Hospital paper material electronic storage management method and electronic device

Abstract

The application provides a hospital paper material electronic storage management method and an electronic device. The method comprises the following steps: converting paper materials provided by corresponding departments of a hospital into digital images; converting the digital images into editable electronic files; performing error and omission detection on the electronic files through a pre-created quality control rule library; issuing an error and omission prompt information when the electronic files have errors and omissions; receiving a corrected electronic file uploaded based on the error and omission prompt information, and repeatedly performing the error and omission detection step based on the corrected electronic file until a preset stop condition is met; associating the last corrected electronic file with the corresponding digital image, and performing classified storage. In this way, the error and omission detection of batch electronic files can be automatically realized, and the detection efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, and more specifically, to a method and electronic device for the electronic storage and management of hospital paper materials. Background Technology

[0002] In the current daily operations of hospitals, a large number of medical documents still exist in paper form, including but not limited to doctor's orders, nursing records, examination and test reports, medical record cover sheets, and various medical approval forms. With the advancement of medical informatization, hospitals' need for electronic management of paper materials is becoming increasingly urgent. The digitization of paper materials largely relies on traditional OCR (Optical Character Recognition) technology for text extraction. However, even after digitization, errors and omissions that are easily overlooked still exist. For example, the content recorded in the paper materials themselves may contain errors or omissions, or errors and omissions may exist during the digitization process. These errors and omissions often require manual checking. Given the large variety and quantity of paper materials, manual checking is inefficient and prone to overlooking errors and omissions due to fatigue or negligence. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method and electronic device for the electronic storage and management of hospital paper materials, which can improve the problem of low efficiency in checking for errors and omissions after the paper materials are digitized in hospitals.

[0004] To achieve the above technical objectives, the technical solution adopted in this application is as follows:

[0005] In a first aspect, embodiments of this application provide a method for electronic storage and management of hospital paper materials, the method comprising:

[0006] S10 converts paper materials provided by the relevant departments of the hospital into digital images;

[0007] S20, convert the digitized image into an editable electronic file;

[0008] S30, using a pre-created quality control rule library, perform error and omission detection on the electronic file;

[0009] S40, when there are errors or omissions in the electronic document, an error or omission prompt message is issued;

[0010] S50: Receive the corrected electronic file uploaded based on the error message, and repeat step S30 based on the corrected electronic file until the preset stop condition is met to obtain the final corrected electronic file.

[0011] S60 associates the final revised electronic file with the corresponding digital image and stores them in categories.

[0012] Secondly, embodiments of this application also provide an electronic device, which includes a processor and a memory coupled to each other. The memory stores a computer program, and when the computer program is executed by the processor, the electronic device is able to perform the above-described method.

[0013] The invention employing the above technical solution has the following advantages:

[0014] The technical solution provided in this application converts paper materials provided by relevant departments of the hospital into digital images, then converts the digital images into editable electronic files, and finally uses a quality control rule base to detect errors and omissions in the electronic files. This facilitates the automatic batch detection of errors and omissions in electronic files, thereby improving detection efficiency. Furthermore, since manual error checking is unnecessary, it helps avoid errors and omissions that are easily overlooked due to fatigue or negligence during manual inspection. Attached Figure Description

[0015] This application can be further illustrated by the non-limiting embodiments given in the accompanying drawings. It should be understood that the following drawings only illustrate some embodiments of this application and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained from these drawings without any inventive effort.

[0016] Figure 1 A flowchart illustrating the electronic storage and management method for hospital paper materials provided in this application embodiment.

[0017] Figure 2 A functional block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that similar or identical parts are referred to by the same reference numerals in the drawings or description. Implementations not shown or described in the drawings are forms known to those skilled in the art. In the description of this application, terms such as "first" and "second" are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0019] Please refer to Figure 1 This application provides a method for electronic storage and management of hospital paper materials. This method can be applied to electronic devices, which execute or implement the steps of the method. The electronic device can be, but is not limited to, personal computers, smartphones, etc. The hospital paper material electronic storage and management method may include the following steps:

[0020] S10 converts paper materials provided by the relevant departments of the hospital into digital images;

[0021] S20, convert the digitized image into an editable electronic file;

[0022] S30, using a pre-created quality control rule library, perform error and omission detection on the electronic file;

[0023] S40, when there are errors or omissions in the electronic document, an error or omission prompt message is issued;

[0024] S50: Receive the corrected electronic file uploaded based on the error message, and repeat step S30 based on the corrected electronic file until the preset stop condition is met to obtain the final corrected electronic file.

[0025] S60 associates the final revised electronic file with the corresponding digital image and stores them in categories.

[0026] In the above implementation, paper materials provided by the relevant departments of the hospital are converted into digital images, and then the digital images are converted into editable electronic files, achieving a precise conversion from paper materials to editable electronic files and avoiding manual secondary editing. A quality control rule base is then used to detect errors and omissions in the electronic files, automating the identification of errors and omissions in medical records. This facilitates the automated batch detection of errors and omissions in electronic files, thereby improving detection efficiency. Through a closed-loop mechanism of "detection-prompt-correction-re-check," the integrity and accuracy of medical records are ensured, reducing medical disputes caused by defects in medical records. Furthermore, since no manual error checking is required, it helps avoid oversights that are easily missed during manual checks due to fatigue or negligence.

[0027] The following is a detailed explanation of each step in the method for electronic storage and management of hospital paper materials:

[0028] In step S10, the paper materials come from various paper medical materials submitted by hospital clinical departments (internal medicine, surgery, etc.), medical technology departments (radiology, ultrasound, etc.), and administrative and logistical departments (quality control department, medical records room, etc.).

[0029] Paper materials: refers to all kinds of paper medical documents generated by various departments of the hospital during the diagnosis and treatment process, including but not limited to doctor's orders, nursing records, examination and test reports, medical record cover sheets, medical approval forms, X-ray films, ultrasound reports and other flat / curved / special-sized paper carrier materials.

[0030] The way to convert paper materials into digital images is to use conventional scanning equipment, cameras, etc., to scan or photograph the paper materials to obtain an initial image, and then preprocess the initial image to obtain standardized, high-definition image data, which is the digital image.

[0031] The initial image is preprocessed in a conventional manner, such as image enhancement, denoising, and geometric correction.

[0032] Specifically, image enhancement can employ an adaptive histogram equalization algorithm to dynamically adjust the grayscale distribution and address issues of local over-darkness or over-brightness, thereby achieving image enhancement.

[0033] Noise reduction: A median filtering algorithm is used to adaptively select a 3×3 window (noise density ≤ 30%) or a 5×5 window (noise density > 30%) based on the noise density to filter speckle noise and Gaussian noise.

[0034] Geometric correction: The tilt angle is determined by detecting straight line edges through Hough transform and then rotated for correction. The SIFT (Scale-Invariant Feature Transform) feature matching algorithm is used to stitch and fuse multiple frames of images to obtain a complete upright image.

[0035] The digitized images include a first type of digitized images obtained by scanning text-based materials, a second type of digitized images obtained by scanning tabular materials, and a third type of digitized images obtained by scanning image-based materials.

[0036] Understandably, the first type of digital image refers to the initial digital image obtained by scanning text-based paper materials (such as medical orders, nursing records, and diagnostic certificates) using a scanner or industrial camera. It mainly consists of text content and does not contain complex tables or image elements.

[0037] The second type of digital image refers to the initial digital image obtained by scanning tabular paper materials (such as inspection and testing reports, medical approval forms) with a scanner or industrial camera, containing table elements with a clear row and column structure and cell partitioning.

[0038] The third category of digital images refers to the initial digital images obtained by scanning paper materials (such as X-ray films, ultrasound images, and pathological slide photographs) with a scanner or industrial camera. The core feature is that the images are the main information and the text content is secondary (or there is no text).

[0039] In step S20, the digitized image is converted into an editable electronic file, including:

[0040] The first type of digitized image is converted into an editable electronic document using an OCR recognition model.

[0041] The second type of digitized images are converted into editable spreadsheets using a table conversion model.

[0042] The third type of digital image is converted into an image of a specified format using an image conversion model, and the image, electronic document, and spreadsheet of the specified format are treated as electronic files.

[0043] Specifically, OCR recognition model: refers to an OCR recognition model based on the Transformer architecture, which has the ability to extract text and restore structure, and is adapted to the font and format features of medical documents.

[0044] Electronic devices can use a Transformer-based OCR recognition model to extract text content from first-class digital images line by line; and extract format features such as line spacing, font size, bold / italic, etc., to restore the original document's title hierarchy (such as first-level title "Medical Orders", second-level title "Long-term Medical Orders"), paragraph breaks, and signature bar position; and encapsulate the recognized text and format features into a DOCX format file as an editable electronic document to ensure that the format is not lost during editing.

[0045] The table conversion model refers to a combined model based on the YOLOv8 object detection model and the OCR recognition model, which has the capabilities of table region detection, cell segmentation, text extraction, and table structure reconstruction. Editable spreadsheets refer to Excel format spreadsheet files.

[0046] Electronic devices can use the YOLOv8 object detection model to detect table regions in second-type digital images and output table boundary coordinates; based on the table boundary coordinates, cells are segmented and the row number, column number, and coordinate information of each cell are output; corresponding OCR recognition is performed on the text in each cell; then, based on the row and column relationship and coordinate overlap relationship of the cells, table borders are automatically generated, cells are merged, an EXCEL format file is constructed, and the text obtained by OCR recognition of each cell (cells in second-type digital images) is written into the cells of the generated table, thereby obtaining an editable spreadsheet, and the content of the spreadsheet corresponds to the content and format of the paper material table.

[0047] Image conversion models can refer to a combination of image quality optimization algorithms (sharpening, deblurring) and BLIP (Bootstrapped Language-Image Pretraining) models. The specified format can be high-resolution PNG (suitable for medical image storage requirements, preserving detailed information).

[0048] Electronic devices can use sharpening algorithms to enhance edge details of third-type digital images and deblurring algorithms to repair blurred areas caused by shooting / scanning, thereby optimizing image quality. The optimized image is then saved as a PNG file. Next, the BLIP model is used to extract key image information (such as patient name, examination type, and examination date), generate image description text, and associate it with the PNG file.

[0049] In this embodiment, different models are used for paper materials with different content formats such as text, tables, and images to convert them into corresponding editable electronic files. This can improve the problem of uneditable scanned digital images, eliminate the need for manual secondary editing, significantly reduce the workload of medical staff, and shorten the processing time for digitizing paper materials.

[0050] Before step S10, the method may further include:

[0051] Configure the first type of rule for basic information consistency checks;

[0052] Configure a second type of rule for checking the completeness of required fields;

[0053] Configure the third type of rule for consistency detection of diagnostic and treatment logic;

[0054] Configure the fourth type of rule for format conformity checks;

[0055] Configure the fifth type of rules for medical record integrity detection, and form a quality control rule base based on the first, second, third, fourth and fifth types of rules.

[0056] Specifically, basic information may include, but is not limited to, patient identity information and information related to treatment affiliation, such as name, medical record number, gender, age, department, and inpatient / outpatient identification.

[0057] The configuration method for the first type of rule is as follows: Define that the name, medical record number, gender, and age must be completely consistent in different electronic files of the same medical record (such as doctor's orders and nursing records). The medical record number is a specified number of digits (such as 10 digits), the age is an integer (such as 0-120 years old), the gender can be "male / female", and the department name is consistent with the hospital department code table.

[0058] In the first type of rules, regular expressions can be used to check whether elements such as name and medical record number are consistent. Each element (such as name and medical record number) has a corresponding regular expression; that is, each regular expression is a sub-rule in the first type of rules. By configuring a corresponding regular expression for each element, the first type of rules for basic information consistency checks can be formed.

[0059] In the second category of rules, required fields are document fields that must be filled in according to medical standards or hospital management requirements. These are divided into general required fields and type-specific required fields. The second category of rules includes rules corresponding to the integrity checks of general required fields and rules corresponding to the integrity checks of type-specific required fields.

[0060] As an example, here's how to configure the rules for general required field integrity checks: Define all medical record documents as requiring the patient's medical record number, creation time, and recorder's signature.

[0061] The rules for integrity checks of required fields specific to configuration types are defined as follows:

[0062] The doctor's order form must include the contents of the order, dosage, usage, prescribing doctor, date of prescribing, and date of administration.

[0063] Nursing records must include vital signs (temperature, blood pressure, etc.), nursing procedures, recording time, and the recording nurse;

[0064] Surgical records must include the name of the surgery, the time of the surgery, the surgeon, the method of anesthesia, and the intraoperative details.

[0065] Specifically, the second type of rule has a corresponding integrity checklist. This checklist lists the items that need to be checked for different types of materials. For example, for a "medical order," the integrity checklist might list items such as "medical order content, dosage, usage, prescribing doctor, prescribing time, and execution time." During the integrity check, the extracted keywords are matched against the items in the corresponding integrity checklist. If some items in the integrity checklist are not matched, it means that there are missing required fields in the corresponding paper material (or corresponding electronic document). If all items in the corresponding integrity checklist are matched, it means that the required fields in the paper material (or corresponding electronic document) are complete and there are no missing required fields.

[0066] In the third category of rules, consistency of diagnosis and treatment logic refers to the causal relationship and rationality between medical orders, nursing records, examination and test results, and diagnostic conclusions, which conform to medical diagnosis and treatment standards.

[0067] As an example, the configuration methods for the third type of rule can include:

[0068] Configure the consistency rule between medical orders and diagnoses, that is, define the matching between the diagnosis result and the content of the medical order (e.g., when diagnosing "diabetes", the medical order should not include "high sugar injection").

[0069] Configure nursing-medical order consistency rules, that is, define that nursing records must match the requirements of medical orders (e.g., if the medical order is "measure body temperature every 4 hours", the nursing record must have corresponding frequency data).

[0070] Configure the consistency rules between test results and diagnosis, that is, define the association between test results and diagnostic conclusions (e.g., if a blood routine test shows "significantly elevated white blood cell count", the diagnosis must include a "contagion" related statement).

[0071] In the fourth category of rules, format standardization refers to the conformity of the date, page number, signature, text layout, etc. of medical records with the unified hospital format standards.

[0072] As an example, the configuration methods for the fourth type of rule can include:

[0073] Configure date format rules, for example, define all time fields to be uniformly "YYYY-MM-DD HH:MM:SS" (e.g., "2025-12-19 14:30:00");

[0074] Configure signature rules, such as defining that relevant medical staff must provide handwritten or electronic signatures, and that the signature should be placed in a specified area of ​​the document;

[0075] Configure page numbering rules, for example, define that multi-page documents should be marked with page numbers (format "Page X of Y"), and the page numbers should be consecutive without any gaps.

[0076] In the fifth category of rules, medical record integrity refers to the fact that, based on the list of documents to be archived corresponding to the diagnosis and treatment / fee items, the actual medical record data collected is complete and meets the requirements for diagnosis and treatment traceability.

[0077] As an example, the configuration methods for the fourth type of rule can include:

[0078] Construct a mapping table of "diagnosis / treatment / fee items - documents to be archived" (example: abdominal ultrasound examination → abdominal ultrasound report; laparoscopic cholecystectomy → surgical record + preoperative discussion record + postoperative nursing record).

[0079] Define integrity verification rules:

[0080] For any examination / test prescribed by a doctor, the corresponding report must be filed.

[0081] The expense breakdown includes charges for examinations / tests / treatments, and the corresponding reports / records must be filed.

[0082] Related data within the same treatment cycle must be complete (e.g., surgical records must be accompanied by preoperative discussion records).

[0083] In this embodiment, a structured storage format (such as JSON) is used to store the rules. Each rule includes a rule number, rule name, applicable document type, validation logic, and violation weight. A quality control rule base is built, allowing quality control departments to add, modify, and delete rules through a backend interface, with operation logs automatically saved. A rule version management mechanism is established to retain historical rule versions and support rollback. This covers all core requirements of medical record quality control, and the rule base is dynamically expandable to adapt to updates in medical standards and the personalized needs of departments, solving the problem of traditional quality control rules being fixed and unable to be iterated. The "Diagnosis / Treatment / Billing Items - Documents to be Archived" mapping table systematizes and standardizes integrity verification, improving the accuracy of identifying missing documents. The structured storage and version management of the quality control rule base facilitates the quality control department's ability to trace rule changes, improving the standardization and operability of medical quality control management.

[0084] In step S30, the electronic document is checked for errors and omissions using a pre-created quality control rule base, including:

[0085] S31, Perform feature extraction on the electronic document to obtain features to be verified. The features to be verified include at least one of the following: basic information features, required field filling features, diagnosis and treatment association features, format features, and integrity verification features.

[0086] S32, Select rules corresponding to the features to be verified from the pre-created quality control rule base for error detection, and calculate the total error risk score using a preset weighted rule matching algorithm. The calculation formula for the weighted rule matching algorithm is as follows:

[0087] (1)

[0088] In the formula, R represents the total error / omission risk score; This represents the weight of the k-th rule; This represents the number of sub-rules contained in the k-th rule category; Let be the matching degree of the i-th sub-rule in the k-th rule category; The importance weight of the i-th sub-rule in the k-th rule category;

[0089] S33. Based on the total error risk score, the error level is determined according to the preset level classification strategy. The error level includes level 0 error, which indicates no error; level 1 error, which indicates a general error; level 2 error, which indicates a serious error; and level 3 error, which indicates a fatal error.

[0090] Specifically, information such as patient name, medical record number, gender, age, and department is extracted from electronic files to form basic information features;

[0091] For electronic file types, extract the required fields corresponding to the second type of rules, mark "filled" (1) or "not filled" (0), and obtain the required field filling features;

[0092] Extract the diagnosis conclusions, doctor's orders, nursing records, and examination and test results in the electronic file to form a diagnosis and treatment semantic vector as the diagnosis and treatment association features;

[0093] Extract the date format, signature status, and page number information of the electronic file as format features to compare with the fourth type of rules;

[0094] Generate a list of materials to be archived based on the "Diagnosis / Treatment / Charging Items - Materials to be Archived" mapping table, compare it with the list of actually collected medical record materials, and construct a "to be archived - already collected" feature vector (already collected = 1, not collected = 0, not required to be archived = -1), thereby obtaining the integrity verification features.

[0095] As an example, the implementation process of step S32 is as follows:

[0096] For each type of feature to be verified, select the corresponding rules for comparison (such as the basic information features matching the first type of rules, and the integrity verification features matching the fifth type of rules) to determine the matching degree of each sub-rule (fully matched = 1, partially matched = 0.5, not matched = 0);

[0097] Load the preset weight parameters, where (consistency of basic information), (completeness of required fields), (consistency of diagnosis and treatment logic), (standardization of format), (completeness of medical records), Assign values according to the severity of errors and omissions (fatal = 5, serious = 3, general = 1);

[0098] Substitute into the above formula (1) to calculate the total score R of the error and omission risk (value range 0 - 100).

[0099] The grading strategy refers to the interval division based on the total score R of the error and omission risk, and clarifies the severity of errors and omissions and the processing priorities corresponding to different grades. As an example, the implementation process of step S33 is as follows:

[0100] Level 0 error and omission: R = 0, the features to be verified completely match all rules, without any errors and omissions;

[0101] Level 1 error and omission: 0 < R < 30, only general errors and omissions such as non-standard format and non-core expressions exist, which do not affect the diagnosis and treatment traceability;

[0102] Level 2 errors: 30≤R<60, such as serious errors such as missing filing of ordinary test reports, missing non-critical required fields, and flaws in minor diagnostic logic, which require manual verification;

[0103] Level 3 errors: R≥60, such as missing archiving of key data such as ultrasound reports / surgical records, contradictions in core diagnosis and treatment logic, inconsistencies in basic information, etc., which are fatal errors that affect medical safety.

[0104] In this embodiment, multi-dimensional features to be verified are beneficial for comprehensively covering key points of medical record quality control and avoiding errors or omissions caused by single features. The weighted rule matching algorithm, through scientific weight allocation, highlights the core position of consistency of diagnosis and treatment logic and integrity of medical records, and the priority of core error and omission identification is increased, which helps to reduce the rate of missed detection of fatal errors and omissions. By quantitatively classifying error and omission levels, error and omission handling becomes more targeted, reducing the decision-making cost of the quality control department and improving the efficiency of quality control review.

[0105] In step S40, when the electronic document contains errors or omissions, an error / omission warning message is issued, including:

[0106] S41, when the electronic document contains a Level 1 error indicating a general error, an error prompt message and a correction suggestion corresponding to the Level 1 error are sent to the review end. When the review end sends an instruction to accept the correction suggestion, the electronic document is corrected according to the correction suggestion to obtain the corrected electronic document.

[0107] S42, when the electronic document contains level 2 or level 3 errors, the corresponding error message is sent to the review terminal so that the reviewer can upload the corrected electronic document based on the error message.

[0108] Understandably, when an electronic document is free of errors or omissions, i.e., it is a level zero error or omission, the electronic document is directly associated with the corresponding digital image and stored in a classified manner.

[0109] In this embodiment, the review terminal can be a terminal device (computer, mobile tablet) deployed within the hospital for use by medical staff and departmental quality control specialists, integrating error and omission notification reception, correction operations, and file upload functions.

[0110] The proposed corrections can be format correction schemes that are automatically generated based on the fourth type of rules (format specification) and can be executed directly.

[0111] As an example, step S41 is implemented as follows:

[0112] When an electronic document contains a Level 1 error indicating a general omission, the electronic device can automatically generate an error message. This message may include the location of the error (e.g., "Date format error on page 3"), the violation rule (e.g., "Formatting rule 4.1"), and a suggested correction (e.g., "Change '2025.12.19' to '2025-12-19 10:00:00'"). The electronic device can then push this message to the relevant medical staff's review terminal via the hospital's intranet. After reviewing the message, if the medical staff clicks "Accept Correction Suggestion," the system will automatically perform the correction. If they click "Manual Correction," they can edit the electronic document on the review terminal and submit it. The electronic device can then automatically correct the document or receive the manually corrected version, generating or obtaining the corrected electronic document.

[0113] As an example, step S42 is implemented as follows:

[0114] When electronic documents contain Level 2 or Level 3 errors, the electronic device can generate a detailed error report, including the error type (e.g., "missing data" or "logical contradiction"), the corresponding diagnostic and treatment items (e.g., "complete blood test"), the deadline for supplementing the missing data (e.g., "within 24 hours"), and the basis for the violation (e.g., "Category 5 Rule 5.2"). The error report is then pushed to the department's quality control specialist and the attending physician's review end, with a simultaneous hospital-wide message reminder. After verification through the review end, if the error is due to missing data (e.g., missing report filing), the paper materials are supplemented and resubmitted to step S10 (digital data collection), generating an electronic document which is then uploaded through the review end. If the error is due to logical contradictions (e.g., incorrect medical orders), the electronic document is edited and corrected on the review end before being uploaded.

[0115] In this embodiment, errors and omissions are handled in a tiered manner. General errors and omissions are automatically corrected, while serious / critical errors and omissions require precise manual intervention, balancing processing efficiency and accuracy. Error and omission alerts can include complete evidence of violations and processing guidelines, which helps lower the operational threshold for auditors. Furthermore, error and omission alerts help prevent delays in error and omission processing.

[0116] In step S50, the preset stopping conditions may include: the electronic document reaches level zero error after re-inspection (no errors); or it is specially approved for filing by the quality control department (in special cases where the data cannot be supplemented). That is, receiving an instruction from the quality control terminal (smartphone or computer) indicating special approval for filing is considered to meet the preset stopping conditions.

[0117] In step S60, the final corrected electronic document is associated with the corresponding digital image. The digital image can serve as a backup file of the original paper material, making it convenient to compare and view the content of the initial scanned image and the corrected electronic document later.

[0118] In step S60, the final revised electronic file is associated with the corresponding digitized image and classified and stored, including:

[0119] S61, extract classification features from the last revised electronic document, the classification features including basic information features, content semantic features, diagnosis and treatment stage features and integrity association features;

[0120] S62, the classification features are standardized to obtain a standardized classification feature vector. The standardization process includes: deduplicating the classification features, performing one-hot encoding on the text features in the classification features, and normalizing the numerical features in the classification features.

[0121] S63, input the standardized basic information features into the trained department classification model to obtain the corresponding department classification label;

[0122] S64, Based on the pre-established mapping relationship between treatment stages and keywords, determine the treatment stage corresponding to the keywords in the treatment stage features, and use it as a treatment stage classification label;

[0123] S65. Input the standardized content semantic features into the trained file classification model to obtain file type classification labels;

[0124] S66. Based on the integrity association features, obtain integrity classification labels, which include a first category label for complete original data and a second category label for complete data after completion.

[0125] S67. Based on the department classification label, treatment stage classification label, document type classification label, and integrity classification label, generate the corresponding storage path, and encrypt and store the last corrected electronic file and the corresponding digital image according to the storage path to form a database, and construct a multi-dimensional index corresponding to the corresponding electronic file. The multi-dimensional index includes medical record number index, classification label index, creation time index, and keyword index.

[0126] As an example, step S61 is implemented as follows:

[0127] Extract information such as the patient's department, inpatient / outpatient identification, document creation time (extracted from the document text or collection date), and original document type label (text / table / image) from the final revised electronic file as basic information features;

[0128] The conventional BERT pre-trained model is used to semantically encode the text content of electronic documents (including table text and image description text) to generate a 768-dimensional semantic feature vector, capturing diagnostic and treatment information such as "gallbladder stones" and "laparoscopic surgery" to obtain content semantic features;

[0129] Keywords such as "preoperative discussion", "intraoperative record" and "postoperative care" are extracted from the document text of electronic files. Combined with HIS treatment cycle data (admission time, operation time, discharge time), preliminary treatment stage identifiers are formed to obtain treatment stage characteristics.

[0130] Based on the fifth type of rule, the integrity verification of electronic documents can yield status information indicating whether the electronic document is "originally complete" or "complete after supplementation", as well as the corresponding diagnosis and treatment item association label (such as "abdominal ultrasound examination association"), which can be used as integrity association features.

[0131] In step S62, standardization refers to performing a unified format conversion on the classification features to eliminate data dimensional differences and redundant information, thereby improving the inference efficiency and accuracy of the classification model.

[0132] As an example, step S62 is implemented as follows:

[0133] Deduplication: Remove duplicate features (e.g., if a patient's medical record number appears repeatedly in the basic information and semantic features, keep only one copy).

[0134] Text feature encoding: Text features such as department names and diagnostic keywords are converted into numerical vectors using one-hot encoding;

[0135] Numerical feature normalization: For numerical features such as semantic feature vectors and creation timestamps, the Min-Max normalization algorithm is used to map them to the [0,1] interval to eliminate the influence of dimensions;

[0136] Vector integration: Integrating all processed features into a standardized classification feature vector.

[0137] In step S63, the department classification model refers to a hybrid model of "rule matching + lightweight MLP classifier", with rule matching taking priority and the classifier assisting in judgment in fuzzy scenarios.

[0138] As an example, step S63 is implemented as follows:

[0139] Rule matching: Directly extract the "patient's department" feature from the basic information, perform preliminary classification by referring to the department code lookup table, and generate department labels;

[0140] Alternatively, for fuzzy scenario verification: if the document does not specify the department (such as a general approval form), input the standardized classification feature vector into a lightweight MLP classifier (768-dimensional input layer, 256-dimensional hidden layer, and the number of departments in the output layer), combine it with the semantic feature vector to determine the department to which it belongs, and output the department classification label.

[0141] In step S64, the mapping relationship between diagnosis and treatment stages and keywords refers to the preset "diagnosis and treatment stage - core keywords" comparison table (e.g., the preoperative stage corresponds to "preoperative discussion, preoperative assessment", and the intraoperative stage corresponds to "surgical record, intraoperative nursing").

[0142] Treatment stage classification labels: These are labels that identify the treatment stage to which a document belongs, including the preoperative stage, intraoperative stage, postoperative stage, outpatient treatment stage, and rehabilitation follow-up stage.

[0143] As an example, step S64 is implemented as follows:

[0144] Keyword matching: Match the core keywords in the "Diagnosis and Treatment Stage - Core Keywords" lookup table with the semantic features of the content to preliminarily determine the diagnosis and treatment stage;

[0145] Time correlation verification: Verify by combining the document creation time with HIS treatment cycle data (e.g., if the creation time is within 1 week before the surgery and contains the keyword "preoperative", confirm "preoperative stage"), and avoid keyword ambiguity;

[0146] Generate tags: Determine the final treatment stage and generate classification tags for that stage.

[0147] In step S65, the file classification model is obtained by training a classification model using semantic feature vectors and document type label encodings. The classification model can be a lightweight MLP (Multilayer Perceptron) model. During the model application inference stage, the semantic feature vectors can be input into the file classification model, and the file classification model can output the document type probability distribution.

[0148] Document type classification tags refer to tags that identify the core type of document (such as "ultrasound report", "surgical record", "outpatient prescription").

[0149] As an example, step S65 is implemented as follows:

[0150] The standardized semantic features of the content are input into the file classification model, and the probability distribution is output through the Softmax function. Combined with the document name comparison (match degree ≥ 90% confirmation), file type classification labels are generated.

[0151] In step S66, for the obtained electronic file, if it is complete from the initial acquisition, it is labeled with the first type of label; if it is complete after supplementation, it is labeled with the second type of label.

[0152] In step S67, the storage path refers to a structured path built based on four categories of labels. The preset format can be "Hospital / Department / Treatment Stage / Completeness Classification / File Type / Patient Medical Record Number / Creation Date / Document Name". The multi-dimensional index refers to an indexing system used for fast retrieval, covering four dimensions: medical record number, category label, creation time, and keywords, thus improving retrieval efficiency.

[0153] In step S67, the final revised electronic file and the corresponding digital image are encrypted and stored, which may include:

[0154] The AES-256 encryption algorithm is used to encrypt and store the final corrected electronic files and corresponding digital images, and access permissions are set according to the corresponding classification labels.

[0155] Specifically, during the encrypted storage process, the AES-256 encryption key can be dynamically generated by the hospital's information security system. The key complies with the Level 3 security protection requirements and is rotated regularly (every 90 days). Electronic documents and digitized images are encrypted separately, and metadata such as file format and file name are preserved during the encryption process to ensure that they can be opened and edited normally after decryption. The encryption key is stored on a separate key management server, which is physically isolated from the storage database and can only be accessed by authorized information security administrators.

[0156] Access permissions refer to the permissions to operate on encrypted files (view, edit, download, delete) based on user roles, and are associated with category tags (such as department tags that restrict access to users within that department).

[0157] Methods for setting access permissions can include:

[0158] Role-based access control (RBAC) model, with preset mapping relationships between roles and permissions (e.g., doctors can only view "preoperative / intraoperative / postoperative" related documents for patients in their department, quality control specialists can view quality control related documents for the entire department, and patients can only view their own outpatient / inpatient summary documents).

[0159] Tag-related permissions: Bind category tags (especially department tags and treatment stage tags) to role permissions. For example, files tagged "Surgery - Intraoperative Stage" can only be viewed by surgeons and surgical quality control specialists.

[0160] Permission storage: Associate permission configuration information with the storage path of encrypted files and store it in the permission management module of the electronic device.

[0161] Category tags are linked to access permissions, enabling fine-grained control over data access, preventing unauthorized access, and allowing flexible permission configuration to adapt to the hospital's organizational structure. It also supports adding roles and adjusting permissions, improving the flexibility and scalability of data management.

[0162] As an example, step S67 is implemented as follows:

[0163] The system concatenates four types of tags, patient medical record number, and creation date according to a preset format to generate a unique storage path. Electronic files and digitized images are encrypted using the AES-256 encryption algorithm and stored in a "distributed main storage + cold backup storage" architecture (the main storage stores data for the past 3 years, and the cold backup is stored off-site). The "distributed main storage + cold backup storage" architecture is a conventional storage architecture. A multi-dimensional index is constructed based on the medical record number, category tags (a combination of four types of tags), creation time, and document keywords (extracted from content semantic features) and associated with the storage path.

[0164] In this embodiment, four-level classification tags enable refined categorization of medical documents, structure storage paths, and multi-dimensional indexes, which helps to shorten document retrieval response time. Encrypted storage helps reduce the risk of data leakage and improves data storage security.

[0165] As an optional implementation, the method may further include:

[0166] Upon receiving a retrieval request for retrieving a corresponding electronic file, the system searches the database for an electronic file that matches the index information carried in the retrieval request, using this file as the target file.

[0167] When the access permission corresponding to the retrieval request is greater than or equal to the access permission of the target file, reading the target file is permitted;

[0168] If the access permission corresponding to the retrieval request is less than the access permission of the target file, the target file will not be read, and a corresponding prompt message will be issued.

[0169] Specifically, a search request refers to a request initiated by hospital staff (doctors, quality control specialists, etc.) through the hospital's search system to find target electronic documents, supporting both fuzzy and precise searches.

[0170] Index information refers to the keywords or identifiers carried in the retrieval request that are used to match files, including medical record number, category label, creation time, and document keywords (such as "abdominal ultrasound").

[0171] The target file refers to an encrypted electronic file and its corresponding digital image that exactly matches (precise retrieval) or highly matches (fuzzy retrieval) the index information in the retrieval request.

[0172] After receiving a search request, the electronic device can parse the request, extract index information (such as "Medical Record Number XXX + Category Tag = Ultrasound Report + Creation Time = 2025-12-19"), obtain a multi-dimensional index, identify the user who initiated the search, and associate their role information (such as "Internal Physician - Role ID 001"). Based on the multi-dimensional index, the electronic device quickly locates the storage path that matches the index information (such as locking the patient folder by the medical record number index and locking the "Ultrasound Report" subfolder by the category tag index); the search returns files that match exactly; the fuzzy search returns a list of the top 10 files with the highest matching degree (sorted in descending order of matching degree); and the encrypted files that match are extracted as the target files.

[0173] Next, the electronic device retrieves the permission level corresponding to the user's role and compares it with the access permission level of the target file. If the permissions are sufficient (user permissions ≥ file permissions), the electronic device automatically calls the decryption key from the key management server to decrypt the target file, generating a viewable file format (such as DOCX or PNG), and sends the result back to the user.

[0174] If permissions are insufficient (user permissions < file permissions): the access request will be denied, and a prompt message will be generated (e.g., "Your role does not have permission to view surgical-related documents. Please contact the quality control department to apply for authorization"). The retrieval and permission verification are linked to ensure compliant data access and prevent unauthorized access to sensitive information.

[0175] As an optional implementation, the method may further include:

[0176] Based on the update operation command, the corresponding rules in the quality control rule base are updated using the new rules obtained.

[0177] Update operation instructions refer to instructions initiated by the hospital's quality control department through the rule base management module to add, modify, or delete quality control rules. These instructions include the operation type (add / modify / delete) and the rule number (when modifying / deleting).

[0178] The new rules refer to the rules generated based on the latest medical standards and departmental requirements, including new quality control dimensions, adjusted rule logic, or newly added mapping relationships between "diagnosis / treatment / fee items and required archived materials".

[0179] Specifically, the electronic device receives update operation instructions and new rules, verifies the submitting user's permissions (only the quality control department administrator has the right to initiate this); after the permission verification is successful, the new rules are parsed, and the rule type (first to fifth categories), rule logic, applicable document type, and weight parameters are extracted. , Finally, it refers to incorporating new rules into the quality control rule base, or modifying or deleting existing rules to ensure that the rule base is consistent with medical standards and hospital requirements.

[0180] As an example, an update may include:

[0181] a. Perform updates based on operation type:

[0182] Add a new rule: Assign a unique rule number to the new rule, store it in the rule library by type, and update the rule count statistics;

[0183] Modify rules: Locate the existing rules based on the rule number, replace them with the new rules, and retain historical versions (mark the modification time and the person who modified them);

[0184] Deleting rules: Locate the original rule based on the rule number and mark it as "deleted" (logical deletion, archived);

[0185] b. Synchronize mapping table: If the new rule involves the fifth category of rules (medical record integrity), update the "Diagnosis / Treatment / Billing Items - Documents to be Archived" mapping table;

[0186] c. Algorithm optimization: Based on the weight parameters corresponding to the new rules, the weighted rule matching algorithm is incrementally trained to optimize the weight allocation and improve the accuracy of error and omission identification corresponding to the new rules.

[0187] Thus, this method supports dynamic updates of quality control rules, which can solve the problems of traditional rule bases being fixed and unable to be iterated.

[0188] Please refer to Figure 2 This application provides an electronic device that may include a processing module and a memory. The memory stores a computer program, which, when executed by the processor, enables the electronic device to perform the corresponding steps in the aforementioned method for electronic storage and management of hospital paper materials.

[0189] In this embodiment, the processor can be an integrated circuit chip with signal processing capabilities. For example, the processor can be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0190] The memory can be, but is not limited to, random access memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, etc. In this embodiment, the memory can be used to store quality control rule bases, electronic documents, etc. Of course, the memory can also be used to store programs, which the processor executes after receiving execution instructions.

[0191] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the electronic device described above can be referred to the corresponding steps in the aforementioned method, and will not be elaborated further here. Additionally, the electronic device may also include... Figure 2 More components may be included, such as a display screen.

[0192] Based on the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product. This software product can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, electronic device, or network device, etc.) to execute the methods described in the various implementation scenarios of this application.

[0193] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which includes one or more executable instructions for implementing a specified logical function. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0194] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for electronic storage and management of hospital paper materials, characterized in that, The method includes: S10 converts paper materials provided by the relevant departments of the hospital into digital images; S20, convert the digitized image into an editable electronic file; S30, using a pre-created quality control rule library, perform error and omission detection on the electronic file; S40, when there are errors or omissions in the electronic document, an error or omission prompt message is issued; S50: Receive the corrected electronic file uploaded based on the error message, and repeat step S30 based on the corrected electronic file until the preset stop condition is met to obtain the final corrected electronic file. S60 associates the final revised electronic file with the corresponding digital image and stores them in categories.

2. The method according to claim 1, characterized in that, The digital images include a first type of digital image obtained by scanning text-based materials, a second type of digital image obtained by scanning tabular materials, and a third type of digital image obtained by scanning image-based materials; Converting the digitized image into an editable electronic file includes: The first type of digitized image is converted into an editable electronic document using an OCR recognition model. The second type of digitized images are converted into editable spreadsheets using a table conversion model. The third type of digital image is converted into an image of a specified format using an image conversion model, and the image, electronic document, and spreadsheet of the specified format are treated as electronic files.

3. The method according to claim 1, characterized in that, Before converting the paper materials provided by the relevant departments of the hospital into digital images, the method further includes: Configure the first type of rule for basic information consistency checks; Configure a second type of rule for checking the completeness of required fields; Configure the third type of rule for consistency detection of diagnostic and treatment logic; Configure the fourth type of rule for format conformity checks; Configure the fifth type of rules for medical record integrity detection, and form a quality control rule base based on the first, second, third, fourth and fifth types of rules.

4. The method according to claim 1, characterized in that, The electronic documents are checked for errors and omissions using a pre-created quality control rule base, including: The electronic document is subjected to feature extraction to obtain features to be verified. The features to be verified include at least one of the following: basic information features, required field filling features, diagnosis and treatment association features, format features, and integrity verification features. From a pre-created quality control rule base, rules corresponding to the features to be verified are selected for error detection, and a preset weighted rule matching algorithm is used to calculate the total error risk score. The calculation formula for the weighted rule matching algorithm is as follows: ； In the formula, R represents the total error / omission risk score; This represents the weight of the k-th rule; This represents the number of sub-rules contained in the k-th rule category; Let be the matching degree of the i-th sub-rule in the k-th rule category; The importance weight of the i-th sub-rule in the k-th rule category; Based on the total error risk score, the error level is determined according to the preset level classification strategy. The error level includes level 0 error, which indicates no error; level 1 error, which indicates a general error; level 2 error, which indicates a serious error; and level 3 error, which indicates a fatal error.

5. The method according to claim 4, characterized in that, When errors or omissions are found in the electronic document, an error / omission warning message is issued, including: When the electronic document contains a Level 1 error indicating a general error, an error message and correction suggestion corresponding to the Level 1 error are sent to the review end. When the review end sends an instruction to accept the correction suggestion, the electronic document is corrected according to the correction suggestion to obtain the corrected electronic document. When the electronic document contains Level 2 or Level 3 errors, a corresponding error message is sent to the reviewer so that the reviewer can upload the corrected electronic document based on the error message.

6. The method according to claim 1, characterized in that, The final revised electronic files are associated with the corresponding digital images and stored in a categorized manner, including: Classification features are extracted from the last revised electronic document, including basic information features, content semantic features, diagnosis and treatment stage features, and integrity association features; The classification features are standardized to obtain a standardized classification feature vector. The standardization process includes: deduplicating the classification features, performing one-hot encoding on the text features in the classification features, and normalizing the numerical features in the classification features. The standardized basic information features are input into the trained department classification model to obtain the corresponding department classification labels; Based on the pre-established mapping relationship between diagnosis and treatment stages and keywords, the diagnosis and treatment stages corresponding to the keywords in the diagnosis and treatment stage features are determined as diagnosis and treatment stage classification labels. The standardized semantic features of the content are input into the trained file classification model to obtain file category labels. Based on the integrity association features, integrity classification labels are obtained, including a first category label for complete original data and a second category label for complete data after completion. Based on department classification labels, treatment stage classification labels, document type classification labels, and integrity classification labels, corresponding storage paths are generated. Then, according to the storage paths, the final corrected electronic files and corresponding digital images are encrypted and stored to form a database. In addition, a multi-dimensional index corresponding to the corresponding electronic files is constructed. The multi-dimensional index includes medical record number index, classification label index, creation time index, and keyword index.

7. The method according to claim 6, characterized in that, The final revised electronic file and the corresponding digital image are encrypted and stored, including: The AES-256 encryption algorithm is used to encrypt and store the final corrected electronic files and corresponding digital images, and access permissions are set according to the corresponding classification labels.

8. The method according to claim 6, characterized in that, The method further includes: Upon receiving a retrieval request for retrieving a corresponding electronic file, the system searches the database for an electronic file that matches the index information carried in the retrieval request, using this file as the target file. When the access permission corresponding to the retrieval request is greater than or equal to the access permission of the target file, reading the target file is permitted; If the access permission corresponding to the retrieval request is less than the access permission of the target file, the target file will not be read, and a corresponding prompt message will be issued.

9. The method according to claim 1, characterized in that, The method further includes: Based on the update operation command, the corresponding rules in the quality control rule base are updated using the new rules obtained.

10. An electronic device, characterized in that, The electronic device includes a processor and a memory coupled together, the memory storing a computer program that, when executed by the processor, causes the electronic device to perform the method as described in any one of claims 1 to 9.

Images