dataset, key change dataset, and auxiliary monitoring dataset) in historical big data at the same stage (such as the diagnosis stage and the monitoring stage) and under the same data type. The setting standard can be unified as follows: take the statistical quantile in the historical error distribution that can distinguish between 'normal fluctuations' and 'abnormal deviations' as the threshold, for example, the 95th percentile. When the first error calculation result at a certain stage exceeds the corresponding threshold for that stage, it indicates that there is a systematic bias in the measurement of core spinal morphological parameters (such as the Cobb angle), and the data reliability needs to be improved by increasing the acquisition frequency. Similarly, if the third error exceeds the dynamic data threshold, it indicates abnormal monitoring of muscle activity function indicators, and the acquisition cycle needs to be shortened to capture more refined changes. When the second or fourth error exceeds the standard, it indicates that the basic reference data or auxiliary monitoring data has excessive noise, and acquisition should be reduced to reduce resource waste. For example, if the first error in the diagnosis stage reaches 0.8° (threshold 0.5°), the X-ray acquisition frequency should be increased from once a month to twice a week. This precise error attribution mechanism effectively improves the efficiency of assessment resource utilization and diagnostic accuracy.

[0119] S3-3: Determine the optimization method for the current assessment and diagnosis cycle based on the increase in the collection frequency of the high-priority assessment dataset, the reduction in the collection cycle of the key change dataset, and the reduction in the collection of the basic reference dataset or auxiliary monitoring dataset.

[0120] The total time reduction due to the increased acquisition frequency during the remaining phase is calculated as the acquisition cycle shortening amount and defined as the core evaluation data acquisition time compression amount.

[0121] Calculate the total time saved due to the reduction in collection points during the remaining phase and define it as the auxiliary data collection time saving amount;

[0122] The sum of the time reduction in the core evaluation data acquisition and the time saving in the auxiliary data acquisition is taken as the total time reduction of the cycle.

[0123] The total cycle duration compression is preferentially allocated to the stages corresponding to the increased collection frequency of the high-priority evaluation dataset, thus obtaining a priority allocation result.

[0124] It should be noted that the reduction in the acquisition cycle has two sources: firstly, it corresponds to the equivalent reduction in the cycle interval of high-priority evaluation datasets due to the increased acquisition frequency; secondly, it corresponds to the direct reduction in the acquisition cycle of key change datasets. Both contribute to the calculation of the core evaluation data acquisition time compression.

[0125] It should be noted that in the process of assessing spinal morphology in children, traditional methods often lead to insufficient collection of key data or excessive collection of auxiliary data due to the lack of a dynamic resource adjustment mechanism. This step achieves efficient reconstruction of the assessment cycle by accurately calculating the optimization of time resources. The remaining stages refer to the various time segments in the current assessment and diagnosis cycle that have not yet been completed.

[0126] The core assessment data acquisition time compression refers to the additional total time resources required in the remaining phase to ensure that high-priority assessment datasets (such as X-ray images) are acquired at a higher frequency, and critical change datasets (such as electromyography signals) are acquired with a shorter acquisition cycle. This value is a virtual, quantitative indicator used for resource allocation, and its calculation includes the increase in time resource requirements caused by the changes in the two types of core data acquisition strategies mentioned above.

[0127] Example 1 (Key Change Dataset): To monitor aberrant muscle activity more intensively, the surface electromyography (sEMG) monitoring interval was reduced from once daily to once per hour. Over the remaining 30-day period, the number of data collections increased from 30 to 720. Assuming an average collection and processing time of 5 minutes per session, the additional time consumed by this strategy change is (720-30) times * 5 minutes / time = 3450 minutes, or 57.5 hours. This additional time will be included in the core assessment data collection time reduction.

[0128] Example 2 (High-Priority Assessment Dataset): To improve the accuracy of spinal morphology measurements, the frequency of X-ray imaging was increased from once a month to once every two weeks. Over the remaining three-month period, the number of imaging sessions increased from three to six. Assuming an average of two hours per imaging and assessment session, the additional time consumed due to this strategy change is (6-3) times * 2 hours / time = 6 hours. This additional time will also be included in the core assessment data acquisition time compression.

[0129] The time saved in auxiliary data acquisition refers to the total time and resources actually saved in the remaining phase due to the reduction or suspension of acquisition of the basic reference dataset and the auxiliary monitoring dataset.

[0130] Example: To reduce noise data collection, the frequency of surface marker measurements (basal reference data) is reduced from 3 times per week to 1 time per week. Over the remaining 4 weeks (28 days), the number of measurements is reduced from 12 to 4. Assuming each measurement takes 0.5 hours, the time saved by this strategy adjustment is (12-4) measurements * 0.5 hours / measurement = 4 hours.

[0131] The reduction in the acquisition cycle corresponds to a decrease in the acquisition interval of key change datasets, resulting in an increase in the compression of core assessment data acquisition time. The reduction in acquisition corresponds to a decrease in the frequency of auxiliary data acquisition, directly contributing to the savings in auxiliary data acquisition time.

[0132] The sum of the core assessment data acquisition time reduction (representing the additional time resources required) and the auxiliary data acquisition time savings (representing the time resources that can be released) is taken as the total time reduction of the cycle. This value represents the net amount of time resources required to implement the optimization strategy.

[0133] Based on the ranking of the stage error contribution sets, the resources represented by the calculated total cycle time compression are preferentially allocated to stages with higher error contributions. For example, if the diagnostic stage has the highest error contribution, the allocated resources (time) are allocated to this stage to support operations such as increasing the frequency of X-ray acquisition. This precise time resource reallocation mechanism effectively improves the acquisition density and evaluation accuracy of key data while avoiding redundant acquisition of auxiliary data.

[0134] S3-4: Based on the priority allocation results, the phase sequence and duration of the current assessment and diagnosis cycle are replanned.

[0135] If the duration of any optimized stage is greater than the average duration of the remaining stages, then that stage is split into two or more consecutive sub-stages.

[0136] If the duration of two consecutive stages after optimization is less than the average duration of the remaining stages, then the two stages are merged into a new stage.

[0137] Generate and output a new assessment and diagnostic cycle scheme optimized in terms of the total number of stages, stage order, duration of each stage, and data collection frequency within each stage;

[0138] The new assessment and diagnosis cycle scheme will be applied to the subsequent assessment process of the current child, and the comprehensive error calculation results will be continuously monitored. If the comprehensive error calculation result of this stage or the cycle error calculation result of the entire assessment and diagnosis cycle is lower than the corresponding value before optimization, the optimization will be confirmed to be effective. Otherwise, a new round of similarity matrix analysis and optimization process should be started.

[0139] It should be noted that traditional static assessment methods for pediatric spinal morphology lack the ability to dynamically optimize the temporal structure of different stages, resulting in insufficient alignment between data collection and clinical progress. This step achieves precise temporal management by intelligently reconstructing the assessment and diagnostic cycle. When the duration of a certain stage is significantly longer than the average duration of other stages, it indicates that the data collection density or concentration of clinical events in that stage exceeds a reasonable range, and it needs to be broken down into multiple consecutive sub-stages to improve the granularity of the assessment. For example, if the original diagnostic stage lasted 60 days, exceeding the average of 40 days, it is broken down into two sub-stages: preliminary diagnosis (30 days) and detailed diagnosis (30 days). The average duration serves as a baseline to ensure the balance of stage division.

[0140] When the duration of two consecutive phases is below the average, it indicates that there is redundancy in the data collection of these two phases or that they have a high clinical relevance. Merging them improves assessment efficiency. For example, if both the monitoring phase (15 days) and the adjustment phase (20 days) are below the average of 35 days, they are merged into a monitoring-adjustment phase (35 days). The new assessment and diagnosis cycle scheme includes the optimized total number of phases, phase order, duration of each phase, and collection frequency, forming an individualized assessment path. This invention verifies the optimization effect by continuously monitoring the comprehensive error calculation results of each phase and the cycle error calculation results of the entire assessment and diagnosis cycle. The comprehensive error of a phase is used to reflect the detailed changes in data quality within a phase, while the cycle error of the cycle is used to measure the overall reliability of the entire assessment process. If the comprehensive error calculation result of the optimized phase or the cycle error calculation result of the entire assessment and diagnosis cycle is lower than the corresponding value before optimization, the optimization is confirmed to be effective; otherwise, the similarity matrix analysis needs to be restarted to further optimize the timing arrangement. This dynamic adjustment mechanism ensures that the assessment process is always synchronized with the child's spinal development status.

[0141] It should be noted that the determination of effective optimization adopts the logical relationship that "the comprehensive error calculation result of this stage or the periodic error calculation result of the entire cycle is lower than the value before optimization," which is based on the progressive and targeted optimization characteristics of this method. A reduction in stage-specific error indicates that local optimization of key issues has taken effect and is a leading indicator of overall error improvement; while a reduction in periodic error signifies the final achievement of global optimization. Satisfying either one confirms effectiveness, promptly verifying the correctness of the optimization strategy, reducing delays caused by waiting for global results, and ensuring evaluation efficiency and response speed, meeting the real-time requirements of dynamic clinical assessment.

[0142] Example 2: Based on Example 1, a children's spinal morphology intelligent assessment system based on big data analysis, such as... Figure 2 As shown, it includes:

[0143] The data acquisition and classification module is used to acquire static and dynamic data, and divide high-priority assessment datasets, basic reference datasets, key change datasets, and auxiliary monitoring datasets. It determines the assessment and diagnosis cycle and data-stage allocation scheme based on the data type.

[0144] The error calculation and analysis module is used to calculate the comprehensive error of each stage based on the data-stage allocation results, construct periodic feature vectors and generate a similarity matrix, and output the stage error contribution set and the key issue stage judgment results.

[0145] The evaluation and optimization module is used to perform error source attribution judgment for key issue stages, adjust the collection frequency of high-priority evaluation datasets, the collection cycle of key change datasets and auxiliary monitoring datasets, calculate the total cycle time compression, and reallocate time resources.

[0146] The cycle reconstruction module is used to perform phase splitting and merging operations on the assessment and diagnosis cycle based on the time resource allocation results, generate a new assessment and diagnosis cycle plan and verify the optimization effect, trigger a new round of similarity matrix analysis until the optimization is effective.

[0147] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent evaluation of children's spinal column shape based on big data analysis, characterized in that, Comprising the following steps: S1: Obtain static data and dynamic data, divide the static data into a high-priority evaluation data set and a basic reference data set, and divide the dynamic data into a key change data set and an auxiliary monitoring data set; The complete time period between the time point when the child's spine morphology first receives medical evaluation and the time point when the system confirms that the child's spine function and morphology reach a steady state is taken as the evaluation and diagnosis cycle, and the static data and the dynamic data are allocated to the evaluation and diagnosis cycle to obtain a data-stage allocation result; S2: Obtain a comprehensive error calculation result according to the data-stage allocation result, comprising: According to the data-stage allocation result, calculate the average error of the high-priority evaluation data set as a first error calculation result; The average absolute deviation of the basic reference data set is taken as a second error calculation result; The mean square error between the dynamic parameter evaluation value of the key change data set and the true value is taken as a third error calculation result; The reciprocal of the signal-to-noise ratio of the auxiliary monitoring data set is taken as a fourth error calculation result; The four error calculation results are weighted and summed to obtain the comprehensive error calculation result of the stage; According to the comprehensive error calculation result of each stage, the cycle error calculation result of each evaluation and diagnosis cycle is obtained by weighted summation; According to the recorded diagnosis and treatment activities and timestamp information in the evaluation and diagnosis cycle, a clinical event sequence is constructed according to the event type and occurrence time sequence; Based on the patient characteristics, data characteristics and the clinical event sequence of the evaluation and diagnosis cycle, a period feature vector is obtained by feature encoding and fusion processing; According to the period feature vector, the similarity values between all the evaluation and diagnosis cycles are calculated using cosine similarity, and a similarity matrix is constructed according to the similarity values; S3: According to the similarity matrix, analyze the stage error distribution and evaluation process of the current evaluation and diagnosis cycle, comprising: According to the matrix vector corresponding to the current evaluation and diagnosis cycle in the similarity matrix, the other evaluation and diagnosis cycles with a similarity value greater than the preset similarity threshold are obtained by screening processing according to the preset similarity threshold, and the other evaluation and diagnosis cycles are defined as comparison evaluation and diagnosis cycles; Compare the cycle error calculation results of the current evaluation and diagnosis cycle and the comparison evaluation and diagnosis cycle; If the cycle error calculation result of the comparison evaluation and diagnosis cycle is less than the cycle error calculation result of the current evaluation and diagnosis cycle, analyze the stage error distribution and evaluation process of the current evaluation and diagnosis cycle; Calculate the proportion of the comprehensive error calculation result of each stage in the cycle error calculation result in the current evaluation and diagnosis cycle to obtain a stage error contribution degree set; Sort the stage error contribution degree set in descending order to obtain a sorting result, and select one or more stages corresponding to the maximum value in the sorting result according to the sorting result, and define the one or more stages as key problem stages; ​ For the key problem stage, perform error source attribution judgment: if the first error calculation result of the stage is greater than the preset first error threshold corresponding to the stage and data type, it is determined that the high-priority evaluation dataset evaluation of the stage exists an abnormality, and the acquisition frequency of the high-priority evaluation dataset is increased; If the third error calculation result of the stage is greater than the preset third error threshold corresponding to the stage and data type, it is determined that the key change dataset evaluation of the stage exists an abnormality, and the acquisition cycle of the key change dataset is shortened; If the second error calculation result or the fourth error calculation result of the stage is greater than the corresponding preset error threshold, it is determined that the noise level of the reference or background data of the stage exceeds the allowed range, and a control instruction is generated to reduce or suspend the acquisition of the auxiliary monitoring dataset or the basic reference dataset; According to the acquisition frequency increase amount of the high-priority evaluation dataset, the acquisition cycle shortening amount of the key change dataset, and the acquisition reduction amount of the basic reference dataset or auxiliary monitoring dataset, an optimization method for the current evaluation and diagnosis cycle is determined, including: Calculating the total time consumption accumulated due to the acquisition frequency increase in the remaining stages and defining it as the core evaluation data acquisition time compression amount; Calculating the total time consumption saved due to the acquisition point reduction in the remaining stages and defining it as the auxiliary data acquisition time saving amount; The sum of the core evaluation data acquisition time compression amount and the auxiliary data acquisition time saving amount is the total cycle length compression amount; The total cycle length compression amount is preferentially allocated to the stage corresponding to the acquisition frequency increase of the high-priority evaluation dataset to obtain a preferential allocation result; If the duration of any stage after optimization is greater than the average duration of the remaining stages, the stage is split into two or more consecutive sub-stages; If the duration of two consecutive stages after optimization is less than the average duration of the remaining stages, the two stages are merged into a new stage; Apply the new evaluation and diagnosis cycle scheme to the subsequent evaluation process of the current child and continuously monitor the comprehensive error calculation result; if the comprehensive error calculation result of the stage or the cycle error calculation result of the entire evaluation and diagnosis cycle is lower than the corresponding value before optimization, it is confirmed that the optimization is effective; otherwise, a new round of similarity matrix analysis and optimization process should be started.

2. The method for intelligent evaluation of children's spine shape based on big data analysis according to claim 1, characterized in that, The acquisition of static data and dynamic data, the division of the static data into high-priority evaluation dataset and basic reference dataset, and the division of the dynamic data into key change dataset and auxiliary monitoring dataset, include: The static data refers to measurement data representing the morphological structure attributes of the child's spine at a single time point; The dynamic data refers to continuous change data recording the functional activity patterns of the child's spine over time; If the static data, according to the morphological structure attributes and evaluation priority of the static data, the static data is divided into high-priority evaluation dataset, and the remaining static data is divided into basic reference dataset; The high-priority evaluation dataset refers to a subset of data that needs to be calculated in priority, which is divided according to morphological indicators with core value for spinal pathology diagnosis in static data; The basic reference dataset refers to a subset of data for analysis reference, which is divided according to auxiliary morphological indicators for providing anatomical context and verification benchmark in static data; If the dynamic data, the dynamic data is divided into a key change dataset according to the functional activity pattern and change significance of the dynamic data, and the rest of the dynamic data is divided into an auxiliary monitoring dataset; The key change dataset refers to a subset of data for key decision, which is divided according to functional change patterns exceeding a preset clinical significance threshold in dynamic data; The auxiliary monitoring dataset refers to a subset of data for long-term trend observation, which is divided according to background functional indicators within a normal fluctuation range in dynamic data.

3. The method of claim 2, wherein the method is characterized by, The complete time period between the time point when the child's spinal morphology is first medically evaluated and the time point when the system confirms that the child's spinal function and morphology reach a steady state is taken as an evaluation and diagnosis cycle, and the static data and the dynamic data are allocated to the evaluation and diagnosis cycle to obtain data-stage allocation results, including: According to the collection time point of the core evaluation data and the clinical events in the evaluation and diagnosis cycle, the evaluation and diagnosis cycle is divided into N stages; The core evaluation data includes the high-priority evaluation dataset and the key change dataset; In each stage, the static data and the dynamic data corresponding to the data are allocated to the stage according to the time stamp of the data to obtain data-stage allocation results.

4. A child spinal morphology intelligent evaluation system based on big data analysis, for realizing the child spinal morphology intelligent evaluation method based on big data analysis according to any one of claims 1-3.

Images