An intelligent neck pressure regulating device based on deep learning

Abstract

The present invention provides an intelligent neck pressure regulating device based on deep learning. The device includes: a learning data acquisition subsystem for acquiring local treatment pressure data of neck lymphatic leakage; a deep learning subsystem for learning the local treatment pressure data of neck lymphatic leakage based on a deep learning model to obtain a simulated pressure model; a simulated pressure parameter determination subsystem for acquiring postoperative condition information of the patient wearing the intelligent neck pressure regulator and determining simulated pressure parameters according to the simulated pressure model; and an intelligent pressure regulating subsystem for triggering a pressure regulating signal of the intelligent neck pressure regulator according to the simulated pressure parameters. The intelligent neck pressure regulating device based on deep learning of the present invention uses deep learning technology to learn local treatment pressure data of neck lymphatic leakage to obtain a simulated pressure model; determines simulated pressure parameters according to the postoperative condition information of the patient and triggers a pressure regulating control instruction of the neck pressure regulator, and the local pressure treatment efficiency of neck lymphatic leakage is higher and more intelligent.

Description

Technical Field

[0001] The present invention relates to the technical field of deep learning, and particularly relates to an intelligent neck pressure regulating device based on deep learning. Background Art

[0002] Thyroid cancer is the most common malignant tumor in the endocrine system and also a common malignant tumor in the head and neck surgery. Surgical radical neck lymph node dissection is one of the best surgical options. Chylous fistula or lymphatic fistula is a complication after neck lateral lymph node dissection. Chylous fistula or lymphatic fistula is usually caused by the rupture of the thoracic duct or lymphatic duct branches. It is especially prone to occur when separating the left jugular venous angle. Most chylous fistulas or lymphatic fistulas occur on the second day after surgery, and a few occur on the third or fourth day after surgery. The drainage volume is less than 1000 ml / d. Patients with chylous fistula or lymphatic fistula can be cured through conservative treatment, but the disadvantage is the prolonged hospital stay.

[0003] Currently, the clinical treatment methods for chylous fistula or lymphatic fistula include connecting a negative pressure drainage ball, local compression, low-fat diet, intravenous administration of somatostatin, and parenteral nutritional support treatment to achieve the effect of accelerating the patient's recovery and shortening the hospital stay. However, the above-mentioned compression operation needs to be manually performed by medical staff according to personal experience, with low processing efficiency and lack of intelligence.

[0004] In view of this, there is an urgent need for an intelligent neck pressure regulating device based on deep learning to at least solve the above deficiencies. Summary of the Invention

[0005] One of the purposes of the present invention is to provide an intelligent neck pressure regulating device based on deep learning, which uses deep learning technology to learn the local treatment compression data of neck lymphatic fistula after neck lateral lymph node dissection of the target patient to obtain a simulated compression model; then automatically determines the simulated compression parameters according to the postoperative condition information of the patient wearing the intelligent neck pressure regulator; the simulated compression parameters are converted through a pressure regulating signal conversion template to output a pressure regulating control instruction for the neck pressure regulator to control the neck pressure, without the need for medical staff to manually apply pressure, and the local compression treatment efficiency of neck lymphatic fistula is higher and more intelligent.

[0006] An intelligent neck pressure regulating device based on deep learning provided by an embodiment of the present invention includes:

[0007] A learning data acquisition subsystem, configured to acquire the local treatment compression data of neck lymphatic fistula after neck lateral lymph node dissection of the target patient by medical staff;

[0008] A deep learning subsystem, configured to learn the local treatment compression data of neck lymphatic fistula based on a deep learning model to obtain a simulated compression model;

[0009] The simulated pressurization parameter determination subsystem is used to obtain the postoperative condition information of the patient wearing the intelligent neck pressure regulator, and determine the simulated pressurization parameters according to the postoperative condition information and the simulated pressurization model;

[0010] The intelligent pressure regulation subsystem is used to determine the pressure regulation signal and perform intelligent pressure regulation control according to the simulated pressurization parameters and the pressure regulation signal conversion template of the intelligent neck pressure regulator.

[0011] Preferably, the learning data acquisition subsystem acquires the local treatment pressurization data of the neck lymphatic leakage after the cervical lymph node dissection of the target patient by medical staff, including:

[0012] Connect to the medical information source;

[0013] Through the medical information source, retrieve the local treatment pressurization data of the neck lymphatic leakage after the cervical lymph node dissection of the target patient by medical staff; among them, the local treatment pressurization data of the neck lymphatic leakage includes: the postoperative condition information of the target patient and the reference pressure parameters.

[0014] Preferably, the deep learning subsystem learns the local treatment pressurization data of the neck lymphatic leakage based on the deep learning model to obtain the simulated pressurization model, including:

[0015] Use the postoperative condition information of the target patient as the input of the ViT model and the reference pressure parameters as the output of the ViT model to train the simulated pressurization model.

[0016] Preferably, the simulated pressurization parameter determination subsystem acquires the postoperative condition information of the patient wearing the intelligent neck pressure regulator, and determines the simulated pressurization parameters according to the postoperative condition information and the simulated pressurization model, including:

[0017] Acquire the clinical data and medical image data of the patient wearing the intelligent neck pressure regulator;

[0018] Use the clinical data and medical image data as the postoperative condition information of the patient wearing it;

[0019] Use the postoperative condition information of the patient wearing it as the input of the simulated pressurization model, and obtain the simulated pressurization parameters output by the simulated pressurization model.

[0020] Preferably, the learning data acquisition subsystem retrieves the local treatment pressurization data of the neck lymphatic leakage after the cervical lymph node dissection of the target patient by medical staff through the medical information source, including:

[0021] Determine the medical item labels of the preselected patients through the medical information source;

[0022] Retrieve the lymphatic leakage treatment item labels from the medical item labels, and obtain the case data associated with the retrieved successful medical item labels;

[0023] The data types for obtaining case data; the data types include: data on the situation after dissection, data on the treatment process of lymphatic leakage, and data on the treatment effect of lymphatic leakage;

[0024] Determine whether the case data meets the data screening criteria;

[0025] If so, use the preselected patients corresponding to the relevant case data as target patients and obtain the local treatment pressure data for cervical lymphatic leakage.

[0026] Preferably, the learning data acquisition subsystem determines whether the case data meets the data screening criteria, including:

[0027] Traverse the sub-case data of each data type in the case data in sequence, and use the currently traversed sub-case data as the target sub-case data;

[0028] When the data type is data on the situation after dissection, obtain the standard situation item description scoring template; the standard situation item description scoring template includes: corresponding standard situation items and standard situation item specification description factors;

[0029] Parse the target sub-case data to obtain the recorded situation items and recorded situation item description factors;

[0030] Obtain the missing situation items of the recorded situation items corresponding to the standard situation items;

[0031] Obtain the missing situation scoring library;

[0032] Determine the missing situation score according to the missing situation scoring library and the missing situation items;

[0033] Obtain the matching situation items of the recorded situation items corresponding to the standard situation items;

[0034] Calculate the factor cosine value of the recorded situation item description factor of the matching situation item and the corresponding matching standard situation item and standard situation item specification description factor;

[0035] Obtain the matching situation scoring library;

[0036] Determine the matching situation score according to the matching situation scoring library and the factor cosine value;

[0037] Sum up the missing situation score and the matching situation score to obtain the first score value of the target sub-case data;

[0038] When the data type is data on the treatment process of lymphatic leakage, obtain the evaluation records during the treatment process;

[0039] Determine the evaluation type of the evaluation record, and the evaluation types include: active evaluation and passive evaluation;

[0040] Obtain the evaluation record scoring strategy corresponding to the evaluation type;

[0041] According to the evaluation record scoring strategy, determine the treatment process scores corresponding to the evaluation records of different evaluation types;

[0042] According to the preset evaluation weight values corresponding to the evaluation types and the treatment process scores, calculate the second score value of the target case sub-data;

[0043] When the data type is the lymphatic fistula treatment effect data, obtain the effect score of the target case sub-data and use it as the third score value;

[0044] When the case sub-data of each data type in the case data has been traversed, accumulate and calculate the first score value, the second score value, and the third score value to obtain the data screening value of the case data;

[0045] If the data screening value is greater than or equal to the preset data screening value threshold, determine that the case data meets the data screening standard.

[0046] Preferably, the learning data acquisition subsystem determines the treatment process scores corresponding to the evaluation records of different evaluation types according to the evaluation record scoring strategy, including:

[0047] When the evaluation type is an active evaluation, parse the evaluation record to obtain the patient evaluation words;

[0048] Obtain the preset discomfort degree value determination library;

[0049] According to the patient evaluation words and the discomfort degree value determination library, determine the discomfort degree value of the evaluation record of the active evaluation;

[0050] Based on the conversion relationship between the preset discomfort degree value and the treatment process score, determine the treatment process score;

[0051] When the evaluation type is a passive evaluation, parse the evaluation record to obtain the medical staff evaluation words;

[0052] Based on the lymphatic fistula treatment knowledge records in the lymphatic fistula treatment knowledge base, train the lymphatic fistula treatment rationality analysis model;

[0053] Characterize the medical staff evaluation words to obtain evaluation word features, and the evaluation word features include: the meaning of the lymphatic fistula treatment process evaluated by the evaluation words, the evaluation meaning, the reverse evaluation meaning, and the target meaning between the evaluation word corresponding to the evaluation meaning and the evaluation word corresponding to the reverse evaluation meaning;

[0054] Input the evaluation word features into the lymphatic fistula treatment rationality analysis model to obtain the rationality degree value, and use the rationality degree value as the treatment process score.

[0055] An intelligent neck pressure regulating device based on deep learning provided by an embodiment of the present invention further includes:

[0056] An offset detection subsystem for detecting the offset of the pressing position during the operation of the intelligent neck pressure regulator.

[0057] Preferably, the offset detection subsystem detects the offset of the pressing position during the operation of the intelligent neck pressure regulator, including:

[0058] During the operation of the intelligent neck pressure regulator, obtain the dynamic information of the intelligent neck pressure regulator;

[0059] If the dynamic information is successfully obtained, perform the offset detection of the pressing position.

[0060] Preferably, if the dynamic information is successfully obtained, the offset detection subsystem performs the offset detection of the pressing position, including:

[0061] If the dynamic information is successfully obtained, obtain the temperature induction data of the pressurized area within the preset duration of the intelligent neck pressure regulator;

[0062] According to the temperature induction data, draw the temperature change gradient map of different pressurized areas;

[0063] Calculate the gradient map similarity of the temperature change gradient maps of different pressurized areas;

[0064] If the gradient map similarity is greater than or equal to the preset gradient map similarity threshold, send an offset reminder to the wearer.

[0065] A method for intelligent neck pressure regulation based on deep learning provided by an embodiment of the present invention includes:

[0066] Step 1: Obtain the local treatment pressurization data of the neck lymphatic leakage after the lymph node dissection in the neck side area of the target patient by medical staff;

[0067] Step 2: Learn the local treatment pressurization data of the neck lymphatic leakage based on a deep learning model to obtain a simulated pressurization model;

[0068] Step 3: Obtain the postoperative condition information of the patient wearing the intelligent neck pressure regulator, and determine the simulated pressurization parameters according to the postoperative condition information and the simulated pressurization model;

[0069] Step 4: Determine the pressure regulation signal according to the simulated pressurization parameters and the pressure regulation signal conversion template of the intelligent neck pressure regulator, and perform intelligent pressure regulation control.

[0070] The beneficial effects of the present invention are:

[0071] The present invention utilizes deep learning technology to learn the local treatment pressurization data of neck lymphatic leakage after neck lateral region lymph node dissection of the target patient to obtain a simulated pressurization model; then automatically determines the simulated pressurization parameters according to the postoperative condition information of the patient wearing the intelligent neck pressure regulator; the simulated pressurization parameters are converted through a pressure regulation signal conversion template to output a pressure regulation control instruction for the neck pressure regulator to control the neck pressure, without manual pressurization by medical staff, and the local pressurization treatment efficiency of neck lymphatic leakage is higher and more intelligent.

[0072] Other features and advantages of the present invention will be described in the following specification, and part of them will be obvious from the specification or understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structure specifically pointed out in this application document.

[0073] The technical solution of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0074] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0075] Figure 1 It is a schematic diagram of an intelligent neck pressure regulating device based on deep learning in an embodiment of the present invention.

[0076] Figure 2 It is a schematic diagram of an intelligent neck pressure regulating method based on deep learning in an embodiment of the present invention. Detailed Embodiments

[0077] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0078] The embodiment of the present invention provides an intelligent neck pressure regulating device based on deep learning, as Figure 1 shown, including:

[0079] A learning data acquisition subsystem 1, configured to acquire the local treatment pressurization data of neck lymphatic leakage after neck lateral region lymph node dissection of the target patient;

[0080] Among them, the learning data acquisition subsystem acquires the local treatment pressurization data of neck lymphatic leakage after neck lateral region lymph node dissection of the target patient, including:

[0081] Dock medical information sources; among them, the medical information sources are data source nodes in the medical information system, including hospital information systems, electronic medical record systems, and picture archiving and communication systems. Through these nodes, medical data of patients can be retrieved and obtained;

[0082] Through the medical information sources, retrieve the local treatment pressure data for neck lymphatic leakage after neck side lymph node dissection of the target patient by medical staff; among them, the local treatment pressure data for neck lymphatic leakage includes: postoperative condition information of the target patient and reference pressure parameters; among them, the postoperative condition information includes: medical imaging data of the target patient (such as: neck ultrasound images of the target patient) and clinical data (such as: lymphatic leakage occurrence time, drainage volume, grading of lymphatic leakage), etc.; the reference pressure parameters are: specific parameters for local pressure treatment of neck lymphatic leakage after neck side lymph node dissection of the target patient, such as: pressure application position, pressure application time, and pressure application pressure, etc.;

[0083] The deep learning subsystem 2 is used to learn the local treatment pressure data for neck lymphatic leakage based on the deep learning model to obtain a simulated pressure model;

[0084] Among them, the deep learning subsystem learns the local treatment pressure data for neck lymphatic leakage based on the deep learning model to obtain a simulated pressure model, including:

[0085] Take the postoperative condition information of the target patient as the input of the ViT model and the reference pressure parameters of the target patient as the output of the ViT model to train the simulated pressure model; among them, the ViT model is the Vision Transformer model;

[0086] The simulated pressure parameter determination subsystem 3 is used to obtain the postoperative condition information of the patient wearing the intelligent neck pressure regulator, and determine the simulated pressure parameters according to the postoperative condition information and the simulated pressure model;

[0087] Among them, the simulated pressure parameter determination subsystem obtains the postoperative condition information of the patient wearing the intelligent neck pressure regulator, and determines the simulated pressure parameters according to the postoperative condition information and the simulated pressure model, including:

[0088] Obtain the clinical data and medical imaging data of the patient wearing the intelligent neck pressure regulator;

[0089] Take the clinical data and medical imaging data as the postoperative condition information of the patient wearing;

[0090] Take the postoperative condition information of the patient wearing as the input of the simulated pressure model, and obtain the simulated pressure parameters output by the simulated pressure model;

[0091] The intelligent pressure regulating subsystem 4 is used to determine the pressure regulating signal and perform intelligent pressure regulating control according to the simulated pressurization parameters and the pressure regulating signal conversion template of the intelligent neck pressure regulating instrument. Among them, the pressure regulating signal conversion template is: a template for obtaining the pressure regulating control instruction (pressure regulating signal) of the intelligent neck pressure regulating instrument by comparing the pressure parameter that needs to be triggered by the intelligent neck pressure regulating instrument. After the pressure regulating signal is triggered, the intelligent neck pressure regulating instrument will automatically apply the pressure corresponding to the pressure parameter to the patient wearing it through its preset pressurization area.

[0092] The working principle and beneficial effects of the above technical solution are:

[0093] The present invention utilizes deep learning technology to learn the local treatment pressurization data of cervical lymph node dissection in the lateral cervical area of the target patient by medical staff to obtain a simulated pressurization model; then automatically determines the simulated pressurization parameters according to the postoperative condition information of the patient wearing the intelligent neck pressure regulator; the simulated pressurization parameters are converted by a pressure regulation signal conversion template to output the pressure regulation control instructions of the neck pressure regulator to control the neck pressure, without the need for manual pressurization by medical staff, and the local pressurization treatment of cervical lymph node leakage is more efficient and intelligent.

[0094] In one embodiment, the learning data acquisition subsystem retrieves the local treatment and compression data of cervical lymph node leakage after lateral cervical lymph node dissection of the target patient by medical staff through the medical information source, including:

[0095] Determine the medical project label of the pre-selected patient through the medical information source; wherein the medical information source is: various data sources containing medical related information, such as: the hospital's electronic medical record system, medical literature database and clinical research data;

[0096] Retrieve lymphatic leakage treatment project tags from medical project tags, and obtain case data associated with the successfully retrieved medical project tags; wherein the medical project tags are keywords for classifying and identifying medical projects; the lymphatic leakage treatment project tags are project tags related to lymphatic leakage treatment, such as "lymphatic leakage treatment"; and the case data are data related to lymphatic leakage treatment of patients of lymphatic leakage treatment projects recorded in the medical information source;

[0097] Data types for obtaining case data; the data types include: data on the situation after dissection, data on the treatment process of lymphatic leakage, and data on the treatment effect of lymphatic leakage; among them, the data on the situation after dissection is: the situation of the preselected patients after neck dissection, such as: neck CT scan images, basic information of the patients (patient age, gender, weight, and historical medical records, etc.); the data on the treatment process of lymphatic leakage is: data during the treatment of lymphatic leakage, such as: information on the compression site, intensity, and time recorded by medical staff when using elastic bandages for compression dressing to treat neck lymphatic leakage; the data on the treatment effect of lymphatic leakage is: data reflecting the treatment effect of lymphatic leakage, such as: changes in drainage volume after treatment, symptom relief, and complication incidence, etc.;

[0098] Judge whether the case data meets the data screening criteria;

[0099] If so, take the preselected patient corresponding to the corresponding case data as the target patient and obtain the local treatment compression data for neck lymphatic leakage;

[0100] Among them, the learning data acquisition subsystem judges whether the case data meets the data screening criteria, including:

[0101] Traverse each sub-case data of each data type in the case data in turn, and take the currently traversed sub-case data as the target sub-case data;

[0102] When the data type is data on the situation after dissection, obtain the standard situation item description scoring template; the standard situation item description scoring template includes: one-to-one corresponding standard situation items and standard situation item specification description factors; among them, the standard situation item is: the postoperative situation to be described, such as: "whether it includes neck CT images", "basic information of the patient"; the standard situation item specification description factor is: the characteristic description vector for standardizing the description of the standard situation item, and the characteristic description is, for example, the neck CT image in the summary list of the post-dissection situation of the preselected patient is marked as "yes", and the source library associated with the neck CT image;

[0103] Analyze the target sub-case data to obtain the recorded situation items and recorded situation item description factors; among them, the recorded situation item is: the item of the post-dissection situation information of the preselected patient recorded in the target sub-case data; the recorded situation item description factor is: the characteristic description vector of the recorded situation item;

[0104] Obtain the missing situation items of the recorded situation items corresponding to the standard situation items; among them, the missing situation item is: the standard situation item that does not exist in the recorded situation item;

[0105] Obtain a missing situation scoring library; among them, the missing situation scoring library includes scoring rules corresponding to multiple missing situations, and the scoring rules corresponding to the missing situations are preset manually. For example: missing the neck CT image information item, the score for determining the missing situation is 40; missing the patient's age information item, the score for the corresponding missing situation is 80;

[0106] Determine the missing situation score according to the missing situation scoring library and the missing situation items;

[0107] Obtain the matching situation item of the record situation item corresponding to the standard situation item; among them, the matching situation item is: the standard situation item existing in the record situation item;

[0108] Calculate the factor cosine value of the record situation item description factor of the matching situation item and the corresponding standard situation item and the standard situation item specification description factor; among them, the factor cosine value is: the cosine similarity of the record situation item description factor and the standard situation item specification description factor;

[0109] Obtain a matching situation scoring library; among them, the matching situation scoring library includes quantization rules for quantifying the cosine similarity corresponding to different matching situations into matching situation scores, which conform to the rule that the greater the cosine similarity, the higher the matching situation score. The specific conversion ratio relationship is set manually according to the type of postoperative cleaning situation information corresponding to the matching situation;

[0110] Determine the matching situation score according to the matching situation scoring library and the factor cosine value;

[0111] Sum up the missing situation score and the matching situation score to obtain the first score value of the target case sub-data;

[0112] When the data type is lymphatic fistula treatment process data, obtain the evaluation records during the treatment process; among them, the evaluation records include: the evaluation of the treatment process by the medical staff who treat the lymphatic fistula of the preselected patient and the feedback evaluation of the preselected patient himself;

[0113] Determine the evaluation type of the evaluation record. The evaluation types include: active evaluation and passive evaluation; among them, the active evaluation is: the patient's own evaluation of the lymphatic fistula treatment process; the passive evaluation is: the evaluation of the lymphatic fistula treatment process by the medical staff who treat the preselected patient for the lymphatic fistula;

[0114] Obtain the evaluation record scoring strategy corresponding to the evaluation type; among them, the evaluation record scoring strategy is: a plan for evaluating the lymphatic fistula treatment process based on the content of evaluation records of different evaluation types. For example: for the active evaluation type, determine the discomfort degree value of the patient according to the discomfort adjectives reported by the patient. The lower the discomfort degree value, the higher the corresponding treatment process score; for the passive evaluation type, determine the reasonable degree value of the evaluation semantics according to the evaluation semantics of medical staff and the preset lymphatic fistula treatment knowledge base. The higher the reasonable degree value, the higher the corresponding treatment process score;

[0115] According to the evaluation record scoring strategy, determine the treatment process scores corresponding to the evaluation records of different evaluation types;

[0116] According to the evaluation weight corresponding to the evaluation type and the treatment process score, calculate the second score value of the target case sub-data; among them, the second score value is the result obtained by multiplying and summing the evaluation weight and the corresponding treatment process score. The evaluation weight corresponding to the active evaluation is less than the evaluation weight corresponding to the passive evaluation;

[0117] When the data type is lymphatic fistula treatment effect data, obtain the effect score of the target case sub-data and use it as the third score value; among them, the effect score is obtained by quantifying based on the preset effect evaluation semantics - effect value quantization rule according to the effect evaluation semantics;

[0118] After traversing all the case sub-data of each data type in the case data, accumulate and calculate the first score value, the second score value and the third score value to obtain the data screening value of the case data;

[0119] If the data screening value is greater than or equal to the preset data screening value threshold, it is determined that the case data meets the data screening standard. Among them, the preset data screening value threshold is set manually in advance.

[0120] The working principle and beneficial effects of the above technical solution are:

[0121] When obtaining the learning data of the deep learning model through the medical information source, not all the treatment pressure data of patients with local treatment of cervical lymphatic fistula are available. For example: relevant data such as incomplete postoperative information of patients, inappropriate pressurization strategies, and poor effects after local treatment of lymphatic fistula. Therefore, it is necessary to screen the data obtained from the medical information source;

[0122] First, determine the medical project label, retrieve the lymphatic fistula treatment project label in the medical project label, and obtain the case data under the label; determine the data type of the case data, specifically including: data on the situation after dissection, lymphatic fistula treatment process data, and lymphatic fistula treatment effect data, and judge whether the case data meets the data screening standard. Among them,

[0123] Determining whether the case data meets the data screening criteria includes: traversing the sub-case data of each data type in the case data in sequence to obtain the target sub-case data currently being traversed. When the data type is the data of the situation after radical dissection, introduce the description scoring template for the standard situation items; the description scoring template for the standard situation items includes the postoperative situations to be described (standard situation items) and their corresponding characteristic description vectors (standard situation item specification description factors) for the standardized description of the standard situation items; compare the recorded situation items obtained from the target sub-case data with the standard situation items to determine the missing situation items and the matching situation items. The missing situation items determine the missing situation score based on the missing situation scoring library; the matching situation score is determined according to the factor cosine value of the recorded situation item description factor of the calculated matching situation item and the standard situation item specification description factor of the corresponding matched standard situation item and the matching situation scoring library. Sum the missing situation score and the matching situation score to obtain the first score value of the target sub-case data;

[0124] When the data type is the data of the treatment process of lymphatic fistula, introduce the evaluation records during the treatment process. The evaluation records include the evaluations actively issued by the patient himself / herself on the treatment process of lymphatic fistula and the evaluations passively issued by the medical staff on the treatment process of lymphatic fistula; according to the different evaluation types, obtain the corresponding evaluation record scoring strategies and assign treatment process scores to the corresponding evaluation records; multiply the evaluation weights preset for the evaluation types and the treatment process scores respectively and then sum them to obtain the second score value;

[0125] Finally, when the data type is the data of the treatment effect of lymphatic fistula, quantify the semantic of the effect evaluation to obtain the effect score. The better the treatment effect represented by the semantic of the effect evaluation, the higher the third score value;

[0126] Sum the first score value, the second score value, and the third score value associated with the case data to obtain the data screening value, and screen out the case data that meets the data screening criteria and whose data screening value is greater than or equal to the preset data screening value threshold, improving the screening quality of the learning data.

[0127] In one embodiment, the learning data acquisition subsystem determines the treatment process scores corresponding to the evaluation records of different evaluation types according to the evaluation record scoring strategy, including:

[0128] When the evaluation type is the active evaluation, parse the evaluation record to obtain the patient evaluation words; among them, the patient evaluation words are obtained by performing word segmentation on the patient evaluation record;

[0129] Obtain the preset discomfort degree value determination library; among them, the discomfort degree value determination library presets multiple target words and discomfort degree values in one-to-one correspondence, which are pre-configured manually;

[0130] According to the patient evaluation words and the discomfort degree value determination library, determine the discomfort degree value of the evaluation record of the active evaluation;

[0131] Determine the treatment process score based on the conversion relationship between the preset discomfort degree value and the treatment process score; wherein, the conversion relationship conforms to an inverse relationship, and the specific proportional relationship is set manually according to requirements;

[0132] When the evaluation type is passive evaluation, parse the evaluation record to obtain the medical staff evaluation words; wherein, the medical staff evaluation words are obtained by performing word segmentation on the medical staff evaluation record.

[0133] Train a lymphatic fistula treatment rationality analysis model based on the lymphatic fistula treatment knowledge records in the lymphatic fistula treatment knowledge base; wherein, the lymphatic fistula treatment rationality analysis model is an AI model that uses a neural network model to train the lymphatic fistula treatment knowledge records and outputs the rationality degree value of the knowledge description based on the input knowledge description.

[0134] Characterize the medical staff evaluation words to obtain evaluation word features, and the evaluation word features include: the meaning of the lymphatic fistula treatment process described by the evaluation words, the evaluation meaning, the reverse evaluation meaning, and the target meaning between the evaluation word corresponding to the evaluation meaning and the evaluation word corresponding to the reverse evaluation meaning; wherein, the meaning of the lymphatic fistula treatment process is: the meaning of the lymphatic fistula treatment process description words in the medical staff evaluation words; the evaluation meaning is: the meaning in the medical staff evaluation words that represents good or bad evaluation, such as: "good", "bad", "right", "wrong", etc.; the reverse evaluation meaning is: the medical staff evaluation word related to the medical staff evaluation word corresponding to the evaluation meaning and representing good or bad evaluation, and the meaning of this word is opposite to the associated evaluation meaning, and the association means that there is no other medical staff evaluation word representing good or bad evaluation between the two medical staff evaluation words, and the number of words between the two medical staff evaluation words is less than the preset word number threshold; the target meaning is: the meaning of the medical staff evaluation words between the evaluation word corresponding to the evaluation meaning and the evaluation word corresponding to the reverse evaluation meaning.

[0135] Input the evaluation word features into the lymphatic fistula treatment rationality analysis model to obtain the rationality degree value, and use the rationality degree value as the treatment process score.

[0136] The working principle and beneficial effects of the above technical solution are:

[0137] The present invention introduces an evaluation type. When the evaluation type is active evaluation, a discomfort degree value determination library corresponding to the patient's evaluation is introduced. According to the patient's evaluation words and the discomfort degree value determination library, the discomfort degree value of the evaluation record of the active evaluation is determined. According to the inverse conversion relationship between the discomfort degree value and the treatment process score, the treatment process score is determined. When the evaluation type is passive evaluation, the lymphatic leakage treatment knowledge records in the lymphatic leakage treatment knowledge base are introduced to train the lymphatic leakage treatment rationality analysis model, and the medical staff's evaluation words are characterized. The target words between the words corresponding to the lymphatic leakage treatment process, the evaluation words corresponding to the lymphatic leakage treatment process, the reverse evaluation words associated with the evaluation words, and the evaluation words corresponding to the evaluation words to the evaluation words corresponding to the reverse evaluation words are obtained. This feature extraction setting takes into account the treatment situation where when the lymphatic leakage treatment process is inappropriate, the medical staff makes a negative evaluation and then corrects it. It should be noted that there are large individual physiological differences, and it is a common situation to adjust according to the process feedback during the pressurization process. The present invention takes into account the adjustment situation and extracts the target words (adjustment situation words) between the evaluation words corresponding to the evaluation words to the evaluation words corresponding to the reverse evaluation words and inputs them into the lymphatic leakage treatment rationality analysis model together, avoiding unreasonable treatment process scores for the lymphatic leakage treatment process after identifying negative evaluation words and thus blindly eliminating learning data, and improving the learning efficiency.

[0138] An embodiment of the present invention provides an intelligent neck pressure regulating device based on deep learning, further including:

[0139] An offset detection subsystem, configured to perform pressing position offset detection during the operation of the intelligent neck pressure regulator;

[0140] Among them, the offset detection subsystem performs pressing position offset detection during the operation of the intelligent neck pressure regulator, including:

[0141] During the operation of the intelligent neck pressure regulator, the dynamic information of the intelligent neck pressure regulator is obtained; among them, the dynamic information is obtained according to the gyroscope inside the intelligent neck pressure regulator. When the gyroscope data mutates, it is determined that the dynamic information acquisition is successful;

[0142] If the dynamic information acquisition is successful, pressing position offset detection is performed;

[0143] Among them, if the dynamic information acquisition is successful, the offset detection subsystem performs pressing position offset detection, including:

[0144] If the dynamic information acquisition is successful, the temperature induction data of the pressurized area within a preset duration of the intelligent neck pressure regulator is obtained; the preset duration is: the preset time length before the dynamic information acquisition is successful and the preset time length after the dynamic information acquisition is successful, and the preset time length is set manually in advance, for example: 10 seconds;

[0145] According to the temperature sensing data, a temperature change gradient diagram of different pressurized areas is drawn; wherein, the temperature change gradient diagram is a visualization chart representing the temperature distribution change in the pressurized area within a preset time period;

[0146] Calculate the gradient diagram similarity of the temperature change gradient diagrams of different pressurized areas; wherein, the gradient diagram similarity is the degree of similarity of image features;

[0147] If the gradient diagram similarity is greater than or equal to the preset gradient diagram similarity threshold, an offset reminder is sent to the wearer. Wherein, the preset gradient diagram similarity threshold is set manually in advance.

[0148] The working principle and beneficial effects of the above technical solution are as follows:

[0149] The present invention detects the dynamic information of the intelligent neck pressure regulator. When the dynamic information is successfully obtained, the temperature sensing data of the pressurized area within the preset time period is introduced, and the temperature sensing data is visualized to obtain the temperature change gradient diagram. Since generally the intelligent neck pressure regulator deflects in one direction, and the temperature of the pressurized skin area before deflection is higher than that after deflection, therefore, when it is detected that the intelligent neck pressure regulator generates relative movement and the temperature change characteristics of different pressurized areas are similar, it indicates that the instrument has deflected, and the wearer is reminded to adjust in time to avoid the delay of local pressure treatment.

[0150] An embodiment of the present invention provides an intelligent neck pressure regulation method based on deep learning, as Figure 2 shown, including:

[0151] Step 1: Obtain the local treatment pressure data of the neck lymphatic leakage after the lymph node dissection in the neck side area of the target patient by the medical staff;

[0152] Step 2: Learn the local treatment pressure data of the neck lymphatic leakage based on the deep learning model to obtain a simulated pressure model;

[0153] Step 3: Obtain the postoperative condition information of the wearer of the intelligent neck pressure regulator, and determine the simulated pressure parameters according to the postoperative condition information and the simulated pressure model;

[0154] Step 4: Determine the pressure regulation signal according to the simulated pressure parameters and the pressure regulation signal conversion template of the intelligent neck pressure regulator, and perform intelligent pressure regulation control.

[0155] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. An intelligent neck pressure regulating device based on deep learning, characterized in that, include: A learning data acquisition subsystem is used to obtain the local treatment and compression data of cervical lymph node leakage after lateral cervical lymph node dissection of target patients by medical staff; The local treatment pressurization data for cervical lymphatic leakage includes: the postoperative condition information and reference pressure parameters of the target patient; the postoperative condition information of the target patient includes: the medical imaging data and clinical data of the target patient, the medical imaging data of the target patient is the target patient's neck ultrasound image, and the clinical data of the target patient is the target patient's lymphatic leakage occurrence time, drainage volume and lymphatic leakage grade; the reference pressure parameters are specific parameters for local pressurization treatment of cervical lymphatic leakage after lateral cervical lymph node dissection of the target patient, and the specific parameters are: pressurization position, pressurization time and pressurization pressure; A deep learning subsystem, used to learn the local treatment pressurization data of cervical lymphatic leakage based on a deep learning model and obtain a simulated pressurization model; A simulated pressurization parameter determination subsystem is used to obtain postoperative condition information of the patient wearing the intelligent neck pressure regulator, and determine the simulated pressurization parameters according to the postoperative condition information and the simulated pressurization model; An intelligent pressure regulating subsystem, used to determine the pressure regulating signal and perform intelligent pressure regulating control according to the simulated pressurization parameters and the pressure regulating signal conversion template of the intelligent neck pressure regulating instrument; The simulated pressurization parameter determination subsystem obtains the postoperative condition information of the patient wearing the intelligent neck pressure regulator, and determines the simulated pressurization parameters according to the postoperative condition information and the simulated pressurization model, including: Obtain clinical data and medical imaging data of patients wearing the intelligent neck pressure regulator; Use clinical data and medical imaging data as postoperative condition information for patients wearing the device; The postoperative condition information of the patient wearing the device is used as the input of the simulated pressurization model to obtain the simulated pressurization parameters output by the simulated pressurization model.

2. The intelligent neck pressure regulating device based on deep learning according to claim 1, wherein The learning data acquisition subsystem acquires the local treatment and compression data of neck lymph node leakage after lateral neck lymph node dissection of the target patient by medical staff, including: Connect with medical information sources; Through the medical information source, the local treatment and pressurization data of cervical lymph node leakage after lateral cervical lymph node dissection performed by medical staff on the target patient are retrieved; wherein, the local treatment and pressurization data of cervical lymph node leakage include: the postoperative condition information and reference pressure parameters of the target patient.

3. The intelligent neck pressure regulating device based on deep learning according to claim 1, wherein, The deep learning subsystem learns the local treatment pressurization data of cervical lymphatic leakage based on the deep learning model to obtain a simulated pressurization model, including: The postoperative condition information of the target patient is used as the input of the ViT model, and the reference pressure parameters are used as the output of the ViT model to train the simulated pressurization model.

4. The intelligent neck pressure regulating device based on deep learning according to claim 2, wherein, The learning data acquisition subsystem retrieves the local treatment and compression data of cervical lymph node leakage after lateral cervical lymph node dissection of the target patient by medical staff through medical information sources, including: Determine the medical item label of the pre-selected patient through the medical information source; Retrieve lymph node leakage treatment project tags from medical project tags, and obtain case data associated with the successfully retrieved medical project tags; The data types of case data are obtained; the data types include: post-clearance condition data, lymphatic leakage treatment process data, and lymphatic leakage treatment effect data; Determine whether the case data meets the data screening criteria; If so, use the preselected patient corresponding to the corresponding case data as the target patient and obtain the local treatment pressure data for cervical lymphatic fistula.

5. The intelligent neck pressure regulating device based on deep learning according to claim 4, characterized in that, The learning data acquisition subsystem determines whether the case data meets the data screening criteria, including: Traverse the sub-case data of each data type in the case data in sequence, and use the currently traversed sub-case data as the target sub-case data; When the data type is the data of the situation after dissection, obtain the description scoring template of the standard situation items; According to the description scoring template of the standard situation items, obtain the first score value of the target sub-case data; When the data type is the data of the lymphatic fistula treatment process, obtain the evaluation records during the treatment process; Determine the evaluation type of the evaluation records, and the evaluation types include: active evaluation and passive evaluation; Obtain the evaluation record scoring strategy corresponding to the evaluation type; According to the evaluation record scoring strategy, determine the treatment process scores corresponding to the evaluation records of different evaluation types; According to the preset evaluation weight corresponding to the evaluation type and the treatment process score, calculate the second score value of the target sub-case data; When the data type is the data of the lymphatic fistula treatment effect, obtain the effect score of the target sub-case data and use it as the third score value; After traversing all the sub-case data of each data type in the case data, accumulate and calculate the first score value, the second score value and the third score value to obtain the data screening value of the case data; If the data screening value is greater than or equal to the preset data screening value threshold, it is determined that the case data meets the data screening criteria.

6. The intelligent neck pressure regulating device based on deep learning according to claim 1, characterized in that It also includes: The offset detection subsystem is used to detect the offset of the pressing position during the operation of the intelligent neck pressure regulator.

7. The intelligent neck pressure regulating device based on deep learning according to claim 6, characterized in that, The offset detection subsystem detects the offset of the pressing position during the operation of the intelligent neck pressure regulator, including: During the operation of the intelligent neck pressure regulator, obtain the dynamic information of the intelligent neck pressure regulator; If the dynamic information is successfully obtained, perform the offset detection of the pressing position.

8. The intelligent neck pressure regulating device based on deep learning according to claim 7, characterized in that, If the dynamic information is successfully obtained, the offset detection subsystem performs the offset detection of the pressing position, including: If the dynamic information is successfully obtained, obtain the temperature sensing data of the pressurized area of the intelligent neck pressure regulator within the preset duration; According to the temperature sensing data, draw the temperature change gradient map of different pressurized areas; Calculate the gradient map similarity of the temperature change gradient maps of different pressurized areas; If the gradient map similarity is greater than or equal to the preset gradient map similarity threshold, send an offset reminder to the wearer.

9. An intelligent neck pressure regulation method based on deep learning, characterized in that, It includes: Step 1: Obtain the local treatment pressure data for cervical lymphatic fistula after the lymph node dissection in the neck side area of the target patient by medical staff; The local treatment pressure data for cervical lymphatic fistula includes: the postoperative situation information of the target patient and the reference pressure parameters; the postoperative situation information of the target patient includes: the medical imaging data of the target patient and the clinical data of the target patient, the medical imaging data of the target patient is the neck ultrasound image of the target patient, and the clinical data of the target patient is the lymphatic fistula occurrence time, drainage volume and the grading of the lymphatic fistula of the target patient; the reference pressure parameters are the specific parameters for the local pressure treatment of the cervical lymphatic fistula after the lymph node dissection in the neck side area of the target patient, and the specific parameters are: the pressing position, the pressing time and the pressing pressure; Step 2: Learn the local treatment pressurization data of neck lymphatic fistula based on the deep learning model to obtain a simulated pressurization model; Step 3: Obtain the postoperative condition information of the patients wearing the intelligent neck pressure regulator, and determine the simulated pressurization parameters according to the postoperative condition information and the simulated pressurization model; Step 4: Determine the pressure regulation signal and perform intelligent pressure regulation control according to the simulated pressurization parameters and the pressure regulation signal conversion template of the intelligent neck pressure regulator; The steps for determining the simulated pressurization parameters are as follows: Obtain the clinical data and medical image data of the patients wearing the intelligent neck pressure regulator; Use the clinical data and medical image data as the postoperative condition information of the patients wearing it; Use the postoperative condition information of the patients wearing it as the input of the simulated pressurization model, and obtain the simulated pressurization parameters output by the simulated pressurization model.

Images