Physical examination center data management method and system based on artificial intelligence

Abstract

The invention relates to the technical field of data management, and discloses a physical examination center data management method and system based on artificial intelligence, and the method comprises the steps: obtaining a multi-dimensional data set; performing normalization and vector conversion according to the multi-dimensional data set, and fusing features to obtain a multi-dimensional feature fusion matrix; slicing the matrix to obtain time slice data; counting the variation amplitude of the slice data, and if the variation amplitude exceeds an amplitude threshold, marking the slice data as abnormal and identifying risk signal distribution to obtain an abnormal distribution feature set; classifying the abnormal data and establishing a risk level evaluation system to obtain risk quantitative indexes; performing early warning level judgment on the risk quantitative index, and generating and obtaining a personalized prevention suggestion according to the early warning level; monitoring the execution effect of the prevention suggestion, comparing the change condition of the diet habits associated with the nighttime blood pressure fluctuation data before and after prevention, judging the effectiveness of the prevention suggestion, and performing adjustment to form a closed-loop precise physical examination health management system. According to the method, intelligent management of data is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data management, and in particular to a physical examination center data management method and system based on artificial intelligence. BACKGROUND

[0002] At present, with the rapid development of artificial intelligence technology and its wide application in various industries, the medical and health field has ushered in a revolutionary change. The application of artificial intelligence in physical examination center data management is becoming an important direction in the medical and health field, which improves the efficiency of disease prevention and diagnosis through intelligent means, and has key significance for promoting precision medicine.

[0003] In one prior art, a physical examination information management system is adopted, which collects physical examination data of patients through the hospital internal network, including basic personal information (age, gender, etc.), medical history, lifestyle questionnaire and various physical examination indexes (such as blood pressure, blood sugar, blood lipid, etc.), and stores them in a structured table form in a relational database. However, this system can only simply arrange and combine structured data (such as age, height, weight, and various physiological index values), and for multi-dimensional data (such as medical history description and lifestyle narrative), it is only simply stored in a column of the electronic table in the form of text, lacking effective integration means and dynamic analysis capability, causing information isolation, and it is difficult to identify long-term health trends or abnormal patterns.

[0004] In summary, the prior art has the problem of insufficient information integration capability. SUMMARY

[0005] The present application provides a physical examination center data management method and system based on artificial intelligence to solve the problem of insufficient information integration capability in the prior art.

[0006] In a first aspect, to solve the above technical problems, the present application provides a physical examination center data management method based on artificial intelligence, comprising: obtaining a multi-dimensional data set, the multi-dimensional data set containing night blood pressure fluctuation data; performing numerical normalization processing and vector conversion on the multi-dimensional data set, and synthesizing a feature matrix to obtain a multi-dimensional feature fusion matrix; performing time slicing processing of different granularities on the multi-dimensional feature fusion matrix to obtain time slicing data; statistically analyzing the change amplitude of the time slicing data, and if the change amplitude exceeds a preset change amplitude threshold, marking the time slicing data as abnormal and identifying potential risk signal distribution to obtain an abnormal distribution feature set; quantitatively scoring the abnormal data in the abnormal distribution feature set to obtain a risk quantitative index; The risk quantitative index is subjected to early warning level judgment to obtain a warning level, and a corresponding personalized prevention suggestion is generated according to the warning level; The execution effect of the personalized prevention suggestion is continuously monitored, the night blood pressure fluctuation data before and after prevention are compared, the effectiveness of the prevention suggestion is judged, and a closed-loop precise physical examination health management system is formed.

[0007] In an optional implementation, the numerical normalization processing and vector conversion are performed on the multi-dimensional data set, and a feature matrix is synthesized to obtain a multi-dimensional feature fusion matrix, including: The multi-dimensional data set is subjected to format conversion to obtain a first data set; The night blood pressure fluctuation data in the first data set is subjected to standardization processing to obtain a night blood pressure fluctuation data set; The food types and time information of the eating habits in the first data set are extracted to obtain an eating habit text; The eating habit text is converted into a vector representation to obtain an eating habit vector; The night blood pressure fluctuation data set and the eating habit vector representation are integrated to obtain a multi-dimensional feature fusion matrix.

[0008] In an optional implementation, the change amplitude of the time slice data is counted, and if the change amplitude exceeds a preset change amplitude threshold, the time slice data is subjected to deep analysis to identify a potential risk signal distribution to obtain an abnormal distribution feature set, including: The change amplitude of the time slice data is counted, and if the change amplitude exceeds a preset change amplitude threshold, the fluctuation of the time slice data is determined to be abnormal and is marked as an abnormal segment to obtain an abnormal fluctuation segment; Features of the abnormal fluctuation segment are extracted and synthesized into a data set to obtain a risk signal feature set; The frequency and amplitude change law of the abnormal data fluctuation in the risk signal feature set are calculated, and the distribution of the abnormal data is counted to obtain an abnormal distribution feature set.

[0009] In an optional implementation, the abnormal data in the abnormal distribution feature set is subjected to quantitative scoring to obtain a risk quantitative index, including: The abnormal data in the abnormal distribution feature set is subjected to classification and aggregation to obtain an abnormal data point set; Threshold value judgment is performed on the frequency or amplitude of the abnormal data points in the abnormal data point set, and if the frequency exceeds a preset frequency threshold or the amplitude exceeds a preset amplitude threshold, the abnormal data point score is calculated to obtain a quantitative score; When the quantification score is in a preset low-risk interval, the data point is determined as low risk, and a risk quantification index with a low risk level is obtained; When the quantification score is in a preset medium-risk interval, the data point is determined as medium risk, and a risk quantification index with a medium risk level is obtained; When the quantification score is in a preset high-risk interval, the data point is determined as high risk, and a risk quantification index with a high risk level is obtained.

[0010] In an optional embodiment, the risk quantification index is subjected to early warning level determination to obtain an early warning level, and a corresponding individualized prevention suggestion is generated according to the early warning level, which comprises: obtaining heart rate data of the individual; verifying the heart rate data for completeness and reliability to obtain reliable heart rate data; calculating an average heart rate of the reliable heart rate data, and when the average heart rate exceeds a preset heart rate threshold and the risk quantification index is high risk, confirming that the early warning level is high, and obtaining an early warning level with a high level; When the average heart rate exceeds the preset heart rate threshold and the risk quantification index is medium risk, the early warning level is confirmed as medium, and an early warning level with a medium level is obtained; When the average heart rate exceeds the preset heart rate threshold and the risk quantification index is low risk, the early warning level is confirmed as low, and an early warning level with a low level is obtained; According to the early warning level, an intervention scheme is searched in a preset intervention scheme database to obtain a corresponding individualized prevention suggestion.

[0011] In an optional embodiment, the execution effect of the individualized prevention suggestion is continuously monitored, the night blood pressure fluctuation data before and after prevention is compared, the effectiveness of the prevention suggestion is determined, and a closed-loop precise physical examination health management system is formed, which comprises: The execution effect of the individualized prevention suggestion is continuously monitored, and night blood pressure fluctuation data and dietary habit data after prevention are collected to obtain blood pressure change data; The night blood pressure fluctuation value and the dietary habit index change value of the data set after prevention are calculated to obtain fluctuation correlation data; If the night blood pressure fluctuation value in the fluctuation correlation data exceeds a preset fluctuation threshold, the correlation strength between the change of the dietary habit index and the night blood pressure fluctuation value is analyzed, and an intervention effect evaluation result is obtained according to the correlation strength; The effectiveness of the prevention suggestion is determined according to the intervention effect evaluation result and dynamically adjusted to form a closed-loop precise physical examination health management system.

[0012] In an alternative embodiment, the multi-dimensional data set comprises: The multi-dimensional data set comprises a structured physiological index data source and an unstructured living habit data source, wherein the structured physiological index data source comprises night blood pressure fluctuation data recorded by a physical examination device, and the unstructured living habit data source comprises diet log text and interview records filled in by a user; In a second aspect, the present application provides an artificial intelligence-based physical examination center data management system, comprising: A data acquisition module is configured to acquire a multi-dimensional data set, wherein the multi-dimensional data set comprises night blood pressure fluctuation data; A data fusion module is configured to perform numerical normalization processing and vector conversion according to the multi-dimensional data set, and synthesize a feature matrix to obtain a multi-dimensional feature fusion matrix; A data slicing module is configured to perform time slicing processing of different granularities on the multi-dimensional feature fusion matrix to obtain time-sliced data; A data statistics module is configured to statistically analyze the change amplitude of the time-sliced data, and if the change amplitude exceeds a preset change amplitude threshold, mark the time-sliced data as abnormal and identify a potential risk signal distribution to obtain an abnormal distribution feature set; A data classification module is configured to quantitatively score abnormal data in the abnormal distribution feature set to obtain a risk quantitative index; A data judgment module is configured to judge a warning level according to the risk quantitative index to obtain the warning level, and generate a corresponding individualized prevention suggestion according to the warning level; A data monitoring module is configured to continuously monitor the execution effect of the individualized prevention suggestion, compare night blood pressure fluctuation data before and after prevention, judge the effectiveness of the prevention suggestion, and form a closed-loop precise physical examination health management system.

[0013] In a third aspect, the present application further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the artificial intelligence-based physical examination center data management method of any one of the above.

[0014] In a fourth aspect, the present application further provides a computer readable storage medium comprising a stored computer program, wherein the computer program controls a device where the computer readable storage medium is located to execute the artificial intelligence-based physical examination center data management method of any one of the above when the computer program is running.

[0015] Compared with the prior art, the present application has the following beneficial effects: The application realizes the deep correlation of "physiological indicators-lifestyle habits" by integrating and fusing the structured night blood pressure fluctuation data (such as systolic pressure 140 mmHg) and the unstructured diet text (such as "ate salted fish fried rice at 8 o'clock last night") into a unified matrix through normalization processing (blood pressure standardization) and vector conversion (diet text into 768-dimensional vector), and breaks the data silos.

[0016] The application can capture short-term blood pressure surges (such as 2-hour fluctuation after high-salt diet) and identify long-term trends (such as abnormal blood pressure pattern caused by late dinner for 1 week) by slicing data according to different scales such as hours, days, etc. and dynamically tracking changes combined with sliding window mechanism, solving the limitations of traditional "static analysis" scheme.

[0017] The application can not only accurately mark abnormalities (such as time slices with change amplitude exceeding threshold) but also locate the causes (such as strong correlation between certain abnormal fragments and "high-salt intake 2 hours ago") by risk signal distribution identification (step 4), combined with pattern matching (such as typical pattern library of "blood pressure surge after high-salt diet") and verification rules (such as time alignment and causal correlation test), upgrading abnormal identification from "single numerical judgment" to "multi-factor correlation analysis" and reducing missed detection and false positives. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a flowchart of the data management method of the physical examination center based on artificial intelligence provided by the first embodiment of the application; Figure 2 is a structural diagram of the data management system of the physical examination center based on artificial intelligence provided by the second embodiment of the application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0020] Referring to Figure 1 The first embodiment of the application provides a data management method of a physical examination center based on artificial intelligence, including the following steps: S11, obtaining a multi-dimensional data set, the multi-dimensional data set containing night blood pressure fluctuation data; S12, performing numerical normalization processing and vector conversion according to the multi-dimensional data set, and synthesizing a feature matrix to obtain a multi-dimensional feature fusion matrix. S13, performing time slicing processing on the multi-dimensional feature fusion matrix with different granularities to obtain time slice data; S14, counting the change amplitude of the time slice data, and if the change amplitude exceeds a preset change amplitude threshold, marking the time slice data as abnormal and identifying a potential risk signal distribution to obtain an abnormal distribution feature set; S15, quantitatively scoring the abnormal data in the abnormal distribution feature set to obtain a risk quantitative indicator; S16, judging a warning level for the risk quantitative indicator to obtain the warning level, and generating a corresponding personalized prevention suggestion according to the warning level; S17, continuously monitoring the execution effect of the personalized prevention suggestion, comparing the night blood pressure fluctuation data before and after prevention, judging the effectiveness of the prevention suggestion, and forming a closed-loop precise physical examination health management system.

[0021] In step S11, a multi-dimensional data set is obtained, which includes night blood pressure fluctuation data.

[0022] In an implementation manner, the multi-dimensional data set includes: The multi-dimensional data set includes a structured physiological index data source and an unstructured living habit data source, wherein the structured physiological index data source includes night blood pressure fluctuation data recorded by a physical examination device, and the unstructured living habit data source includes a diet log text and an interview record filled by a user. It should be noted that a multi-source fusion scheme is adopted for data collection: night blood pressure fluctuation data is collected by a wrist blood pressure meter certified by a medical instrument, the measurement range is 60-280 mmHg for systolic pressure and 40-199 mmHg for diastolic pressure, the static accuracy is ±3 mmHg, the pulse number is ±5%, the measurement time points are 22:00, 23:00, 00:00, 01:00, 02:00, 03:00, 04:00, 05:00, and 06:00, a total of 9 time points, three parameters of systolic pressure, diastolic pressure, and heart rate are recorded each time, and the data accuracy requirement is ±2 mmHg. The unstructured living habit data source includes a diet log text and an interview record, both of which are collected through a natural language processing interface for free description. The diet log text is recorded by the user daily, and is collected through a “diet log” module of a physical examination center APP. The user needs to input a description text through natural language, and the system receives the free description text in real time through a natural language processing interface, such as “ate salted fish fried rice of about 200 grams at 8 o'clock last night”. The interview record is collected through an “interview record” module: according to the night blood pressure abnormal data of the user (such as a sudden rise in blood pressure at a certain time period), the user's free description is input through a natural language processing interface to record the “diet details before and after the abnormal period”.

[0023] In step S12, numerical normalization processing and vector conversion are performed on the multi-dimensional data set, and a feature matrix is synthesized to obtain a multi-dimensional feature fusion matrix.

[0024] In an implementation manner, the numerical normalization processing and vector conversion are performed on the multi-dimensional data set, and the feature matrix is synthesized to obtain the multi-dimensional feature fusion matrix, including: Performing format conversion on the multi-dimensional data set to obtain a first data set; Performing standardization processing on the night blood pressure fluctuation data in the first data set to obtain a night blood pressure fluctuation data set; Extracting food types and time information of eating habits in the first data set to obtain an eating habit text; Converting the eating habit text into a vector representation to obtain an eating habit vector; Integrating the night blood pressure fluctuation data set and the eating habit vector representation to obtain a multi-dimensional feature fusion matrix.

[0025] It should be noted that the format conversion on the multi-dimensional data set is performed by a data preprocessing module to uniformly convert structured physiological index data (such as blood pressure values) and unstructured living habit data (such as eating text) into a JSON format, and a HL7 FHIR R4 standard protocol is used for data format conversion. The blood pressure data is mapped to an Observation resource, wherein a LOINC code (8480-6 represents systolic pressure, and 8462-4 represents diastolic pressure) is used for the code field, the eating data is mapped to a NutritionOrder resource, a SNOMED CT code is used for the foodType field, and the data is uniformly stored in a JSON format.

[0026] The night blood pressure fluctuation data in the first data set is standardized by Z-score standardization method, and the standardized value is calculated by the mean and standard deviation of the night blood pressure fluctuation data. The formula is standardized value=(original value-mean) / standard deviation. For example, the standardized value of a user's systolic pressure of 140 mmHg is (140-120) / 20=1.0, and the standardized night blood pressure value is obtained. Then, the food category and time information of the diet habit in the first data set is extracted, and the BiLSTM-CRF (Bidirectional Long Short-Term Memory Network-Conditional Random Field) named entity recognition algorithm is used. The algorithm first preprocesses the unstructured diet log text and interview record, splits the text into word sequences and assigns initial part-of-speech tags, then inputs the text sequence into the BiLSTM layer, captures the semantic dependency relationship from front to back and from back to front through the bidirectional neuron, generates high-dimensional semantic feature vectors for each word to distinguish entity and non-entity semantic differences, and then inputs the semantic feature vectors into the CRF layer to globally optimize the initial tag sequence and avoid entity fragmentation tagging. Finally, the food category information labeled as "FOOD" and the time information labeled as "TIME" are outputted, and the food entity (such as "salted fish fried rice") and the time entity (such as "8 o'clock last night") are extracted from the diet text. The time format is matched through regular expression matching, and invalid characters are filtered to obtain structured diet habit text.

[0027] The diet habit text is converted into vector representation, and a pre-trained BERT-base model (bert-base-chinese) is used. The diet text is tokenized and input into the model, and the hidden state of the [CLS] token is taken as the text vector representation. A 768-dimensional diet habit vector is obtained. The semantic correlation degree of "salted fish fried rice" and "high-salt diet" is 0.83. The cosine value of the included angle of the two vectors is calculated based on the cosine similarity formula, which is the dot product of the two vectors divided by the product of the lengths of the two vectors. For vector A ("salted fish fried rice" vector) and vector B ("high-salt diet" vector), the similarity can be calculated as follows: Finally, the night blood pressure value and the diet habit vector representation are integrated, and a horizontal splicing method is used to combine the 2-dimensional blood pressure value vector and the 768-dimensional text vector to form a 770-dimensional feature vector. A feature matrix with a dimension of N×770 (N is the number of samples) is constructed. To eliminate the dimensional influence, the L2 norm of each row vector is normalized to 1, and a multi-dimensional feature fusion matrix is obtained.

[0028] It is worth mentioning that the pre-trained BERT-base model (bert-base-chinese) for converting dietary habit text into vector representation is used in the application, and the structure is designed around the dietary text processing needs. Model structure: the model input layer receives the dietary habit text, after WordPiece segmentation, combined with word embedding, sentence embedding (single sentence fixed value), position embedding (maximum sequence length 128) to form an input vector; the core is a 12-layer bidirectional Transformer encoder (12 self-attention heads + GELU activation feedforward network per layer), which captures sub-word semantic association; the output layer takes the 768-dimensional hidden state of the [CLS] token as the dietary text vector. Training process: load the official pre-training weight of bert-base-chinese, fix the bottom 6-layer Transformer parameters to retain the general semantic ability, only fine-tune the upper 6-layer and output layer parameters, adapt to the dietary text processing scene in the two files, balance the generality and field adaptability. According to the dietary data logic in the two files, use 10000 "dietary text - risk label" data (labels are high salt / high sugar / high fat / ordinary) in the physical examination center, divide them into training, validation and test sets according to the ratio of 8:1:1, distinguish different risk dietary text vectors, use cross-entropy loss function, AdamW optimizer (learning rate 2e-5), batch 32 iteration; measure the accuracy rate with the validation set every round (≥90% required), stop early if there is no improvement for 3 consecutive rounds, and prevent explosion with gradient clipping. The difference between the test set accuracy and the validation set accuracy should be less than 3%, and the parameters are fixed after validation, and deployed to the system for text vector conversion.

[0029] In step S13, the multi-dimensional feature fusion matrix is subjected to time slicing processing of different granularities to obtain time slicing data.

[0030] It should be noted that, in order to perform time slicing processing on the multi-dimensional feature fusion matrix with different granularities, it is necessary to first set a time window parameter, which mainly covers three core sub-parameters of "window granularity", "sliding step" and "edge completion rule". The three core sub-parameters can be dynamically adjusted through a configuration file, and can also be customized by a medical institution. The "window granularity" can be taken as an hour, and 24 hours of data can be divided into 24 time slices, each of which contains the average blood pressure and diet record in the hour. The "sliding step" parameter is used to define the time interval between adjacent two time windows, and determines the coverage density of the slices. By default, the "sliding step" parameter is consistent with the "window granularity" (for example, a 1-hour window corresponds to a 1-hour step). The "edge completion rule" parameter is used to define the data completion method in the case that the data of the edge period is less than one window during the slicing process. The zero padding method is adopted, and the padding value is marked as -1. After the data slicing operation, according to the currently configured window parameters, the feature fusion matrix is sliced, and each window contains all the feature vectors of the corresponding period. For example, starting from 22:00, the slicing is performed every hour to generate 22:00-23:00, 23:00-00:00, etc. The data of the edge period less than one window is completed by using the zero padding method, and the padding value is marked as -1 to distinguish from the real data.

[0031] In step S14, the change amplitude of the time-sliced data is counted. If the change amplitude exceeds a preset change amplitude threshold, the time-sliced data is marked as abnormal and the potential risk signal distribution is identified, and an abnormal distribution feature set is obtained.

[0032] In an implementation manner, the counting of the change amplitude of the time-sliced data and the identification of the potential risk signal distribution if the change amplitude exceeds a preset change amplitude threshold to obtain an abnormal distribution feature set include: The change amplitude of the time-sliced data is counted. If the change amplitude exceeds a preset change amplitude threshold, the fluctuation abnormality of the time-sliced data is determined and the time-sliced data is marked as an abnormal segment, and an abnormal fluctuation segment is obtained. The features of the abnormal fluctuation segment are extracted and a data set is synthesized, and a risk signal feature set is obtained. The frequency and amplitude change law of the abnormal data fluctuation in the risk signal feature set are calculated, the distribution of the abnormal data is counted, and an abnormal distribution feature set is obtained.

[0033] It should be noted that the change amplitude is calculated, and the relative change rate of the feature vector of each time slice to the previous slice is calculated, and the formula is change amplitude = |current window mean - previous window mean| / previous window mean x 100%, and the change amplitudes of three key indicators of systolic blood pressure, diastolic blood pressure and diet correlation degree are calculated respectively. Secondly, set the change amplitude threshold, and dynamically adjust according to age grouping: the systolic blood pressure threshold of 18-44 year-old group is 20%, the diastolic blood pressure threshold of 45-64 year-old group is 15%; the systolic blood pressure threshold of 45-64 year-old group is 25%, the diastolic blood pressure threshold of 65 year-old group is 20%; the systolic blood pressure threshold of 65 year-old group is 30%, and the diastolic blood pressure threshold of 65 year-old group is 25%, and the indicators exceeding the threshold are marked as abnormal. Extract the key features related to nighttime blood pressure fluctuation and eating habits from the abnormal fluctuation segment, including blood pressure features: systolic / diastolic blood pressure fluctuation peak, fluctuation duration, peak occurrence time; eating features: food types (such as high salt, high fat) within 2 hours before the abnormal period, intake time, food intake; correlation features: time difference between diet intake and blood pressure fluctuation (such as blood pressure surge 1 hour after high-salt diet). The above features are integrated into a structured data set, and each record contains "abnormal period + blood pressure fluctuation parameters + diet parameters + time correlation parameters", for example: "22:00-23:00, systolic blood pressure peak 140mmHg (fluctuation 20%), 21:00 intake high-salt food (5g), time difference 1 hour". Analyze the frequency and amplitude change rule of abnormal data points in the risk signal set, for example, analyze the blood pressure data from 22:00 to 2:00 the next day within a week, and find that the frequency of abnormal fluctuation (amplitude exceeding 15mmHg) is 1-2 times a day, mainly concentrated in 2-3 hours after high-salt diet, and the amplitude change rule shows that after high-salt intake, the blood pressure fluctuation amplitude increases by an average of 12mmHg. Finally, extract the abnormal distribution characteristics, calculate the distribution characteristics set of the confirmed risk signals: count the number of abnormal windows in 24 hours to get the abnormal occurrence frequency, count the number of windows from the beginning of the abnormality to the recovery to get the duration, calculate the maximum change amplitude to determine the fluctuation amplitude, and count the proportion of abnormal windows in each period to determine the occurrence period distribution. Integrate the frequency, duration, amplitude and period distribution data obtained by the above quantitative statistics to form the abnormal distribution characteristic set.

[0034] In step S15, the abnormal data in the abnormal distribution characteristic set is quantitatively scored to obtain a risk quantitative index.

[0035] In an implementation manner, the quantitatively scoring the abnormal data in the abnormal distribution characteristic set to obtain a risk quantitative index comprises: Classify and aggregate the abnormal data in the abnormal distribution characteristic set to obtain an abnormal data point set; According to the frequency or amplitude of the abnormal data points in the set of abnormal data points, a threshold judgment is made. If the frequency exceeds a preset frequency threshold or the amplitude exceeds a preset amplitude threshold, the abnormal data point score is calculated to obtain a quantitative score; When the quantitative score is in a preset low-risk interval, the data point is determined to be a low-risk, and a risk quantification index with a low-risk level is obtained. When the quantitative score is in a preset medium-risk interval, the data point is determined to be a medium-risk, and a risk quantification index with a medium-risk level is obtained. When the quantitative score is in a preset high-risk interval, the data point is determined to be a high-risk, and a risk quantification index with a high-risk level is obtained.

[0036] It should be noted that the abnormal data in the abnormal distribution feature set is classified and aggregated to obtain the set of abnormal data points. In combination with the structured data of the abnormal distribution feature set in step S14 (such as “22:00-23:00, systolic pressure fluctuation 20%, 1 hour after high-salt diet” “23:00-00:00, diastolic pressure fluctuation 18%, 2 hours after high-fat diet”), a double classification rule based on “abnormal inducement - occurrence period” is adopted: first, the abnormal data is divided into 4 categories according to the diet correlation degree feature (such as high salt, high fat, high sugar, and no clear diet inducement), and then each category is further subdivided according to the occurrence period (22:00-24:00, 00:00-02:00, 02:00-06:00), and the same feature is aggregated to form the set of abnormal data points (such as “high-salt diet correlation - 22:00-24:00 abnormal data point set” “high-fat diet correlation - 00:00-02:00 abnormal data point set”).

[0037] According to the frequency and amplitude of the abnormal data points in the set of abnormal data points, a threshold judgment is made: the frequency threshold is calculated daily and set to “≥3 times per day” (i.e. the number of abnormal window in 24 hours is ≥3); the amplitude threshold follows the age grouping standard of step S14 (systolic pressure 20% for 18-44 age group, diastolic pressure 15% for 18-44 age group, etc.); for each abnormal data point, if any of the “frequency exceeds the threshold” and “amplitude exceeds the threshold” is met, the quantitative score calculation is performed; if both thresholds are not exceeded, the data point is not included in the score and is marked as “to be observed data point” to avoid low-risk data interference with the score result. The abnormal amplitude is calculated as “actual amplitude / threshold for this age group”x10 (e.g. for 45-64 age group, the actual systolic pressure fluctuation is 30%, then 30% / 25%x10=12 points, and the upper limit is set to 10 points); the occurrence frequency is calculated as “actual daily frequency / 3 times”x10 (e.g. actual 4 times, then 4 / 3x10≈13.3 points, upper limit 10 points) Final quantification score = (abnormal amplitude score x 0.5) + (occurrence frequency score x 0.5).

[0038] According to the interval of the quantification score, the risk level is determined, and a risk quantification index is obtained. Three risk intervals are preset: a quantification score of 1-3 is a low risk interval, when the score falls into the interval, the data point is determined to be a low-level risk, and a risk quantification index of "low-level risk (score X points)" is generated (such as "low-level risk (2.5 points)"); 4-7 is a medium risk interval, and when the score falls into the interval, it is determined to be a medium-level risk, and a risk quantification index of "medium-level risk (score X points)" is generated; 8-10 is a high risk interval, and when the score falls into the interval, it is determined to be a high-level risk, and a risk quantification index of "high-level risk (score X points)" is generated (such as "high-level risk (8.3 points)").

[0039] In step S16, the risk quantification index is judged for the warning level, the warning level is obtained, and the corresponding individual prevention suggestion is generated according to the warning level.

[0040] In an implementation mode, the risk quantification index is judged for the warning level, the warning level is obtained, and the corresponding individual prevention suggestion is generated according to the warning level, including: Obtaining heart rate data of an individual; Verifying the integrity and reliability of the heart rate data to obtain reliable heart rate data; Calculating the average heart rate of the reliable heart rate data, when the average heart rate exceeds a preset heart rate threshold, and the risk quantification index is a high-level risk, confirming that the warning level is high, and obtaining a warning level of high level; When the average heart rate exceeds the preset heart rate threshold, and the risk quantification index is a medium-level risk, the warning level is confirmed to be medium, and a warning level of medium level is obtained; When the average heart rate exceeds the preset heart rate threshold, and the risk quantification index is a low-level risk, the warning level is confirmed to be low, and a warning level of low level is obtained; According to the warning level, an intervention scheme is searched in a preset intervention scheme database, and a corresponding individual prevention suggestion is obtained.

[0041] It should be noted that the individual heart rate data is first acquired, and the data source is consistent with the nighttime blood pressure fluctuation data collection logic in step S11, and is from a monitoring device (such as a wrist blood pressure meter, an electrocardiogram monitor) or a user wearable device certified by a physical examination center, and the collection period is synchronized with the nighttime blood pressure monitoring period (22:00-6:00 the next day), and 1 record is recorded every 30 minutes, to ensure alignment with the time dimension of blood pressure and diet data. The heart rate data is verified for integrity and reliability: for integrity, check the continuity of the timestamp, for the period with a missing interval ≤1 hour and valid data before and after, first convert the timestamp (HH:MM) to the number of minutes, according to the previous valid heart rate and its corresponding time (recorded as the previous valid time), the next valid heart rate and its corresponding time (recorded as the previous valid time), the missing point time, calculate the total time interval = the next valid time - the previous valid time and the time interval of the missing point and the previous point = the missing point time - the previous valid time, then according to the formula missing heart rate = the previous valid heart rate + (the next valid heart rate - the previous valid heart rate) x (the missing point time / total time interval) to obtain the missing heart rate; if the missing heart rate is within the normal range of nighttime resting heart rate (50-70 times / minute), mark “interpolation supplement” and fill in, if it is outside the range, use (the previous valid heart rate + the next valid heart rate) / 2 to supplement, if there is no record in a certain period and the interval exceeds 1 hour, mark it as “invalid data segment”; for reliability, remove abnormal values outside the physiological range (nighttime resting heart rate normal range 50-70 times / minute, less than 40 times / minute or more than 100 times / minute marked as invalid), obtain reliable heart rate data. Finally, the nighttime average heart rate of the reliable heart rate data is calculated, the formula is “average heart rate = total sum of nighttime valid heart rate values / number of valid records”, which is the core physiological indicator for early warning judgment, to judge the early warning level, the preset nighttime resting heart rate threshold (dynamic adjustment according to age: ≥80 times / minute for 18-44 year-old group, ≥75 times / minute for 45-64 year-old group, ≥70 times / minute for 65 year-old group), when the average heart rate exceeds the preset threshold, the heart rate abnormality and the risk quantification indicator form a risk superposition, and the early warning level directly corresponds to the risk level: when the risk quantification indicator is high risk (8-10 points), the early warning level is high; when it is medium risk (4-7 points), the early warning level is medium; when it is low risk (1-3 points), the early warning level is low. When the average heart rate does not exceed the preset threshold, the heart rate is normal and the original risk is reduced, and the early warning level is judged according to the rule of “risk quantification indicator reduced by one level”: high risk reduced to medium early warning, medium risk reduced to low early warning, low risk reduced to no early warning, to avoid excessive intervention.

[0042] Matching intervention scheme, retrieving a matching scheme from a preset scheme database based on the warning level, the database storing four categories of intervention schemes corresponding to no / low / medium / high level warning, each category of scheme associated with individual attributes such as age grouping, dietary triggers, heart rate characteristics, etc., generating suggestions according to the "warning level priority + individual characteristics fine-tuning" logic matching. High-level warning (heart rate threshold + high risk) matches strong intervention scheme, including quantitative dietary restrictions (such as daily salt intake ≤3g), medical collaboration suggestions (contacting cardiovascular doctors within 12 hours) and high-frequency monitoring requirements (heart rate-blood pressure monitoring every 15 minutes); medium-level warning (including heart rate threshold + medium risk, normal heart rate + high risk) matches moderate intervention scheme, including phased dietary goals, low-intensity exercise guidance (such as walking for 30 minutes before sleep) and regular monitoring (once every 1 hour); low-level warning (including heart rate threshold + low risk, normal heart rate + medium risk) matches mild intervention scheme, mainly dietary tips and lifestyle adjustments, with 3 times of monitoring per week; no warning (normal heart rate + low risk) matches regular monitoring scheme, only pushing health manuals and monthly follow-ups, and the final suggestion includes intervention goals, specific measures, execution frequency and monitoring indicators.

[0043] In step S17, the execution effect of the personalized prevention suggestion is continuously monitored, the night blood pressure fluctuation data before and after prevention is compared, the effectiveness of the prevention suggestion is judged, and a closed-loop precise physical examination health management system is formed.

[0044] In an implementation manner, the continuous monitoring of the execution effect of the personalized prevention suggestion, the comparison of the night blood pressure fluctuation data before and after prevention, the judgment of the effectiveness of the prevention suggestion, and the formation of the closed-loop precise physical examination health management system include: The execution effect of the personalized prevention suggestion is continuously monitored, and the night blood pressure fluctuation data and dietary habit data after prevention are collected to obtain blood pressure change data; The night blood pressure fluctuation value and the dietary habit index change value of the data set after prevention are calculated to obtain fluctuation correlation data; If the night blood pressure fluctuation value in the fluctuation correlation data exceeds a preset fluctuation threshold, the correlation strength between the dietary habit index change and the night blood pressure fluctuation value is analyzed, and an intervention effect evaluation result is obtained according to the correlation strength; The effectiveness of the prevention suggestion is judged according to the intervention effect evaluation result and dynamically adjusted to form a closed-loop precise physical examination health management system.

[0045] It should be noted that first, the effect monitoring mechanism is established, and the data of night blood pressure and dietary habits after intervention is continuously collected. The monitoring period is 4 weeks, and the evaluation is conducted once a week. The monitoring indicators include blood pressure compliance rate (the proportion of days with systolic pressure < 130 mmHg and diastolic pressure < 80 mmHg), dietary compliance rate (the proportion of meals performed according to the recommendations), abnormal frequency reduction rate (the number of abnormal windows after intervention / the number of abnormal windows before intervention), risk level change, data collection frequency is consistent with that before intervention, and a new execution log recording function is added. The user submits the execution of the intervention measures every day, and the system automatically associates the blood pressure data at the corresponding time point. Second, the intervention effect is evaluated. The self-before-and-after control design is adopted, and the 3σ control chart is used to monitor the trend of the indicators (the center line is the baseline value before intervention, and the control limit = baseline value ± 3 x baseline standard deviation). If the last 3 points fall on the same side of the center line or 6 points increase / decrease, it is considered to have a statistically significant change. The comprehensive scoring method is used for effect evaluation: significant effect (≥ 2 indicators improved and no deterioration), effective (1-2 indicators improved), and ineffective (no improvement or ≥ 1 deterioration). The scoring standard is determined by 10 experts through the Delphi method. Then, the intervention scheme is adjusted. Based on the evaluation results, the PDCA cycle is started: in the planning stage, the reasons for poor effect are analyzed (such as low compliance due to high execution difficulty); in the execution stage, targeted adjustments are made (such as changing "walking 10000 steps every day" to "30 minutes"); in the inspection stage, the effect after adjustment is evaluated (re-evaluation after 2 weeks of observation); in the treatment stage, effective measures are solidified, and ineffective schemes start expert consultation (cardiovascular doctors, nutritionists, and health managers form an expert group). Next, the closed-loop management process is optimized. According to the compliance, the intervention intensity is dynamically adjusted: if the compliance is > 80%, the reminder frequency is reduced, and the self-management authority is increased; if the compliance is 50-80%, the intensity is maintained, and the success case incentive is increased; if the compliance is < 50%, the scheme is simplified to core measures (≤ 3), the daily reminder is increased, and a user incentive mechanism (health points exchange for health examination discount) is established. If the compliance is continuously met for 4 weeks, it is upgraded to "self-management" mode. Finally, the health management report is generated, including the comparison of indicators before and after intervention, abnormal pattern improvement, dietary behavior change analysis, risk level trend, next step suggestion, data change is displayed using visual charts, and mechanism is explained in simple language (such as "reducing sodium intake reduces blood volume, making the average night systolic pressure decrease by 12 mmHg"). The suggestions are divided into short-term (1 month), medium-term (3 months), and long-term (6 months) goals, forming a closed-loop health management system for continuous improvement.

[0046] In summary, the present application integrates structured night blood pressure fluctuation data (such as systolic blood pressure 140 mmHg) and unstructured diet text (such as "ate salted fish fried rice at 8 pm last night") through normalization processing (blood pressure standardization) and vector conversion (diet text converted to 768-dimensional vector) to realize the deep correlation of "physiological indicators - living habits", and break the data silos. Through multi-granularity time slicing, the data is cut according to different scales such as hours and days, combined with the sliding window mechanism to dynamically track changes, which can not only capture short-term blood pressure surges (such as 2 hours after high-salt diet), but also identify long-term trends (such as abnormal blood pressure patterns caused by eating late dinner for a week), solving the limitations of traditional "static analysis". Through risk signal distribution identification (step 4), combined with pattern matching (such as a typical pattern library of "blood pressure surge after high-salt diet") and verification rules (such as time alignment and causal correlation test), not only can the abnormality (such as time slices with change amplitude exceeding the threshold) be accurately marked, but also the cause (such as strong correlation between a certain abnormal segment and "high-salt intake 2 hours ago") can be located, so that the abnormality identification is upgraded from "single numerical judgment" to "multi-factor correlation analysis", reducing missed detection and misjudgment.

[0047] Reference Figure 2 The second embodiment of the present application provides a physical examination center data management system based on artificial intelligence, comprising: A data acquisition module is configured to acquire a multi-dimensional data set, wherein the multi-dimensional data set comprises night blood pressure fluctuation data. A data fusion module is configured to perform numerical normalization processing and vector conversion according to the multi-dimensional data set, and synthesize a feature matrix to obtain a multi-dimensional feature fusion matrix. A data slicing module is configured to perform time slicing processing of different granularities on the multi-dimensional feature fusion matrix to obtain time slicing data. A data statistics module is configured to count the change amplitude of the time slicing data, and if the change amplitude exceeds a preset change amplitude threshold, mark the time slicing data as abnormal and identify potential risk signal distribution to obtain an abnormal distribution feature set. A data classification module is configured to quantitatively score the abnormal data in the abnormal distribution feature set to obtain a risk quantitative index. A data judgment module is configured to judge the risk quantitative index to obtain a warning level, and generate a corresponding individualized prevention suggestion according to the warning level. A data monitoring module is configured to continuously monitor the execution effect of the individualized prevention suggestion, compare the night blood pressure fluctuation data before and after prevention, judge the effectiveness of the prevention suggestion, and form a closed-loop precise physical examination health management system.

[0048] It should be noted that the artificial intelligence-based physical examination center data management device provided in the embodiments of the present application is used to execute all process steps of the artificial intelligence-based physical examination center data management method of the above-mentioned embodiments, and the working principles and beneficial effects of the two are one-to-one correspondence, so they will not be repeated here.

[0049] The embodiments of the present application also provide an electronic device. The electronic device includes a processor, a memory, and a computer program, such as an algorithm program, stored in the memory and executable on the processor. The processor implements the steps in each of the above artificial intelligence-based physical examination center data management method embodiments when executing the computer program, for example Figure 1 The steps S11 shown. Alternatively, the processor implements the functions of each module / unit in each of the above device embodiments when executing the computer program, for example, the data monitoring module.

[0050] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments that can complete a specific function, which are used to describe the execution process of the computer program in the electronic device.

[0051] The electronic device can be a desktop computer, a notebook, a palm computer, and a smart tablet, etc. The electronic device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device, and can include more or fewer components than the above, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, etc.

[0052] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The processor is the control center of the electronic device, which connects all parts of the electronic device through various interfaces and lines.

[0053] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the electronic device by running or executing the computer program and / or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, an image playing function, etc.), and the like; and the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), and the like. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0054] The modules / units integrated in the electronic device, if realized in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can realize the steps of the above-mentioned various method embodiments when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the contents included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0055] It should be noted that the apparatus embodiments described above are merely illustrative, and the units described as separate units can or can not be physically separate, and the units displayed as units can or can not be physical units, i.e. can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. In addition, the connection relationship between the modules in the apparatus embodiment provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0056] The above specific embodiments further illustrate the purpose, technical scheme and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. It is particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An artificial intelligence-based physical examination center data management method, characterized by, The method comprises the following steps: acquiring a multi-dimensional data set containing night blood pressure fluctuation data; performing numerical normalization and vector conversion on the multi-dimensional data set, and synthesizing a feature matrix to obtain a multi-dimensional feature fusion matrix; performing time slicing processing of different granularities on the multi-dimensional feature fusion matrix to obtain time slice data; statistically analyzing the change amplitude of the time slice data, and if the change amplitude exceeds a preset change amplitude threshold, marking the time slice data as abnormal and identifying a potential risk signal distribution to obtain an abnormal distribution feature set; quantitatively scoring the abnormal data in the abnormal distribution feature set to obtain a risk quantification index; judging the warning level of the risk quantification index to obtain a corresponding warning level, and generating a corresponding personalized prevention suggestion according to the warning level; continuously monitoring the execution effect of the personalized prevention suggestion, comparing the night blood pressure fluctuation data before and after prevention, judging the effectiveness of the prevention suggestion, and forming a closed-loop precision physical examination health management system. 2.The artificial intelligence-based physical examination center data management method of claim 1, wherein, The method comprises the following steps: performing format conversion on the multi-dimensional data set to obtain a first data set; performing standardization processing on the night blood pressure fluctuation data in the first data set to obtain a night blood pressure fluctuation data set; extracting food types and time information of eating habits in the first data set to obtain an eating habit text; converting the eating habit text into a vector representation to obtain an eating habit vector; integrating the night blood pressure fluctuation data set and the eating habit vector representation to obtain a multi-dimensional feature fusion matrix. 3.The AI-based physical examination center data management method of claim 1, wherein, The method comprises the following steps: statistically analyzing the change amplitude of the time slice data, and if the change amplitude exceeds a preset change amplitude threshold, determining the fluctuation abnormality of the time slice data and marking it as an abnormal segment to obtain an abnormal fluctuation segment; extracting features of the abnormal fluctuation segment and synthesizing a data set to obtain a risk signal feature set; calculating the frequency and amplitude variation law of the abnormal fluctuation data in the risk signal feature set, and statistically analyzing the distribution of the abnormal data to obtain an abnormal distribution feature set. 4.The method of claim 1, wherein, The method comprises the following steps: classifying and aggregating the abnormal data in the abnormal distribution feature set to obtain an abnormal data point set; performing threshold judgment on the frequency or amplitude of the abnormal data points in the abnormal data point set, and if the frequency exceeds a preset frequency threshold or the amplitude exceeds a preset amplitude threshold, calculating the abnormal data point score to obtain a quantitative score; when the quantitative score is in a preset low-risk interval, judging that the data point is a low-risk, and obtaining a risk quantification index with a low-risk level; when the quantitative score is in a preset medium-risk interval, judging that the data point is a medium-risk, and obtaining a risk quantification index with a medium-risk level. When the quantification score is in a preset high-risk interval, the data point is determined as high risk, and a risk quantification index with a high risk level is obtained. 5.The method of claim 1, wherein, The risk quantification index is subjected to early warning level judgment to obtain an early warning level, and corresponding individualized prevention suggestions are generated according to the early warning level, including: obtaining heart rate data of the individual; verifying the integrity and reliability of the heart rate data to obtain reliable heart rate data; calculating the average heart rate of the reliable heart rate data, and when the average heart rate exceeds a preset heart rate threshold and the risk quantification index is high risk, confirming that the early warning level is high, and obtaining an early warning level with a high level; when the average heart rate exceeds the preset heart rate threshold and the risk quantification index is medium risk, confirming that the early warning level is medium, and obtaining an early warning level with a medium level; when the average heart rate exceeds the preset heart rate threshold and the risk quantification index is low risk, confirming that the early warning level is low, and obtaining an early warning level with a low level; According to the early warning level, searching for an intervention scheme in a preset intervention scheme database to obtain corresponding individualized prevention suggestions. 6.The method of claim 1, wherein, The execution effect of the individualized prevention suggestions is continuously monitored, and the night blood pressure fluctuation data before and after prevention is compared to judge the effectiveness of the prevention suggestions, forming a closed-loop precise physical examination health management system, including: continuously monitoring the execution effect of the individualized prevention suggestions, and collecting night blood pressure fluctuation data and dietary habit data after prevention to obtain a post-prevention data set; calculating the night blood pressure fluctuation value and dietary habit index change value of the post-prevention data set to obtain fluctuation correlation data; if the night blood pressure fluctuation value in the fluctuation correlation data exceeds a preset fluctuation threshold, the correlation strength between the change of the dietary habit index and the night blood pressure fluctuation value is analyzed, and the intervention effect evaluation result is obtained according to the correlation strength; According to the intervention effect evaluation result, the effectiveness of the prevention suggestions is judged and dynamically adjusted to form a closed-loop precise physical examination health management system. 7.The AI-based physical examination center data management method of claim 1, wherein, The multi-dimensional data set includes: The multi-dimensional data set includes structured physiological index data sources and unstructured life habit data sources, wherein the structured physiological index data sources include night blood pressure fluctuation data recorded by physical examination equipment, and the unstructured life habit data sources include dietary log texts and interview records filled in by users.

8. An artificial intelligence-based physical examination center data management system, characterized by, including: a data acquisition module for acquiring a multi-dimensional data set, the multi-dimensional data set containing night blood pressure fluctuation data; a data fusion module for performing numerical normalization processing and vector conversion according to the multi-dimensional data set, and synthesizing a feature matrix to obtain a multi-dimensional feature fusion matrix; a data slicing module for performing time slicing processing of different granularities on the multi-dimensional feature fusion matrix to obtain time slicing data; a data statistics module for counting the change amplitude of the time slicing data, and if the change amplitude exceeds a preset change amplitude threshold, marking the time slicing data as abnormal and identifying potential risk signal distribution to obtain an abnormal distribution feature set; a data classification module for quantifying the abnormal data in the abnormal distribution feature set to obtain a risk quantification index; The data judgment module is configured to judge the early warning level of the risk quantification index, obtain an early warning level, and generate a corresponding personalized prevention suggestion according to the early warning level. The data monitoring module is configured to continuously monitor the execution effect of the personalized prevention suggestion, compare the night blood pressure fluctuation data before and after prevention, judge the effectiveness of the prevention suggestion, and form a closed-loop precise physical examination health management system.

9. An electronic device, comprising: The computer readable storage medium comprises a stored computer program, wherein the computer program, when executed, controls a device in which the computer readable storage medium is located to perform the artificial intelligence-based physical examination center data management method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises a stored computer program, wherein the computer program, when executed, controls a device in which the computer readable storage medium is located to perform the artificial intelligence-based physical examination center data management method according to any one of claims 1 to 7.

Images