Research data processing method, system and medium for scoliosis full course

Abstract

The present application provides a scientific research data processing method, system and medium for the whole course of scoliosis, which can be applied to the field of big data processing technology. The present application processes the clinical case images and data into scoliosis whole course time series data based on a time series management method, simultaneously writes the sub-data in the pre-processed scoliosis whole course time series data into the corresponding layer in the multi-level case database, and then converts the natural language screening requirements into database executable conditions based on a large language model. After that, the candidate case set is obtained from the clinical case image and data multi-level case database based on a double-channel screening mechanism, and the target scientific research data output package is obtained by quality control processing of the candidate case set. The present application can reduce the processing workload of researchers in the case screening process, reduce the complexity of the multi-stage artificial screening process, and improve the efficiency, accuracy and traceability of the screening, sorting and analysis of idiopathic scoliosis scientific research data.

Description

Technical Field

[0001] This invention relates to the field of big data processing technology, and in particular to a method, system and medium for processing scientific research data throughout the entire course of scoliosis. Background Technology

[0002] Currently, data related to scoliosis patients is mostly managed through existing hospital clinical information systems, electronic medical record systems, or general clinical databases. However, existing management methods are primarily geared towards clinical diagnosis and treatment processes, typically using routine clinical information such as medical record number, hospitalization number, patient name, diagnosis name, examination date, and hospitalization records as the main search fields. While these methods can meet the needs of daily diagnosis and treatment, medical record retrieval, and basic statistics, they still offer limited support for the in-depth data required for long-term follow-up, refined classification, treatment efficacy evaluation, and research cohort construction for a specific disease (such as scoliosis). Especially in research scenarios, researchers often need to not only retrieve basic patient information and diagnosis names but also perform multi-dimensional screening based on factors such as disease stage, imaging parameters, treatment methods, follow-up time, changes in efficacy, special analysis results, and data completeness. Therefore, when conducting research based on case data obtained using existing case data management methods, researchers need to spend a significant amount of time and effort, and the efficiency and accuracy of pathology data processing are also relatively low. Summary of the Invention

[0003] The main objective of this invention is to provide a method, system, and medium for processing scientific research data throughout the entire course of scoliosis, which can effectively improve the efficiency and accuracy of scientific research data analysis for scoliosis cases.

[0004] To achieve the above objectives, the present invention provides a method for processing research data throughout the entire course of scoliosis, the method comprising the following steps:

[0005] Obtain clinical case images and data of patients with scoliosis;

[0006] Based on the time-series management method, all the images and data of the clinical cases were processed to obtain the time-series data of the entire course of scoliosis.

[0007] A multi-level case database is constructed, which includes a patient identity management layer, a clinical time-series follow-up data layer, an image measurement data layer, a three-dimensional morphological analysis data layer, a biomechanical analysis data layer, a machine learning prediction data layer, and a research adaptation status data layer.

[0008] The time-series data of the entire course of scoliosis were preprocessed;

[0009] Each sub-data point within the preprocessed scoliosis full-course time-series data is written into the corresponding layer of the multi-level case database;

[0010] Obtain the current natural language screening requirements;

[0011] Based on a large language model, the current natural language screening requirements are converted into database executable conditions;

[0012] Based on the database's executable conditions, a candidate case set is obtained by querying the multi-level case database containing clinical case images and data using a dual-channel screening mechanism.

[0013] The candidate case set is subjected to quality control processing to obtain the target scientific research data output package.

[0014] In some embodiments, after acquiring the clinical case images and data of scoliosis patients, the method further includes the following steps:

[0015] The clinical case images and data are structurally classified, and hierarchical identifiers are determined for the clinical case images and data; wherein, the hierarchical identifiers include clinical identity identifiers, research identifiers, and model call identifiers.

[0016] In some embodiments, the processing of all clinical case images and data based on a time-series management approach to obtain time-series data of the entire course of scoliosis includes:

[0017] The clinical process corresponding to the clinical case images and data is divided into multiple time points, and each time point is associated with one type of data in the clinical case images and data.

[0018] A clinical time-series data chain of the clinical case images and data is constructed based on the aforementioned time points;

[0019] The time points in the clinical time series data chain are subjected to time series consistency quality control to obtain the time series data of the entire course of scoliosis.

[0020] In some embodiments, the preprocessing of the time-series data of the entire course of scoliosis includes:

[0021] Field matching is performed on each data point in the time-series data of the entire course of scoliosis.

[0022] The data for which no matching field was found in the time series data of the entire course of scoliosis were manually determined, along with the recognition structure of the large language model.

[0023] Data conflict analysis was performed on the time-series data of the entire course of scoliosis.

[0024] Modify record information based on field matching results, manual determination results, and data conflict analysis results.

[0025] In some embodiments, the step of querying a candidate case set from a multi-level case database containing written clinical case images and data based on a dual-channel screening mechanism, according to the database's executable conditions, includes:

[0026] Based on the database's executable conditions, research data is located in a multi-level case database after clinical case images and data have been written.

[0027] The research data is screened through a preset rule-based screening channel to obtain the first screening result;

[0028] The research data was screened using a large language model screening channel to obtain a second screening result;

[0029] The first screening result and the second screening result are subjected to channel consistency verification to obtain the candidate case set.

[0030] In some embodiments, the quality control processing of the candidate case set to obtain the target scientific research data output package includes:

[0031] Channel and consistency quality control were performed on the candidate case set;

[0032] Calculate the multi-task fit of each data point in the candidate case set based on the quality control results;

[0033] The target scientific research data output package is generated based on the multi-task adaptability and the candidate case set.

[0034] In some embodiments, the multi-task adaptability includes three-dimensional classification adaptability, finite element analysis adaptability, machine learning training set adaptability, and brace efficacy prediction adaptability.

[0035] To achieve the above objectives, another aspect of the present invention provides a research data processing system for the entire course of scoliosis, the system comprising:

[0036] The first acquisition module is used to acquire clinical case images and data of patients with scoliosis.

[0037] The time-series processing module is used to process all the images and data of the clinical cases based on the time-series management method to obtain the time-series data of the entire course of scoliosis.

[0038] The module is used to build a multi-level case database, which includes a patient identity management layer, a clinical time-series follow-up data layer, an image measurement data layer, a three-dimensional morphological analysis data layer, a biomechanical analysis data layer, a machine learning prediction data layer, and a research adaptation status data layer.

[0039] The preprocessing module is used to preprocess the time-series data of the entire course of scoliosis.

[0040] The writing module is used to write each sub-data in the preprocessed scoliosis full-course time series data into the corresponding layer of the multi-level case database.

[0041] The second acquisition module is used to acquire the current natural language screening requirements;

[0042] The conversion module is used to convert the current natural language screening requirements into database executable conditions based on a large language model;

[0043] The filtering module is used to query a set of candidate cases from a multi-level case database containing clinical case images and data based on the database's executable conditions and a dual-channel filtering mechanism.

[0044] The quality control processing module is used to perform quality control processing on the candidate case set to obtain the target scientific research data output package.

[0045] To achieve the above objectives, another aspect of the present invention provides a research data processing system for the entire course of scoliosis, comprising:

[0046] At least one processor;

[0047] At least one memory for storing at least one program;

[0048] When the at least one program is executed by the at least one processor, the at least one processor implements the method.

[0049] To achieve the above objectives, another aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described above.

[0050] The present invention provides the following beneficial effects: After acquiring clinical case images and data of scoliosis patients, all clinical case images and data are processed using a time-series management method to obtain time-series data of the entire course of scoliosis, giving the clinical data temporal characteristics. A multi-level case database is then constructed. Simultaneously, after preprocessing the time-series data of the entire course of scoliosis, each sub-data within the preprocessed time-series data is written into the corresponding layer of the multi-level case database. After obtaining the current natural language screening requirements, the requirements are converted into database executable conditions based on a large language model. Based on these executable conditions, a candidate case set is retrieved from the multi-level case database containing the clinical case images and data using a dual-channel screening mechanism. The candidate case set is then subjected to quality control processing to obtain the target research data output package. This reduces the need for researchers to repeatedly perform initial case screening, image measurement, result verification, and secondary screening in the traditional case screening process, reducing the complexity of the multi-stage manual screening process and improving the efficiency, accuracy, and traceability of idiopathic scoliosis research data screening, organization, and analysis. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0052] Figure 1 This is a flowchart of a research data processing method for the entire course of scoliosis provided in this application;

[0053] Figure 2 This is a flowchart of the method for constructing a time-series data chain for the entire course of scoliosis provided in this application;

[0054] Figure 3 This is a flowchart provided in this application for intelligent case screening, reuse warning, and data output for scientific research tasks. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0056] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are used only for the convenience of describing this application and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, as a fixed connection, a detachable connection, or an integral connection; a mechanical connection or an electrical connection; a direct connection or an indirect connection through an intermediate medium; or a connection within two elements. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0057] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects have an "or" relationship.

[0058] Before describing specific embodiments of this application, the terms used in the embodiments of this application are explained as follows:

[0059] Adolescent idiopathic scoliosis (AIS) refers to scoliosis that occurs during adolescence and whose etiology is still unclear. AIS is a relatively common type of scoliosis and is an important subject for bracing treatment, long-term follow-up, and three-dimensional morphological research.

[0060] Large Language Models (LLMs) include GPT, DeepSeek, and Qwen. These models can understand natural language, extract text information, identify fields, generate structured results, and can be used to assist in case data classification and research screening condition analysis.

[0061] The Large Language Model API (Application Programming Interface) refers to the interface that calls the capabilities of models such as GPT and DeepSeek, enabling the system to automatically complete tasks such as text parsing, field extraction, and natural language screening.

[0062] Natural language screening refers to a process where users can input a single sentence to request screening without having to manually set complex fields.

[0063] Three-dimensional classification refers to a method for classifying scoliosis based on three-dimensional morphological data of the spine. Its data foundation may include the vertebral body center point, three-dimensional centerline, curve projection, apical vertebra position, curve span, and spatial rotation characteristics.

[0064] Finite Element Analysis (FEA) is a numerical simulation method that discretizes structures such as the spine, intervertebral discs, ligaments, or braces into mesh elements and calculates their stress, strain, and deformation responses under specific boundary conditions and loads.

[0065] In related technologies, with the continuous advancement of hospital informatization, clinical diagnosis and treatment data has gradually shifted from paper-based medical record management to electronic, structured, and platform-based management. Currently, most medical institutions have established relatively complete hospital information systems, electronic medical record systems, image archiving and communication systems, laboratory testing systems, and follow-up management systems. These systems can store and retrieve patient identity information, outpatient records, inpatient records, diagnoses, examination and test results, surgical records, discharge summaries, and follow-up records, providing important support for clinical diagnosis and treatment, medical record archiving, medical quality control, and basic statistics.

[0066] Existing hospital databases are typically designed around clinical workflows, with the primary goal of ensuring continuity of medical activities and traceability of medical data. Clinicians can quickly retrieve patients' past medical records, imaging data, and treatment processes using fields such as patient name, medical record number, outpatient number, inpatient number, examination date, diagnosis name, and department information. For routine medical management, this type of database can already meet the needs of querying basic patient information, tracing examination results, accessing medical records, and conducting clinical work statistics.

[0067] However, existing hospital databases are typically designed with a focus on "clinical record management" rather than "disease-specific research data organization." Their field systems are largely centered around treatment processes and medical record writing requirements, recording basic information such as whether a patient sought medical attention, underwent examinations, was hospitalized, received treatment, and received a diagnosis. However, they remain limited in providing the in-depth data support needed for long-term follow-up, refined subtyping, treatment efficacy evaluation, and the construction of research cohorts for a specific disease. Especially in research settings, researchers often need to retrieve not only basic patient information and diagnosis names but also perform multi-dimensional filtering based on disease stage, imaging parameters, treatment methods, follow-up time, changes in efficacy, special analysis results, and data completeness.

[0068] Taking scoliosis as an example, this disease is not a common condition that can be managed with a single diagnosis and treatment; rather, it is a type of structural spinal deformity characterized by long-term follow-up. Scoliosis patients typically undergo a lengthy clinical management process, from screening, initial diagnosis, imaging assessment, risk stratification, treatment selection, brace fitting, rehabilitation training, to long-term follow-up. Clinical decisions not only rely on the patient's age, gender, chief complaint, and diagnosis, but also require a comprehensive assessment based on multidimensional information such as the Cobb angle, apical vertebra position, end vertebra position, curve type, bone maturity, coronal balance, sagittal parameters, body surface morphology, previous rate of progression, brace correction rate, brace compliance, and changes in the Cobb angle during follow-up.

[0069] Meanwhile, scoliosis is a disease that highly relies on imaging assessment and morphological analysis. Traditional clinical management typically relies on standing full-spine X-rays or EOS images to assess Cobb angle, sagittal parameters, and trunk balance. With the development of 3D image reconstruction, statistical shape modeling, finite element analysis, and machine learning technologies, research data on scoliosis has been further expanded to include 3D reconstruction models, vertebral body center point coordinates, 3D spinal centerlines, 3D curve classification, local vertebral body morphological parameters, finite element models, material parameters, boundary conditions, stress distribution results, brace action parameters, machine learning input features, and prediction results. This data is often generated by different software, different researchers, or different analysis processes, and is stored in various formats such as tables, image files, 3D model files, finite element result files, and machine learning datasets.

[0070] Therefore, scoliosis data management has distinct characteristics compared to general disease databases. First, its data has long-term follow-up attributes, requiring continuous management across different time points such as initial diagnosis, treatment, bracing, follow-up, and outcome. Second, its data has multimodal attributes, including clinical medical records, as well as X-ray, EOS, CT, MRI, body surface morphology, 3D models, finite element analysis results, and machine learning data. Third, its data has strong research attributes; the same case may be used for different research tasks such as clinical follow-up, 3D classification, finite element analysis, machine learning training set construction, bracing efficacy prediction, and long-term outcome analysis. Fourth, its data has strong correlation; whether a case can be used for a particular study depends not only on whether the basic diagnosis meets the requirements but also on factors such as the completeness of imaging, the continuity of follow-up, the existence of a 3D model, the completeness of the finite element file, and the availability of machine learning labels.

[0071] Under the current working model, scoliosis research groups often need to first screen relevant cases from hospital databases, then manually review imaging data, export measurement results, verify follow-up times, locate 3D reconstruction files, confirm the status of finite element models, and organize machine learning data tables. Each time a research cohort is screened, researchers need to repeatedly switch between medical record systems, imaging systems, local Excel spreadsheets, 3D reconstruction software, finite element analysis folders, and statistical analysis data tables. This process is not only time-consuming but also prone to problems such as missed cases, data mismatches, confusion of imaging time points, discrepancies between 3D models and source images, missing finite element results, and incomplete follow-up outcomes.

[0072] Furthermore, existing hospital databases typically support structured searches based on fixed fields, but they are not adept at handling complex screening requests from researchers using natural language. For example, researchers might want to directly input "screen for moderate adolescent idiopathic scoliosis patients suitable for brace efficacy prediction, requiring baseline imaging, brace-in-brace imaging, follow-up of more than 6 months, and brace wearing records." Ordinary databases usually cannot understand such natural language requests, requiring researchers to break them down into multiple fields such as age, diagnosis, Cobb angle, imaging type, treatment method, follow-up time, and brace records, and then filter them item by item.

[0073] Currently, data related to scoliosis patients is mostly managed through existing hospital clinical information systems, electronic medical record systems, or general clinical databases. These systems are primarily geared towards clinical diagnosis and treatment processes, typically using routine clinical information such as medical record number, hospital number, patient name, diagnosis, examination date, and hospitalization record as primary search fields, which can meet the needs of daily diagnosis and treatment, medical record retrieval, and basic statistics. However, for scoliosis-related research, existing systems still have the following shortcomings:

[0074] First, existing clinical databases are mostly centered on clinical diagnosis and treatment information, lacking data structure designs geared towards scientific research analysis. Scoliosis research requires not only basic patient information and diagnostic information, but also multi-dimensional information such as Cobb angle, apical vertebra position, bone maturity, brace treatment information, follow-up time points, imaging measurement parameters, 3D reconstruction results, 3D classification results, finite element analysis status, and machine learning prediction data. Existing systems typically cannot uniformly manage the aforementioned clinical data, imaging data, 3D morphological data, biomechanical analysis data, and research tags, forcing researchers to repeatedly search and manually match data among medical record systems, imaging systems, measurement tables, 3D reconstruction folders, and finite element analysis results.

[0075] Second, the screening methods of existing systems are relatively fixed, usually only supporting basic fields such as medical record number, hospitalization number, diagnosis name, examination time, and department information. This makes it difficult to support multi-factor combination screening under complex research conditions. For example, when researchers need to screen cases that are "10–18 years old, Cobb angle 25°–45°, have intra-brace imaging, have been followed up for more than 6 months, and have 3D reconstruction data that can be used for finite element analysis," existing systems often cannot complete the screening directly. They need to manually check each item, which is time-consuming and prone to omissions or misscreening.

[0076] Third, existing clinical databases often lack data suitability assessments for specific research tasks. For the same scoliosis case, it may be suitable for clinical follow-up analysis but not necessarily for 3D subtyping studies; it may have complete imaging data but not the 3D model, material parameters, or boundary conditions required for finite element modeling; it may have partial follow-up information but not meet the requirements for input features and outcome labels for machine learning training set construction. Existing systems typically only record the existence of data and cannot automatically determine whether a case is suitable for different research tasks such as 3D subtyping, finite element analysis, machine learning training, efficacy prediction, or long-term follow-up outcome analysis. Furthermore, due to the lack of case usage records specific to research projects, research tasks, and data versions, the same case may be repeatedly retrieved, included, and processed by different researchers, different sub-projects, or the same team at different stages, leading to repeated measurements of imaging indicators, repeated reconstruction of 3D models, repeated construction of finite element models, and repeated processing of follow-up data. For multiple research projects with sample overlap, there may also be duplicate counting, overlap between training and validation sets, and the same patient being mistakenly identified as an independent sample at different time points, affecting the efficiency of research data processing, sample independence, and the reliability of research results.

[0077] Fourth, existing systems lack automated quality control for the consistency and integrity of data from multiple sources. Scoliosis patient data comes from complex sources, including clinical records, imaging examinations, surface assessments, brace treatment records, rehabilitation training records, follow-up results, 3D reconstruction files, finite element model files, and machine learning data tables. These data contain correspondences between patient IDs, examination times, treatment times, follow-up time points, image sources, model file paths, and analysis results. Existing systems typically cannot automatically check for consistency between these data, easily leading to problems such as mismatches between imaging time points and follow-up records, discrepancies between 3D models and source images, missing finite element result files, and incomplete machine learning labels.

[0078] Fifth, existing clinical information systems typically rely on known patients, existing diagnoses, and pre-entered structured fields for queries, lacking a large language model interface and natural language parsing capabilities for research case discovery. In actual research screening, researchers usually need prior knowledge of potential cases, or initially obtain a preliminary patient list based on limited criteria such as diagnosis name, patient number, and department. They then review medical records, medical images, and follow-up data for each case, manually performing image measurements, disease progression assessments, and inclusion criteria verification before conducting a final screening. For research indicators not pre-recorded in routine clinical practice, such as changes in Cobb angle at specific segments, brace correction rates, three-dimensional morphological features, follow-up time windows, or outcome labels, existing systems typically cannot directly retrieve these parameters, requiring researchers to remeasure, organize, and enter them. Therefore, traditional systems are essentially more suitable for retrieving information on known cases and are less effective at proactively identifying potential cases meeting complex research criteria from large amounts of clinical data.

[0079] Meanwhile, researchers also need to manually break down research requirements and convert them into database fields and logical search conditions. For complex needs such as "screening patients with moderate idiopathic scoliosis who are suitable for brace efficacy prediction and have complete follow-up data," existing systems struggle to simultaneously understand multiple dimensions such as disease severity, treatment methods, data completeness, follow-up time, and research suitability. This often requires multiple steps, including initial case screening, data retrieval, imaging measurements, manual judgment, and secondary screening. This process not only increases the workload of researchers and reduces case screening efficiency but also easily leads to inconsistent screening results among different researchers due to differences in the definition of indicators, measurement methods, and understanding of inclusion criteria.

[0080] Sixth, existing systems primarily serve clinical diagnosis and treatment record management, lacking a hierarchical management mechanism among clinical identity, research identifiers, and model call identifiers. Clinical follow-up requires retaining patients' real identity information for follow-up visits and treatment management, while research analysis and model call processes should minimize direct exposure of identity information. Existing systems often struggle to simultaneously meet the three requirements of clinical traceability, research analyzability, and secure isolation during model calls.

[0081] Furthermore, current research data applications typically rely on hospital office automation approval processes. Researchers submit project materials, ethical approvals, and applications for data usage scope and permissions sequentially, which are then manually reviewed and processed by multiple management departments, resulting in lengthy approval cycles. Even after a research project has completed ethical review and has clearly defined data usage permissions, researchers may still need to repeatedly submit information such as the scope of cases, data fields, and purpose of use, impacting the efficiency of research data acquisition. Existing systems lack an automated verification mechanism that links ethical approval results, research project identification, data usage permissions, and personnel identity, making it difficult to establish a standardized, auditable, and rapid authorization channel for research tasks that meet predetermined conditions. Therefore, while ensuring ethical compliance, access control, and full process traceability, reducing redundant approvals and manual processing steps, and providing an efficient data access and retrieval mechanism for approved research projects, has become a problem that needs to be addressed in clinical research data management.

[0082] In view of this, the embodiments of this application provide a method, system and medium for processing scientific research data for the entire course of scoliosis, which can effectively reduce the workload of researchers in screening data and improve the efficiency and accuracy of scientific research data analysis of scoliosis cases.

[0083] Reference Figure 1 This application provides a method for processing research data throughout the entire course of scoliosis, which includes the following steps:

[0084] Step S101: Obtain clinical case images and data of patients with scoliosis;

[0085] Step S102: Process all clinical case images and data based on time-series management to obtain time-series data of the entire course of scoliosis.

[0086] Step S103: Construct a multi-level case database, which includes a patient identity management layer, a clinical time-series follow-up data layer, an image measurement data layer, a three-dimensional morphological analysis data layer, a biomechanical analysis data layer, a machine learning prediction data layer, and a research adaptation status data layer.

[0087] Step S104: Preprocess the time-series data of the entire course of scoliosis;

[0088] Step S105: Write each sub-data in the preprocessed scoliosis full-course time series data into the corresponding layer in the multi-level case database;

[0089] Step S106: Obtain the current natural language screening requirements;

[0090] Step S107: Based on the large language model, convert the current natural language screening requirements into database executable conditions;

[0091] Step S108: Based on the database's executable conditions, a candidate case set is obtained by querying the multi-level case database containing clinical case images and data using a dual-channel filtering mechanism.

[0092] Step S109: Perform quality control processing on the candidate case set to obtain the target scientific research data output package.

[0093] It is understood that, in this embodiment, after obtaining the clinical case images and data of scoliosis patients, the clinical case images and data are structurally classified and their hierarchical labels are determined. Specifically, the clinical case images and data in this embodiment include structured data, semi-structured data, and unstructured data. Structured data includes field data from hospital databases, research forms, or standardized entry forms. Examples include the following field data: patient number; name; outpatient number; inpatient number; gender; age; diagnosis type; Cobb angle; Risser sign; Sanders stage; image type; treatment method; brace wearing time; follow-up time; 3D reconstruction status; finite element model status; machine learning label status. Semi-structured data includes image reports, brace record forms, follow-up forms, rehabilitation training records, research measurement forms, etc. For example:

[0094] Imaging report: Cobb angle of the main thoracic curve 38°, apical vertebra T8, Risser 1;

[0095] Braces record: Worn for 18 hours daily, Cobb angle inside the brace 22°;

[0096] Follow-up record: At the 6-month follow-up examination, the Cobb angle changed from 38° to 35°;

[0097] Finite element record: The T1-L5 model has been completed, and material parameter files and stress result files exist.

[0098] Unstructured data includes doctors' written notes, researchers' comments, case discussion records, and natural language processing (NLP) screening requirements. For example:

[0099] This patient has moderate AIS, good brace treatment compliance, baseline imaging, and a 6-month follow-up, making them suitable for predictive analysis of brace efficacy.

[0100] To balance real-world clinical use, research analysis, and model call security, this embodiment employs a tiered identifier system including clinical identity identifiers, research identifiers, and model call identifiers. Clinical identity identifiers are used for clinical diagnosis, follow-up appointments, long-term follow-up, and treatment management, and include information such as name, outpatient number, inpatient number, and contact information. This information is used by clinicians to trace the identity of actual patients. Research identifiers are used for research analysis, case cohort construction, statistical analysis, and data export. Examples include: SC_001; AIS_001; SCO_2026_001. A mapping relationship is established between research identifiers and clinical identity identifiers, but research identifiers are prioritized for research export. Model call identifiers are used for calling large language model APIs, natural language screening, and machine learning prediction processes. Model call identifiers only associate with structured fields required for model processing and do not directly access direct identity information such as name, ID number, phone number, or address.

[0101] This embodiment, through the aforementioned hierarchical identification management, enables traceability at the clinical end, analysis at the research end, and isolation at the model end.

[0102] It is understandable that scoliosis is a disease requiring long-term follow-up and dynamic management. Patient data is not data at a single point in time, but rather spans multiple stages, including screening, initial diagnosis, imaging examinations, treatment intervention, bracing, rehabilitation training, follow-up examinations, and outcome evaluation. Therefore, this embodiment establishes a timeline-based clinical disease progression data chain to record the patient's clinical status and treatment changes at different stages. Specifically, such as... Figure 2 As shown, this embodiment, after obtaining clinical case images and data, processes all clinical case images and data based on a time-series management method to obtain time-series data of the entire course of scoliosis. The processing procedure of this embodiment includes the following steps:

[0103] The clinical workflow corresponding to clinical case images and data is divided into multiple time points, and each time point is associated with one type of data in the clinical case images and data.

[0104] Constructing a clinical time-series data chain of clinical case images and data based on time points;

[0105] Temporal consistency quality control was performed on the time points in the clinical time series data chain to obtain the time series data of the entire course of scoliosis.

[0106] Specifically, multiple time points are included: T0: screening time point; T1: initial diagnosis time point; T2: baseline imaging time point; T3: treatment plan formulation time point; T4: brace fabrication or brace wearing start time point; T5: in-brace imaging time point; T6: rehabilitation training start time point; T7: first follow-up time point; T8: subsequent follow-up time point; T9: outcome evaluation time point; T10: surgical referral or treatment pathway adjustment time point. Each time point can be associated with corresponding clinical information, imaging information, treatment information, brace information, rehabilitation information, follow-up information, and research analysis information.

[0107] This embodiment constructs a clinical time-series data chain of clinical case images and data based on time points to manage the patient's entire disease course information in chronological order. This clinical time-series data chain includes screening records; initial diagnosis records; baseline imaging records; treatment decision records; brace treatment records; in-brace imaging records; rehabilitation training records; follow-up records; outcome evaluation records; and treatment pathway adjustment records. For example, a patient's timeline can be represented as follows:

[0108] 2026-01-10: Positive result detected during school screening;

[0109] 2026-01-20: Initial outpatient visit;

[0110] 2026-01-21: Standing full spine X-ray, Cobb angle 32°;

[0111] 2026-02-05: Bracing treatment begins;

[0112] 2026-02-20: Image inside the brace, Cobb angle 20°;

[0113] 2026-05-20: 3-month follow-up, Cobb angle 30°;

[0114] 2026-08-20: 6-month follow-up, Cobb angle 28°.

[0115] This embodiment uses the timeline to determine whether the data is continuous, whether the treatment is performed on time, whether the follow-up is missing, and whether the treatment effect has changed.

[0116] This embodiment also performs time-series consistency quality control on time points in the clinical time-series data chain. This time-series consistency quality control includes: whether the screening time is earlier than the initial diagnosis time; whether the baseline image is earlier than the treatment start time; whether the in-brace image is later than the brace wearing start time; whether the follow-up image is later than the baseline image; whether the rehabilitation training record corresponds to the treatment stage; whether the outcome evaluation is later than the follow-up time point; and whether the treatment adjustment record has prior follow-up evidence. For example, when "the in-brace image time is earlier than the brace fabrication time," the case is marked as a time-series abnormality, and the user is prompted to manually verify it.

[0117] After completing time-series consistency quality control, the entire time-series data of scoliosis is output. This data includes: the patient's entire disease timeline; current treatment stage; date of the most recent follow-up; recommended date of the next follow-up; Cobb angle change trend; brace compliance changes; treatment response status; progression risk indicators; and treatment pathway assistance. This output is used to help clinicians quickly understand the patient's complete treatment process.

[0118] It is understood that the multi-level case database in this embodiment includes a patient identity management layer, a clinical time-series follow-up data layer, an image measurement data layer, a three-dimensional morphological analysis data layer, a biomechanical analysis data layer, a machine learning prediction data layer, and a research adaptation status data layer. Specifically, the patient identity management layer is used to manage clinical identity identifiers, research identifiers, model call identifiers, and permission information.

[0119] The clinical time-series follow-up data layer is used to record patient screening, initial diagnosis, treatment, bracing, rehabilitation, follow-up examination and outcome information in chronological order.

[0120] The image measurement data layer is used to record X-ray, EOS, CT, MRI, and brace images and corresponding measurement parameters at different time points, including: Cobb angle; principal curve region; apical vertebra; end vertebra; coronal plane balance; thoracic kyphosis; lumbar lordosis; pelvic parameters; axial rotation parameters; image date; image type; and image stage.

[0121] The 3D morphological analysis data layer is used to manage data such as 3D reconstruction, vertebral body center point, 3D spinal centerline, and 3D classification, including: 3D reconstruction status; 3D model file path; vertebral body center point coordinates; 3D spinal centerline; coronal projection curve; sagittal projection curve; transverse projection curve; long or short curve type; main curve region; 3D classification results; cluster classification results; cluster center distance; maximum membership degree; PCA principal component score; and SSM result file path.

[0122] The biomechanical analysis data layer is used to store the following file paths: geometric model file path; material parameter file path; mesh file path; boundary condition file path; brace action parameters; vertebral body stress results; intervertebral disc stress results; endplate stress results; brace internal deformation response; finite element result file path; and finite element result export status.

[0123] The machine learning prediction data layer is used to manage machine learning datasets and prediction results, including the input feature table path; training label status; training set status; validation set status; test set status; model version; progression risk prediction results; treatment response prediction results; brace efficacy prediction results; surgical referral risk prediction results; and prediction result file path.

[0124] The research fit status data layer is used to record whether a case is suitable for different research tasks, including: 3D classification fit status; finite element analysis fit status; machine learning training set fit status; brace efficacy prediction fit status; long-term follow-up outcome analysis fit status; research data export status; and research cohort version number.

[0125] Simultaneously, this embodiment preprocesses the time-series data of the entire scoliosis course. This preprocessing includes: field matching for each data point in the time-series data of the entire scoliosis course; manual determination of data points in the time-series data of the entire scoliosis course that do not match fields and the large language model recognition structure; data conflict analysis of the time-series data of the entire scoliosis course; and modification of record information based on the field matching results, manual determination results, and data conflict analysis results.

[0126] Specifically, this embodiment addresses the issue that real-world clinical research databases cannot rely entirely on automatic import or large language model recognition. Due to the complexity of data sources, imported data may contain inconsistencies in field names, units, time points, case numbers, data gaps, or model recognition errors. Therefore, in addition to automatic import and large language model recognition, manual editing, correction, confirmation, and version tracking functions are implemented.

[0127] The data import methods in this embodiment include: manual entry; importing Excel spreadsheets; importing CSV files; importing image report text; importing follow-up records; importing brace wearing records; importing 3D reconstruction result file indexes; importing finite element models and result file indexes; importing machine learning training data tables; importing unstructured text after recognition via the Large Language Model API; and importing via the hospital information system or PACS interface.

[0128] This embodiment automatically matches data based on a field dictionary after importing the entire time-series data of scoliosis. For example:

[0129] “Main bend Cobb” → Cobb_main;

[0130] "In-brace Cobb angle" → In_brace_Cobb;

[0131] "Wearing time" → Brace_wear_time;

[0132] "Follow-up months" → Followup_months;

[0133] “3D reconstruction complete” → Reconstruction_status;

[0134] “Finite Element Results” → FEA_result_status;

[0135] "Machine Learning Labels" → ML_label_status.

[0136] In this embodiment, data that does not match a field will proceed to the manual confirmation process.

[0137] Furthermore, this embodiment also allows users to manually modify tables and case data after import, including changing field names; adding fields; deleting invalid fields; merging duplicate fields; splitting composite fields; setting field units; setting field data types; modifying case field values; supplementing missing data; modifying image time points; modifying finite element file paths; modifying machine learning labels; and modifying research adaptation status. For example:

[0138] The imported fields are:

[0139] Brace wearing status: 18 hours per day, for 6 months;

[0140] Users can be divided into:

[0141] Brace_wear_time_per_day = 18;

[0142] Brace_duration_months = 6.

[0143] In addition, this embodiment will also manually confirm the results obtained from the large language model API recognition, including: automatic recognition not confirmed; manual confirmation passed; confirmation after manual modification; recognition conflict; unable to recognize; data to be supplemented. For example, the large language model recognizes: Cobb_main = 38°; Apex = T8; Risser = 1. The user can choose to confirm, modify, mark as error, or return for re-recognition.

[0144] Finally, this embodiment also logs all automatic identification, manual modification, batch update, and research export operations. The log content includes: case number; field name; data before modification; data after modification; source of modification; modifier; modification time; reason for modification; whether identified by LLM; whether manually confirmed; model version; source of imported files; research cohort version number.

[0145] Once the research data has been confirmed, users can mark it as "data locked" and generate a research export version, for example:

[0146] Scoliosis_3DClassification_v1.0;

[0147] FEA_Candidate_Cases_v1.2;

[0148] BracePrediction_Cohort_v2.0;

[0149] ML_Training_Dataset_v1.1.

[0150] Understandably, after obtaining the preprocessed time-series data of the entire scoliosis course, this embodiment writes each sub-data within the preprocessed time-series data into the corresponding layer of a multi-level case database. This allows for autonomous classification and structured archiving via a large language model API. Specifically, this embodiment uses GPT, DeepSeek, or other large language model APIs to autonomously classify, identify fields, and structure the input data, thereby reducing manual processing workload.

[0151] This embodiment first receives preprocessed time-series data of the entire course of scoliosis input by the user, including medical record summary, imaging report, follow-up record, brace record, rehabilitation record, research notes or table content.

[0152] Subsequently, the input content is preprocessed, including: identifying the date; identifying the numerical value; identifying the unit; identifying the image type; identifying the medical terminology; identifying the patient number; identifying the follow-up stage; and identifying possible field names.

[0153] Then, the large language model API was called to classify the input content into: basic information; diagnostic information; imaging measurement; bone maturity; treatment intervention; brace compliance; rehabilitation training; follow-up outcome; three-dimensional morphological analysis; finite element analysis; machine learning prediction; and scientific research notes.

[0154] For example, the input text is: Patient is a 13-year-old female with a Cobb angle of 38° in the main thoracic curve, a Risser 1, and has been wearing a brace for 6 months. The Cobb angle inside the brace has decreased to 22°. She wears the brace for about 18 hours a day.

[0155] Output structured results:

[0156] {

[0157] "basic_information": {

[0158] "sex": "female",

[0159] "age": 13

[0160] },

[0161] "imaging_measurement": {

[0162] "main_curve": "thoracic",

[0163] "cobb_main": 38

[0164] },

[0165] "skeletal_maturity": {

[0166] "risser": 1

[0167] },

[0168] "brace_treatment": {

[0169] "brace_duration_months": 6,

[0170] "in_brace_cobb": 22,

[0171] "daily_wear_time_hours": 18

[0172] }

[0173] }

[0174] Finally, the structured results are written to the corresponding data layer, and model recognition logs and manual confirmation status are generated.

[0175] After completing the data archiving, when the user inputs the current natural language screening requirements, such as: screening scoliosis cases suitable for finite element analysis, requiring CT or high-quality 3D reconstruction, complete geometric model, material parameters, mesh file, boundary conditions and stress results.

[0176] This example will call the Large Language Model API to identify the filtering intent:

[0177] Research task type = Finite element analysis; Disease type = Scoliosis;

[0178] Imaging requirements = CT or high-quality 3D reconstruction;

[0179] Geometric model = complete;

[0180] Material parameters = Yes;

[0181] Grid file = Yes;

[0182] Boundary condition = Yes;

[0183] Result file = Stressed results.

[0184] Convert the intent into a database executable condition:

[0185] Diagnosis = Scoliosis;

[0186] CT_status = Available OR Reconstruction_quality = High;

[0187] Geometry_model_status = Complete;

[0188] Material_parameter_file = Available;

[0189] Mesh_file_status = Available;

[0190] Boundary_condition_file = Available;

[0191] FEA_stress_result = Available.

[0192] For example, the user inputs the current natural language screening requirements as: screening for brace treatment cases suitable for constructing machine learning training sets, requiring baseline images, images inside the brace, brace wearing records, follow-up for more than 6 months, and final outcome labels.

[0193] Convert the current natural language screening requirements to:

[0194] Treatment_type = Brace;

[0195] Baseline_image = Yes;

[0196] In_brace_image = Yes;

[0197] Brace_compliance_record = Yes;

[0198] Followup_months >= 6;

[0199] Outcome_label = Available;

[0200] ML_input_feature_status = Complete.

[0201] like Figure 3 As shown, in this embodiment, after obtaining the database executable conditions corresponding to the current natural language screening requirements, a candidate case set is obtained by querying the multi-level case database containing clinical case images and data based on a dual-channel screening mechanism. Specifically, this embodiment can locate research data in the multi-level case database containing clinical case images and data based on the database executable conditions; screen the research data through a preset rule screening channel to obtain a first screening result; screen the research data through a large language model screening channel to obtain a second screening result; and perform channel consistency verification on the first and second screening results to obtain the candidate case set.

[0202] It is understood that this embodiment can locate relevant data in a multi-level database based on the database's executable conditions, including case number; clinical time series records; image measurement results; three-dimensional model path; three-dimensional classification results; finite element model path; finite element stress result path; machine learning input feature table; machine learning labels; follow-up outcomes; and research cohort version.

[0203] For each candidate relevant case data point, not only is the eligibility criterion displayed, but also its location within the research data. For example:

[0204] Case SC_025:

[0205] 3D reconstruction file: / 3DModel / SC_025 / model.stl;

[0206] Centerline file: / Centerline / SC_025 / centerline.csv;

[0207] Finite element model: / FEA / SC_025 / model.cdb;

[0208] Stress results: / FEA / SC_025 / stress_result.csv;

[0209] Machine learning input table: / ML / input / SC_025_features.csv;

[0210] Follow-up outcome: stable.

[0211] This embodiment allows researchers to directly locate finite element results, machine learning data, and 3D morphological results without having to manually search through multiple software programs or folders.

[0212] Then, based on the screening results, the research data is automatically exported, including a case list; reasons for inclusion / exclusion; missing fields; imaging measurement table; three-dimensional typing result table; finite element result index table; machine learning input feature table; machine learning label table; follow-up outcome table; quality control report; and research cohort version number.

[0213] For example, for finite element analysis tasks, the following research data can be exported: Patient_ID; Cobb_main; Curve_type; CT_status; Geometry_model_path; Material_file_path; Mesh_file_path; Boundary_condition_path; Stress_result_path; FEA_quality_status.

[0214] For machine learning training tasks, the following research data can be exported: Patient_ID; Age; Sex; Cobb_main; Risser; Curve_type; Brace_compliance; In_brace_correction_rate; Followup_months; Outcome_label; ML_feature_status; Missing_fields.

[0215] like Figure 3 As shown, after obtaining the research data, this embodiment screens the data through a preset rule screening channel and a large language model screening channel. Specifically, during the preset rule screening process, this embodiment can screen according to the research group's preset rules. For example, the rules for predicting brace efficacy include: diagnosis of AIS; age 10–18 years; Cobb angle 25°–45°; baseline imaging available; in-brace imaging available; follow-up time ≥6 months; brace wearing record available; and changes in Cobb angle during follow-up.

[0216] In the large language model screening process, this embodiment can generate structured screening conditions based on the user's natural language needs, so that non-technical personnel can quickly screen out results that meet the needs of scientific research screening.

[0217] This embodiment generates two case sets after completing the two-channel screening:

[0218] The first screening result R_set: the result of the rule-based screening;

[0219] The second screening result L_set: Screening results of the large language model.

[0220] Then compare the two results:

[0221] R_set ∩ L_set: Both channels consistently include candidate cases in the set;

[0222] R_set - L_set: Cases included in rule-based screening but not included in LLM;

[0223] L_set - R_set: Cases included in LLM but not included in rule screening.

[0224] For inconsistent cases, output the difference field and the reason, for example:

[0225] Difference field: Follow-up time;

[0226] Reason for the discrepancy: LLM recognizes "follow-up examination after six months" as 6 months, but the database date is calculated as 5.5 months.

[0227] Then, quality control processing is performed on the candidate case set to obtain the target scientific research data output package. Specifically, in this embodiment, channel and consistency quality control is performed on the candidate case set, and then the multi-task fit degree of each data in the candidate case set is calculated based on the quality control results. The target scientific research data output package is generated based on the multi-task fit degree and the candidate case set.

[0228] Specifically, the channel and consistency quality control in this embodiment enables quantitative control of clinical and research data. This includes clinical time-series quality control and research data quality control. Clinical time-series quality control includes screening time earlier than initial diagnosis time; baseline images earlier than treatment start time; images within the brace later than brace wearing start time; follow-up images later than baseline images; and outcome evaluation later than the follow-up time point. Research data quality control includes whether the 3D model corresponds to the source images; whether the vertebral body centerline is complete; whether the 3D classification results exist; whether the finite element geometric model is complete; whether the material parameter file exists; whether the boundary condition file exists; whether the stress result file exists; whether the machine learning input features are complete; whether the machine learning labels exist; and whether the follow-up outcome can be used as a label.

[0229] Specifically, the multi-task adaptation in this embodiment includes 3D classification adaptation, finite element analysis adaptation, machine learning training set adaptation, and brace efficacy prediction adaptation. The 3D classification adaptation is calculated based on the 3D reconstruction state, centerline integrity, curve projection integrity, and classification result state. The finite element analysis adaptation is calculated based on the geometric model integrity, material parameter integrity, mesh quality, boundary condition integrity, and stress result integrity. The machine learning training set adaptation is calculated based on the input feature integrity, label integrity, follow-up outcome integrity, missing field ratio, and sample quality. The brace efficacy prediction adaptation is calculated based on baseline images, in-brace images, brace compliance, follow-up time, and outcome indicators.

[0230] Specifically, the target research data output package in this embodiment is output in the form of a comprehensive report, including: patient clinical timeline; case screening results; included cases; excluded cases; cases pending confirmation; reasons for inclusion or exclusion; missing fields; discrepancies; quality control anomalies; 3D subtyping fit; finite element analysis fit; machine learning training set fit; efficacy prediction fit; finite element result path; machine learning data path; 3D model path; research cohort export table; follow-up plan auxiliary suggestions; progression risk warning; and treatment pathway auxiliary prompts. The clinical end focuses on outputting the patient's entire disease course timeline, follow-up status, and treatment pathway auxiliary prompts; the research end focuses on outputting the case cohort, research fit, data file paths, and data tables that can be directly used for analysis.

[0231] As can be seen from the above, the method of the embodiments of this application has the following beneficial effects:

[0232] First, this embodiment can organize clinical data according to the entire timeline of scoliosis, reflecting the continuous process of patients from screening, initial diagnosis, treatment to long-term follow-up.

[0233] Secondly, this embodiment can autonomously classify, extract fields, and archive structured data from input case data through the large language model API, thereby improving data processing efficiency.

[0234] Third, this embodiment allows users to manually modify and confirm table fields, case data, model recognition results, and research status after importing data, thereby improving the accuracy of the database.

[0235] Fourth, this embodiment can support researchers to quickly screen cases using natural language through a dual-channel mechanism of rule-based screening and large language model screening, and reduce the risk of false screening caused by relying solely on model screening.

[0236] Fifth, this embodiment can quickly locate 3D fractal results, finite element models, finite element stress results, machine learning input features, and follow-up outcomes, reducing the workload for researchers to repeatedly search through multiple folders and software.

[0237] Sixth, this embodiment can output case fit scores according to different research tasks, assisting researchers in quickly constructing three-dimensional classification, finite element analysis, machine learning training, and brace efficacy prediction cohorts.

[0238] Seventh, this embodiment can export scientific research data tables, file paths, missing fields, and quality control reports, improving the efficiency and traceability of scientific research dataset construction.

[0239] This application also provides a scientific research data processing system for the entire course of scoliosis, the system comprising:

[0240] The first acquisition module is used to acquire clinical case images and data of patients with scoliosis.

[0241] The time-series processing module is used to process all clinical case images and data based on time-series management to obtain time-series data of the entire course of scoliosis.

[0242] The building module is used to construct a multi-level case database, which includes a patient identity management layer, a clinical time-series follow-up data layer, an imaging measurement data layer, a three-dimensional morphological analysis data layer, a biomechanical analysis data layer, a machine learning prediction data layer, and a research adaptation status data layer.

[0243] The preprocessing module is used to preprocess the time-series data of the entire course of scoliosis.

[0244] The writing module is used to write each sub-data in the preprocessed scoliosis full-course time series data into the corresponding layer of the multi-level case database;

[0245] The second acquisition module is used to acquire the current natural language screening requirements;

[0246] The conversion module is used to convert current natural language screening requirements into database executable conditions based on a large language model.

[0247] The filtering module is used to query a set of candidate cases from a multi-level case database containing clinical case images and data based on a dual-channel filtering mechanism, according to the database's executable conditions.

[0248] The quality control processing module is used to perform quality control processing on the candidate case set to obtain the target scientific research data output package.

[0249] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0250] This application also provides a scientific research data processing system for the entire course of scoliosis, including:

[0251] At least one processor;

[0252] At least one memory for storing at least one program;

[0253] When the at least one program is executed by the at least one processor, the at least one processor implements Figure 1 The method shown.

[0254] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0255] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a computer program, which is implemented when executed by a processor. Figure 1 The method shown.

[0256] It is understood that the content of the above method embodiments is applicable to this medium embodiment. The specific functions implemented in this medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0257] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0258] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0259] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical coding feature maps; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A research data processing method for the entire course of scoliosis, characterized in that, The method includes the following steps: Obtain clinical case images and data of patients with scoliosis; Based on the time-series management method, all the images and data of the clinical cases were processed to obtain the time-series data of the entire course of scoliosis. A multi-level case database is constructed, which includes a patient identity management layer, a clinical time-series follow-up data layer, an image measurement data layer, a three-dimensional morphological analysis data layer, a biomechanical analysis data layer, a machine learning prediction data layer, and a research adaptation status data layer. The time-series data of the entire course of scoliosis were preprocessed; Each sub-data point within the preprocessed scoliosis full-course time-series data is written into the corresponding layer of the multi-level case database; Obtain the current natural language screening requirements; Based on a large language model, the current natural language screening requirements are converted into database executable conditions; Based on the database's executable conditions, a candidate case set is obtained by querying the multi-level case database containing clinical case images and data using a dual-channel screening mechanism. The candidate case set is subjected to quality control processing to obtain the target scientific research data output package.

2. The method according to claim 1, characterized in that, After acquiring the clinical case images and data of scoliosis patients, the method further includes the following steps: The clinical case images and data are structurally classified, and hierarchical identifiers are determined for the clinical case images and data; wherein, the hierarchical identifiers include clinical identity identifiers, research identifiers, and model call identifiers.

3. The method according to claim 1, characterized in that, The time-series management method is used to process all the clinical case images and data to obtain time-series data of the entire course of scoliosis, including: The clinical process corresponding to the clinical case images and data is divided into multiple time points, and each time point is associated with one type of data in the clinical case images and data. A clinical time-series data chain of the clinical case images and data is constructed based on the aforementioned time points; The time points in the clinical time series data chain are subjected to time series consistency quality control to obtain the time series data of the entire course of scoliosis.

4. The method according to claim 1, characterized in that, The preprocessing of the time-series data of the entire course of scoliosis includes: Field matching is performed on each data point in the time-series data of the entire course of scoliosis. The data for which no matching field was found in the time series data of the entire course of scoliosis were manually determined, along with the recognition structure of the large language model. Data conflict analysis was performed on the time-series data of the entire course of scoliosis. Modify record information based on field matching results, manual determination results, and data conflict analysis results.

5. The method according to claim 1, characterized in that, The step of obtaining a candidate case set by querying a multi-level case database containing written clinical case images and data based on the database's executable conditions and a dual-channel screening mechanism includes: Based on the database's executable conditions, research data is located in a multi-level case database after clinical case images and data have been written. The research data is screened through a preset rule-based screening channel to obtain the first screening result; The research data was screened using a large language model screening channel to obtain a second screening result; The first screening result and the second screening result are subjected to channel consistency verification to obtain the candidate case set.

6. The method according to claim 1, characterized in that, The quality control process performed on the candidate case set yields the target scientific research data output package, including: Channel and consistency quality control were performed on the candidate case set; Calculate the multi-task fit of each data point in the candidate case set based on the quality control results; The target scientific research data output package is generated based on the multi-task adaptability and the candidate case set.

7. The method according to claim 6, characterized in that, The multi-task adaptation includes three-dimensional classification adaptation, finite element analysis adaptation, machine learning training set adaptation, and brace efficacy prediction adaptation.

8. A research data processing system for the entire course of scoliosis, characterized in that, The system includes: The first acquisition module is used to acquire clinical case images and data of patients with scoliosis. The time-series processing module is used to process all the images and data of the clinical cases based on the time-series management method to obtain the time-series data of the entire course of scoliosis. The module is used to build a multi-level case database, which includes a patient identity management layer, a clinical time-series follow-up data layer, an image measurement data layer, a three-dimensional morphological analysis data layer, a biomechanical analysis data layer, a machine learning prediction data layer, and a research adaptation status data layer. The preprocessing module is used to preprocess the time-series data of the entire course of scoliosis. The writing module is used to write each sub-data in the preprocessed scoliosis full-course time series data into the corresponding layer of the multi-level case database. The second acquisition module is used to acquire the current natural language screening requirements; The conversion module is used to convert the current natural language screening requirements into database executable conditions based on a large language model; The filtering module is used to query a set of candidate cases from a multi-level case database containing clinical case images and data based on the database's executable conditions and a dual-channel filtering mechanism. The quality control processing module is used to perform quality control processing on the candidate case set to obtain the target scientific research data output package.

9. A research data processing system for the entire course of scoliosis, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images