A method and system for prompting screening management of high-risk groups for lung cancer

Abstract

The present invention discloses a method and system for prompting screening management of a high-risk group of lung cancer, comprising the following steps: obtaining initial lung cancer medical data and constructing an initial lung cancer knowledge graph; constructing a first type of lung cancer screening questionnaire according to the initial lung cancer knowledge graph, and screening out a first type of suspected persons according to the first type of lung cancer screening questionnaire; setting a data update cycle, obtaining current lung cancer medical data at a corresponding time node, and constructing a current lung cancer knowledge graph according to the current lung cancer medical data; comparing the current lung cancer knowledge graph with the initial lung cancer knowledge graph, determining whether to update the first type of lung cancer screening questionnaire, and screening out a second type of suspected persons according to the update result; sending prompting screening information to the suspected persons, and formulating a follow-up plan for them; the present invention overcomes the problem that the screening accuracy of the high-risk group of lung cancer cannot be dynamically improved, and reduces the screening burden on medical workers.

Description

Technical Field

[0001] The present invention belongs to the field of data analysis, and in particular relates to a method and system for prompting screening management of a high-risk group of lung cancer. Background Art

[0002] Lung cancer is a malignant tumor originating from the bronchial mucosa or glands of the lungs. It is one of the malignant tumors with the fastest growing incidence and mortality rate, and poses the greatest threat to people's health and life. In the past 50 years, many countries have reported that the incidence and mortality of lung cancer have increased significantly. The incidence and mortality of lung cancer in men rank first among all malignant tumors, while the incidence and mortality of lung cancer in women rank second.

[0003] Lung cancer is preventable and controllable. It is an important task for health systems in various places to identify people with a high probability of developing lung cancer through rich lung cancer data, give them reminders, recommend them to undergo early screening, clarify or exclude related risks, and include them in subsequent related management.

[0004] However, the medical systems used by major hospitals are relatively independent, and the same patient data uses different recording rules. The information of lung cancer patients and the data of tumor patients cannot be effectively utilized. The case data of each hospital have the phenomenon of information islands. In addition, the case data of lung cancer patients is constantly updated. It is impossible to formulate high-precision lung cancer risk judgment standards based on historical lung cancer data, thereby reducing the screening accuracy of high-risk groups for lung cancer and increasing the workload of medical workers. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present application provides a method and system for prompting screening management of people at high risk of lung cancer.

[0006] In the first aspect, the present application proposes a method for prompting screening and management of a high-risk group of lung cancer, comprising the following steps:

[0007] Obtain initial lung cancer medical data and construct an initial lung cancer knowledge graph;

[0008] Constructing a first type of lung cancer screening questionnaire according to the initial lung cancer knowledge graph, and screening out a first type of suspected persons according to the first type of lung cancer screening questionnaire;

[0009] Setting a data update cycle, obtaining current lung cancer medical data at a corresponding time node, and constructing a current lung cancer knowledge graph based on the current lung cancer medical data;

[0010] Comparing the current lung cancer knowledge graph with the initial lung cancer knowledge graph, determining whether to update the first type of lung cancer screening questionnaire, and screening out the second type of suspected persons according to the update result;

[0011] Sending a prompt screening message to the first category of suspected persons / the second category of suspected persons;

[0012] Develop a follow-up plan for the first category suspected persons / second category suspected persons.

[0013] In some embodiments, the obtaining of initial lung cancer medical data and constructing an initial lung cancer knowledge graph includes:

[0014] The initial lung cancer medical data is manually corrected and entity extraction and relationship extraction are performed according to the dictionary and data properties, a first type of triples is constructed according to the extraction results, and the initial lung cancer knowledge graph is constructed according to the first type of triples.

[0015] In some embodiments, the initial lung cancer medical data includes lung cancer medical literature and lung cancer patient case data before the current time point.

[0016] In some embodiments, constructing a first type of lung cancer screening questionnaire according to the initial lung cancer knowledge graph, and screening out a first type of suspected persons according to the first type of lung cancer screening questionnaire includes:

[0017] Combining the knowledge content of the initial knowledge graph and the question template to construct the first type of lung cancer screening questionnaire;

[0018] The first type of lung cancer screening questionnaire includes a preset number of initial questions, each of the initial questions is provided with at least two initial options, and each of the initial options is provided with a corresponding score;

[0019] A first category score is obtained according to the corresponding score of the subject after completing all the initial questions, the first category score is averaged and compared with the preset score, and the first category of suspected persons is screened out from the subject according to the comparison result.

[0020] In some embodiments, the setting of the data update period includes:

[0021] The data update cycle includes an entity update cycle and a relationship update cycle. The entity update cycle is obtained according to the change cycle of the entity in the initial knowledge graph, and the relationship update cycle is obtained according to the change cycle of the relationship in the initial knowledge graph;

[0022] The relationship update cycle and the entity update cycle are compared, and the update cycle with the shortest time is selected as the current update cycle.

[0023] In some embodiments, the current lung cancer medical data of the corresponding time node is obtained, and the current lung cancer knowledge graph is constructed according to the current lung cancer medical data;

[0024] Acquire the current lung cancer medical data of the corresponding time node according to the current update cycle;

[0025] Entity extraction and relationship extraction are performed on the current lung cancer medical data according to the dictionary and data properties, a second type of triples are constructed according to the extraction results, and the current lung cancer knowledge graph is constructed according to the second type of triples.

[0026] In some embodiments, the current lung cancer medical data includes lung cancer medical literature and lung cancer patient case data between a time point of a previous data acquisition and a current time point.

[0027] In some embodiments, the step of extracting entities from the initial lung cancer medical data / the current lung cancer medical data includes:

[0028] The entity extraction includes entity extraction of normalized data and entity extraction of non-normalized data;

[0029] When performing entity extraction on normalized data, entities are extracted directly based on text matching rules;

[0030] When performing entity extraction on non-normalized data, first mark the parts of speech and extract the nouns that match the field. Then, through text-based rule matching, obtain the entities in the field of lung cancer medicine and add them to the dictionary. Finally, use the dictionary to match the words in the sentence and extract the entities based on the matching results.

[0031] In some embodiments, the step of extracting a relationship from the initial lung cancer medical data / the current lung cancer medical data includes:

[0032] The relationship extraction includes structured data relationship extraction and unstructured data relationship extraction;

[0033] Structured data relationship extraction includes relationship extraction between entities and attribute relationship extraction of some entities;

[0034] Unstructured data relation extraction uses a word vector convolutional neural network-based extraction method to perform relation extraction.

[0035] In some embodiments, the current lung cancer knowledge graph is compared with the initial lung cancer knowledge graph to determine whether to update the first type of lung cancer screening questionnaire, and the second type of suspected persons are screened out according to the update result. The comparison and judgment step includes:

[0036] Traversing the initial lung cancer knowledge graph and the current lung cancer knowledge graph respectively, and extracting the corresponding initial entity set and current entity set;

[0037] A similarity scoring function is set to perform similarity scoring calculation on the entities in the initial entity set and the current entity set to obtain a similarity scoring result, and a preset update threshold is used to compare the result with the similarity scoring result.

[0038] If the update threshold is exceeded, the first type of lung cancer screening questionnaire needs to be updated.

[0039] If the update threshold is not exceeded, it is directly determined that a prompt screening information needs to be sent to the first category of suspected persons.

[0040] In some embodiments, the current lung cancer knowledge graph is compared with the initial lung cancer knowledge graph to determine whether to update the first type of lung cancer screening questionnaire, and the second type of suspected persons are screened out according to the update result, including the graph update step:

[0041] Comparing the phrase type of the initial entity with the current entity set to determine a current entity that matches the initial entity;

[0042] Traversing the initial lung cancer knowledge graph to extract an initial path starting from any initial entity in the initial lung cancer knowledge graph;

[0043] Traversing the current lung cancer knowledge graph, extracting a path starting from the current entity as an update path;

[0044] Comparing the initial path with the updated path, and obtaining difference features according to the comparison result;

[0045] The initial lung cancer knowledge graph is updated according to the difference features to obtain the latest lung cancer knowledge graph.

[0046] In some embodiments, the step of screening out the second category of suspected persons according to the update results includes:

[0047] Combining the knowledge content of the latest lung cancer knowledge graph and the question template to construct a second type of lung cancer screening questionnaire;

[0048] The second type of lung cancer screening questionnaire includes a preset number of latest questions, each of the latest questions is provided with at least two latest options, and each of the latest options is provided with a corresponding score;

[0049] The second category scores are obtained according to the corresponding scores of the first category of suspected persons / a new batch of test subjects after completing all the latest questions, the second category scores are averaged and compared with the preset scores, and the second category of suspected persons are screened out from the first category of suspected persons according to the comparison results.

[0050] In some embodiments, the sending of prompt screening information to the first category suspected person / the second category suspected person includes:

[0051] It is suggested that the CT imaging information of the first category of suspected persons / the second category of suspected persons needs to be obtained, and a subsequent assessment of the risk of lung cancer is performed based on the analysis results of the CT imaging information of the first category of suspected persons / the second category of suspected persons.

[0052] In some embodiments, the subsequent assessment of lung cancer risk based on the analysis results of the CT image information of the first category of suspected persons / the second category of suspected persons includes:

[0053] Construct a lung cancer CT image training set;

[0054] Processing the lung cancer CT image training set to generate required training samples;

[0055] Use training samples to train a deep learning comparison model based on CT images and generate an image comparison analysis model;

[0056] Collecting CT images of the first category of suspected persons / the second category of suspected persons as an atlas to be analyzed;

[0057] The atlas to be analyzed is processed by an image comparison analysis model to obtain a further assessment of the risk of lung cancer.

[0058] In a second aspect, this embodiment further discloses a high-risk group of lung cancer prompt screening management system, including:

[0059] The data acquisition module is used to obtain initial lung cancer medical data and construct an initial lung cancer knowledge graph;

[0060] A first grassroots institution screening module is used to construct a first type of lung cancer screening questionnaire according to the initial lung cancer knowledge graph, and screen out a first type of suspected persons according to the first type of lung cancer screening questionnaire;

[0061] A graph updating module, which is used to set a data update cycle, obtain current lung cancer medical data at a corresponding time node, and construct a current lung cancer knowledge graph based on the current lung cancer medical data;

[0062] A second grassroots institution screening module is used to compare the current lung cancer knowledge graph with the initial lung cancer knowledge graph, determine whether to update the first type of lung cancer screening questionnaire, and screen out the second type of suspected persons according to the update result;

[0063] A screening and analysis module, used for sending prompt screening information to the first category suspected persons / second category suspected persons;

[0064] A plan-making module is used to make a follow-up plan for the first category suspected person / the second category suspected person.

[0065] In a third aspect, the present disclosure also discloses a computer device, the computer device comprising:

[0066] A processor is used to implement the steps of any of the above methods when executing a computer program stored in a memory.

[0067] In a fourth aspect, the embodiments of the present disclosure further disclose a computer-readable storage medium on which computer instructions are stored, and when the computer instructions are executed by a processor, the steps of any of the above methods are implemented.

[0068] Beneficial effects of the present invention:

[0069] First, the initial lung cancer medical literature, lung cancer patient information and tumor patient data, and lung cancer patient case data are obtained. Then, based on the above data, an initial lung cancer knowledge graph is constructed. Based on the initial lung cancer knowledge graph, a preliminary lung cancer screening questionnaire (i.e., the first type of lung cancer screening questionnaire) is formed to conduct the first stage of screening on the population to be tested. Subsequently, the current lung cancer medical data is obtained according to the update cycle, and it is compared with the initial lung cancer medical data to determine whether it needs to be updated. If no update is required, it proves that in the first stage of screening, the first type of suspected persons screened out have met the current lung cancer screening standards. If an update is required, the latest lung cancer knowledge graph is constructed based on the updated data, and the first type of lung cancer screening questionnaire is used to screen the population to be tested. The image recognition spectrum is used to form an updated lung cancer screening questionnaire (i.e., the second type of lung cancer screening questionnaire), and the first type of suspected persons are further screened / a new batch of persons to be tested are screened to obtain the screening results, thereby forming a closed-loop update of the screening criteria. The advantage of this is that the lung cancer screening criteria can be dynamically updated in real time, and the screening accuracy can be continuously improved. Finally, a prompt screening information is sent to the first type of suspected persons / the second type of suspected persons, and the corresponding CT analysis images are obtained. Then, a follow-up plan is formulated for the first type of suspected persons / the second type of suspected persons based on the CT analysis images. In this way, the problem that the screening accuracy of the high-risk population for lung cancer cannot be dynamically improved is overcome, and the screening burden on medical workers is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is the overall flow chart of the present invention.

[0071] Figure 2 Flowchart for constructing the initial knowledge graph.

[0072] Figure 3 Update the judgment flow chart for the questionnaire.

[0073] Figure 4Updated flowchart for knowledge graph comparison.

[0074] Figure 5 Flow chart for image inspection analysis.

[0075] Figure 6 This is the system structure diagram. DETAILED DESCRIPTION

[0076] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0077] In the first aspect, the present application proposes a method for prompting screening and management of a high-risk group of lung cancer, comprising the following steps:

[0078] S100: Obtain initial lung cancer medical data and construct an initial lung cancer knowledge graph;

[0079] Among them, the initial lung cancer medical data includes lung cancer medical literature and lung cancer patient case data before the current time point.

[0080] For example, lung cancer medical literature includes literature data from major literature collection websites such as CNKI and VIP; lung cancer patient case data includes: various disease indicators of lung cancer patients, systemic status examination, lung function test, vascular invasion, tumor quantity, tumor size, basic patient information, laboratory test items, imaging examination, surgery-related information, drug-related information, prognosis indicators, etc.

[0081] S101: manually correcting the initial lung cancer medical data and performing entity extraction and relationship extraction according to a dictionary and data properties;

[0082] Among them, entity extraction includes:

[0083] The entity extraction includes entity extraction of normalized data and entity extraction of non-normalized data;

[0084] When performing entity extraction on normalized data, entities are extracted directly based on text matching rules;

[0085] Among them, the rule base is specified based on the text characteristics of lung cancer medical literature and lung cancer patient case data;

[0086] For entity extraction of normalized data, entities are extracted directly based on content matching rules. The content matching rules are as follows:

[0087] For example, for lung cancer symptoms, lung cancer or symptom entities are extracted according to the rules of lung cancer complications and suspected lung cancer symptoms;

[0088] For lung cancer names, lung cancer entities and alias attributes are obtained according to the medical name, English name, and slang name rules.

[0089] When performing entity extraction on non-normalized data, first mark the parts of speech and extract the nouns that match the field. Then, through text-based rule matching, obtain the entities in the field of lung cancer medicine and add them to the dictionary. Finally, use the dictionary to match the words in the sentence and extract the entities based on the matching results.

[0090] When extracting entities from non-normalized data, the parts of speech are marked using the Jieba word segmentation to extract nouns that match the field. Then, through text-based rule matching, entities in the field of lung cancer medicine are obtained and added to the dictionary.

[0091] First, match the words in the sentence through the dictionary. If a match can be found, locate the phrase and the predicate verb closest to the matched entity. The nouns before and after the predicate verb are very likely to be entities in the medical field.

[0092] Among them, relationship extraction includes:

[0093] Relation extraction of structured data includes extraction of relations between entities and extraction of attribute relations of some entities;

[0094] The relation extraction of unstructured data is performed by using a word vector convolutional neural network-based extraction method.

[0095] S102: constructing a first type of triple according to the extraction result;

[0096] Among them, according to the results of entity extraction and relationship extraction, a triple of entity-predicate-entity, entity-relationship-entity or entity-relationship-attribute is constructed.

[0097] S103: Constructing the initial lung cancer knowledge graph according to the first type of triples.

[0098] The initial lung cancer knowledge graph is constructed based on the above triple data.

[0099] S200: constructing a first type of lung cancer screening questionnaire according to the initial lung cancer knowledge graph, and screening out a first type of suspected persons according to the first type of lung cancer screening questionnaire;

[0100] Among them, according to the triples in the knowledge graph, the first type of lung cancer screening questionnaire is generated based on the set template;

[0101] For example, set the template to:

[0102] The questionnaire survey included patient information, name, ID number, gender, date of birth, place of origin, ethnicity, personal contact number, emergency contact number, permanent address, and work unit; questions included whether you smoke (smoking is defined as smoking if you smoke more than one cigarette a day for more than six consecutive months or cumulatively), age at which you started smoking, how many cigarettes you smoke a day if you are still smoking or have smoked before (1 liang of tobacco leaves = 50 cigarettes), how long you have smoked in total (years) after deducting the years since you quit smoking if you are still smoking or have smoked before, how long you have smoked in total (years) if you have quit smoking now, how long has this quit lasted (years), for non-smokers, whether you have lived with a smoking family member for more than 20 years, for non-smokers, whether you have worked in the same room with a smoking colleague for more than 20 years, whether you have chronic obstructive pulmonary disease, whether you have a history of occupational exposure to harmful substances (more than one year), what kind of harmful substances you have been exposed to in your occupation (multiple choices), whether your parents, children, or siblings (same as parents) have lung cancer (diagnosed by a regular medical institution), etc.

[0103] For example, the first type of lung cancer screening questionnaire finally constructed is:

[0104] Initial question:

[0105] 1. Age?

[0106] Initial options:

[0107] A. 20-39 years old (1 point). B. 40-54 years old (2 points). C. 55 years old and above (3 points).

[0108] 2. Gender?

[0109] A. Female (1 point). B. Male (2 points)

[0110] 3. Do you smoke? (Smoking is defined as smoking more than one cigarette a day for more than 6 consecutive months or cumulatively)

[0111] A. No, never smoke (0 points)

[0112] B. Yes, I am still smoking (2 points)

[0113] C. Used to smoke, but has quit smoking now (1 point)

[0114] 4. Age of starting to smoke

[0115] A. Before 18 years old (3 points);

[0116] B. 18-40 years old (2 points);

[0117] C. Over 40 years old (1 point).

[0118] 5. If you are still smoking or have smoked in the past, how many cigarettes do you smoke every day (1 liang of tobacco leaves = 50 cigarettes)?

[0119] A. Less than 1 pack (2 points);

[0120] B. At least one pack (3 points).

[0121] 6. If you are currently smoking or have smoked in the past, how long have you smoked in total, excluding the years since you quit smoking?

[0122] A. Less than 5 years (1 point);

[0123] B. 6-10 years (2 points);

[0124] C. More than 11 years (3 points).

[0125] More questions, options, and scores are not described here;

[0126] The above questions and options are pushed to the test subject, and the corresponding scores after the test subject completes all the initial questions are obtained to obtain the total score (i.e., the first category score); the scores are averaged to obtain A1, and then a preset score value α1 is set. When A1 is greater than α1, the test subject is judged to be a first category suspected person, where α1 is the score set by experts and historical lung cancer judgment standards, and can also be changed according to needs based on continuous updating of data.

[0127] S300: Setting a data update cycle, acquiring current lung cancer medical data at a corresponding time node, and constructing a current lung cancer knowledge graph according to the current lung cancer medical data;

[0128] Among them, the data update cycle includes an entity update cycle and a relationship update cycle. The entity update cycle is obtained according to the change cycle of the entity in the initial knowledge graph, and the relationship update cycle is obtained according to the change cycle of the relationship in the initial knowledge graph; the relationship update cycle is compared with the entity update cycle, and the update cycle with the shortest time is selected as the current update cycle. When the relationship update cycle is greater than the entity update cycle, the entity update cycle is selected as the current update cycle, otherwise, the relationship update cycle is selected as the current update cycle. The period update unit of the current update cycle can be second / min / hour / day / month / year.

[0129] Acquire the current lung cancer medical data of the corresponding time node according to the current update cycle;

[0130] Among them, the current lung cancer medical data includes lung cancer medical literature and lung cancer patient case data between the time node of the previous data acquisition and the current time node.

[0131] Entity extraction and relationship extraction are performed on the current lung cancer medical data according to the dictionary and data properties, a second type of triples are constructed according to the extraction results, and the current lung cancer knowledge graph is constructed according to the second type of triples.

[0132] Among them, the current lung cancer knowledge graph is constructed according to the steps of constructing the initial lung cancer knowledge graph mentioned above, and the construction process will not be repeated here.

[0133] S400: Compare the current lung cancer knowledge graph with the initial lung cancer knowledge graph to determine whether to update the first type of lung cancer screening questionnaire, and screen out the second type of suspected persons according to the update result;

[0134] The step of comparing the current lung cancer knowledge graph with the initial lung cancer knowledge graph includes:

[0135] S410: traversing the initial lung cancer knowledge graph and the current lung cancer knowledge graph respectively, extracting the corresponding initial entity set and current entity set;

[0136] For the initial lung cancer knowledge graph, after traversal, entities of the target type can be selected one by one as initial entities for subsequent matching operations, and entity sets are created from the entities of the target type as the initial entity set so that each entity can be extracted from the set in sequence.

[0137] For the current lung cancer knowledge graph, when performing the above operation on the initial lung cancer knowledge graph, a traversal operation can also be performed on the current lung cancer knowledge graph, by traversing the current lung cancer knowledge graph, to extract at least one current entity of the target type in the current lung cancer knowledge graph, and create a current entity set based on the entity, and then use these entities for comparison in step S420.

[0138] S420: Setting a similarity scoring function to calculate similarity scores for the entities in the initial entity set and the current entity set, obtaining a similarity scoring result, and comparing the similarity scoring result with a preset update threshold.

[0139] Among them, the basis for comparison is the calculation result of the similarity scoring function:

[0140] The similarity score function is: PF = γ|1 / logG|+β|1 / ∑ilogTi|

[0141] Among them, γ and β are similarity weights, G is the similarity between the initial entity and the current entity, and Ti is the correlation between the i-th relationship data of the initial entity and the current entity. The update threshold ranges from 0 to 1. The closer it is to 1, the higher the similarity. In this example, the update threshold is set to 0.5. Therefore, when it exceeds the update threshold, it means that an update is needed.

[0142] Furthermore, a similarity weight between two entities can be set. The larger the similarity weight ratio is, the higher the similarity between the two entities is, and the greater the probability of being associated with each other is. The initial entity and the current entity can be randomly selected between the two entities, or can be specified according to actual conditions.

[0143] Furthermore, the degree of association between the relationship data of the initial entity and the current entity is mainly used to indicate the probability that the relationship data set of the initial entity and the relationship data corresponding to the current entity exist.

[0144] If the update threshold is exceeded, S430 is executed:

[0145] S430: Update the first type of lung cancer screening questionnaire.

[0146] Before that, update the knowledge graph to obtain the latest lung cancer knowledge graph. The steps are as follows:

[0147] S431: comparing the phrase type of the initial entity with the current entity set to determine a current entity that matches the initial entity;

[0148] For the initial entity in the selected initial lung cancer knowledge graph, we can first compare the preferred phrase of the entity, that is, use the preferred phrase in the initial lung cancer graph with the current entity set in the current lung cancer knowledge graph to determine whether there is an identical current entity. If there is no current entity identical to the preferred phrase, we can use the unified phrase in the phrase set corresponding to the initial entity for comparison, so as to determine the matching entity of the target type in the current lung cancer knowledge graph, that is, align the starting point of the path.

[0149] S432: traversing the initial lung cancer knowledge graph to extract an initial path starting from any initial entity in the initial lung cancer knowledge graph;

[0150] Traversing the initial lung cancer knowledge graph, extracting path data starting from any initial entity, thereby constructing an initial path starting from any initial entity;

[0151] S433: traversing the current lung cancer knowledge graph to extract a path starting from the current entity as an update path;

[0152] For the current lung cancer knowledge graph, path data starting from any current entity is extracted, so as to construct an update path starting from any current initial entity;

[0153] S434: Compare the initial path with the updated path, and obtain difference features according to the comparison result;

[0154] For the initial path and the updated path, if the initial path and the updated path structure completely correspond, the current entity in the updated path all exists in the initial path, and the upper and lower order is the same, that is, the upper and lower structure is the same. At this time, the synonym data not included in the initial path is extracted from the updated path as the difference feature;

[0155] If the initial path and the updated path structure do not match and do not contain each other after the preferred words of entities at all levels are aligned with synonym entities, then the synonym data not included in the target path can be extracted as the difference feature.

[0156] S435: The initial lung cancer knowledge graph is updated according to the difference features to obtain the latest lung cancer knowledge graph.

[0157] Among them, the above-mentioned difference features can be synonym data and hyponymous and hyponymous structures. The obtained synonym data and hyponymous and hyponymous structures can be converted into triple data using a preset script, and then the conversion results are imported into the initial knowledge graph to obtain the latest lung cancer knowledge graph.

[0158] After obtaining the latest lung cancer knowledge graph, constructing a second type of lung cancer screening questionnaire in combination with the knowledge content of the latest lung cancer knowledge graph and the question template;

[0159] Wherein, the second type of lung cancer screening questionnaire includes a preset number of latest questions, each of the latest questions is provided with at least two latest options, and each of the latest options is provided with a corresponding score;

[0160] The construction of the second type of lung cancer screening questionnaire is consistent with the construction of the first type of lung cancer screening questionnaire;

[0161] After updating the first type of lung cancer screening questionnaire, the second type of lung cancer screening questionnaire example is as follows:

[0162] Latest Questions:

[0163] 1. For non-smokers, have you lived with a smoking family member for more than 20 years?

[0164] Latest options:

[0165] A. No (skip to 3) - 0 points; B. Yes - 1 point.

[0166] 2. Is the family member currently quitting smoking?

[0167] A. No—2 points; B. Yes—2 points.

[0168] 3. If so, has the family member quit smoking for less than 15 years?

[0169] A. No—1 point; B. Yes—2 points.

[0170] 4. Do you suffer from chronic obstructive pulmonary disease?

[0171] A. No—0 points; B. Yes—1 point.

[0172] 5. Do you have a history of occupational exposure to hazardous substances (one year or more)?

[0173] A. No—0 points; B. Yes—2 points.

[0174] More questions, options, and scores are not described here;

[0175] The second category scores are obtained based on the corresponding scores of the first category of suspected persons / a new batch of test subjects after completing all the latest questions, the second category scores are averaged and compared with the preset scores, and the second category of suspected persons are screened out from the first category of suspected persons / a new batch of test subjects based on the comparison results.

[0176] Among them, the above questions and options are pushed to the subjects, and the corresponding scores of the first category of suspected persons / a new batch of subjects after completing all the latest questions are obtained to obtain the total score (i.e., the second category score); the scores are averaged to obtain A2, and then a preset score value α2 is set. When A2 is greater than α2, the second category of suspected persons are screened out from the first category of suspected persons / a new batch of subjects, wherein α2 is the score set by experts and historical lung cancer judgment standards, and can also be changed according to needs based on the continuous updating of data.

[0177] If the update threshold is not exceeded, S440 is executed:

[0178] S440: Directly determine and send a prompt screening message to the first type of suspected persons.

[0179] S500: Sending a prompt screening message to the first category suspected person / the second category suspected person;

[0180] Among them, it is suggested that the CT imaging information of the first category of suspected persons / the second category of suspected persons needs to be obtained, and the lung cancer risk is subsequently evaluated based on the analysis results of the CT imaging information of the first category of suspected persons / the second category of suspected persons.

[0181] The step of performing a subsequent assessment of the risk of lung cancer based on the analysis results of the CT image information of the first category of suspected persons / the second category of suspected persons includes:

[0182] S501: Construct a lung cancer CT image training set;

[0183] CT images containing tumors were screened out, the tumor areas were manually annotated, and the CT images were cropped to a size of 100*100 to retain the tumor areas to form a lung cancer CT image training set.

[0184] S502: Processing the lung cancer CT image training set to generate required training samples;

[0185] S503: Using the training samples to train a deep learning comparison model based on CT images to generate an image comparison analysis model;

[0186] S504: Acquire CT images of the first category suspected persons / the second category suspected persons as an atlas to be analyzed;

[0187] S505: Processing the atlas to be analyzed through an image comparison analysis model to obtain a further assessment of the risk of lung cancer.

[0188] Performing data analysis on the atlas to be analyzed includes any one or more of the following steps:

[0189] The method comprises the steps of zooming in or out the images in the atlas to be analyzed; rotating the images in the atlas to be analyzed; translating the images in the atlas to be analyzed; and enhancing the contrast of the images in the atlas to be analyzed.

[0190] It is worth mentioning that the purpose of CT examination for the first type of suspected persons / second type of suspected persons is not to determine whether the person has lung cancer, but to further accurately assess the risk of disease. Because from a medical point of view, unless the suspected lesions are punctured and sampled for pathological examination, there is no way to confirm whether it is lung cancer by indirect observation means such as CT. The main purpose of CT examination is to further screen the high-risk population, so as to improve the accuracy of subsequent judgment of high-risk population, and to classify such persons for the convenience of formulating follow-up plans. For example, when the doctor checks the CT images of the first type of suspected persons / second type of suspected persons and finds that the lung shadow is large, it is judged that the possibility of lung cancer is high, and a lung puncture sampling and testing plan is formulated for such persons. When the doctor checks the CT images of the first type of suspected persons / second type of suspected persons and finds that the lung shadow is small, it is judged that the possibility of lung cancer is low (it may be suspected of inflammatory lesions, tuberculosis or benign tumors), a conservative observation plan is formulated for such persons. Therefore, this is only a means to improve the accuracy of risk assessment, and it is not a direct basis for judging lung cancer, but a risk assessment, so as to facilitate subsequent prompts and management of various types of personnel.

[0191] S600: Formulate a follow-up plan for the first category suspected person / the second category suspected person.

[0192] Among them, experts formulate follow-up plans, and the specific content is customized by experts. Experts formulate follow-up plans for Category I suspected persons / Category II suspected persons, including: risk level, treatment plan, notes and some suggestions, etc.

[0193] In a second aspect, this embodiment further discloses a high-risk group of lung cancer prompt screening management system, including:

[0194] The data acquisition module is used to obtain initial lung cancer medical data and construct an initial lung cancer knowledge graph;

[0195] A first grassroots institution screening module is used to construct a first type of lung cancer screening questionnaire according to the initial lung cancer knowledge graph, and screen out a first type of suspected persons according to the first type of lung cancer screening questionnaire;

[0196] A graph updating module, which is used to set a data update cycle, obtain current lung cancer medical data at a corresponding time node, and construct a current lung cancer knowledge graph based on the current lung cancer medical data;

[0197] A second grassroots institution screening module is used to compare the current lung cancer knowledge graph with the initial lung cancer knowledge graph, determine whether to update the first type of lung cancer screening questionnaire, and screen out the second type of suspected persons according to the update result;

[0198] A screening and analysis module, used for sending prompt screening information to the first category of suspected persons / the second category of suspected persons;

[0199] A plan-making module is used to make a follow-up plan for the first category suspected person / the second category suspected person according to the inspection and analysis results.

[0200] In some embodiments, the data acquisition module includes an initial graph construction unit, which is used to manually correct the initial lung cancer medical data and perform entity extraction and relationship extraction according to a dictionary and data properties, construct a first type of triples according to the extraction results, and construct the initial lung cancer knowledge graph according to the first type of triples.

[0201] In some embodiments, the first grassroots institution screening module includes a first questionnaire construction unit and a first screening unit;

[0202] The first questionnaire construction unit is used to construct the first type of lung cancer screening questionnaire in combination with the knowledge content of the initial knowledge graph and the question template;

[0203] The first screening unit is used to obtain a first category score according to the corresponding score of the subject after completing all the initial questions, compare the first category score with the preset score after mean processing, and screen out the first category of suspected persons from the subject according to the comparison result.

[0204] In some embodiments, the atlas update module includes an update period acquisition unit;

[0205] The update cycle acquisition unit is used to obtain the entity update cycle according to the change cycle of the entity in the initial knowledge graph, and to obtain the relationship update cycle according to the change cycle of the relationship in the initial knowledge graph; compare the relationship update cycle with the entity update cycle, and select the update cycle with the shortest time as the current update cycle.

[0206] In some embodiments, the atlas updating module further includes a current data acquisition unit and a current atlas construction unit;

[0207] The current data acquisition unit is used to acquire the current lung cancer medical data of the corresponding time node according to the current update cycle;

[0208] The current graph construction unit is used to perform entity extraction and relationship extraction on the current lung cancer medical data according to the dictionary and data properties, construct a second type of triples according to the extraction results, and construct the current lung cancer knowledge graph according to the second type of triples.

[0209] In some embodiments, the second grassroots institution screening module includes a graph comparison and analysis unit;

[0210] The graph comparison and analysis unit is used to traverse the initial lung cancer knowledge graph and the current lung cancer knowledge graph respectively, extract the corresponding initial entity set and the current entity set; set a similarity scoring function to perform similarity scoring calculation on the entities in the initial entity set and the current entity set, obtain a similarity scoring result, and compare the similarity scoring result with the similarity scoring result according to a preset update threshold. If the update threshold is exceeded, the first type of lung cancer screening questionnaire needs to be updated; if the update threshold is not exceeded, a prompt screening information is directly sent to the first type of suspected person.

[0211] In some embodiments, the second grassroots institution screening module further includes a map updating unit;

[0212] The graph updating unit is used to compare the phrase type of the initial entity with the current entity set to determine the current entity that matches the initial entity; traverse the initial lung cancer knowledge graph to extract an initial path starting from any initial entity in the initial lung cancer knowledge graph; traverse the current lung cancer knowledge graph to extract a path starting from the current entity as an update path; compare the initial path with the update path, and obtain difference features according to the comparison results; update the initial lung cancer knowledge graph according to the difference features to obtain the latest lung cancer knowledge graph.

[0213] In some embodiments, the second grassroots institution screening module further includes a second questionnaire construction unit and a second screening unit;

[0214] The second questionnaire construction unit is used to construct a second type of lung cancer screening questionnaire in combination with the knowledge content of the latest lung cancer knowledge graph and the question template; the second type of lung cancer screening questionnaire includes a preset number of latest questions, each of the latest questions is provided with at least two latest options, and each of the latest options is provided with a corresponding score;

[0215] The second screening unit is used to obtain a second category of scores based on the corresponding scores of the first category of suspected persons / a new batch of test subjects after completing all the latest questions, perform mean processing on the second category of scores and compare them with the preset scores, and screen out the second category of suspected persons from the first category of suspected persons based on the comparison results.

[0216] In some embodiments, the screening analysis module includes a risk assessment unit;

[0217] The risk assessment unit is used to prompt the need to obtain CT image information of the first category suspected person / the second category suspected person, and to perform subsequent assessment of the lung cancer risk based on the analysis results of the CT image information of the first category suspected person / the second category suspected person.

[0218] In a third aspect, an embodiment of the present invention provides a computer device, the computer device comprising: a processor, the processor is used to implement the steps of the method for constructing a knowledge graph as described above when executing a computer program stored in a memory. The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the computer to perform desired functions. The memory may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on a computer-readable storage medium, and the processor may run the program instructions to implement the method steps and / or other desired functions of the various embodiments of the present application described above.

[0219] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method for updating a knowledge graph as described above.

[0220] In addition to the above-mentioned methods and apparatus, the embodiments of the present application may also be a computer program product, which includes computer program instructions. When the computer program instructions are executed by a processor, the processor executes the method steps of various embodiments of the present application.

[0221] The computer program product may be written in one or more programming languages ​​in any combination to write program code for performing the operations of the embodiments of the present invention, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0222] In addition, the embodiments of the present application may also be a computer-readable storage medium, on which computer program instructions are stored, and the computer program instructions, when executed by the processor, cause the processor to execute the method steps of various embodiments of the present application. The computer-readable storage medium may adopt any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may, for example, include but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples (non-exhaustive lists) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the elements. The above description is only a specific embodiment of the present disclosure, so that those skilled in the art can understand or implement the present disclosure. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure will not be limited to these embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for prompting screening and management of a high-risk population for lung cancer, characterized in that: The following steps are involved: Obtain initial lung cancer medical data and build an initial lung cancer knowledge graph, including: Manually correct the initial lung cancer medical data and extract entities and relationships according to the dictionary and data properties, construct a first type of triples according to the extraction results, and construct the initial lung cancer knowledge graph according to the first type of triples; A first type of lung cancer screening questionnaire is constructed according to the initial lung cancer knowledge graph, and a first type of suspected persons are screened out according to the first type of lung cancer screening questionnaire, including: Combining the knowledge content of the initial knowledge graph and the question template to construct the first type of lung cancer screening questionnaire; Obtaining a first category score according to the corresponding scores of the test subject after completing all the initial questions, performing mean processing on the first category score and comparing it with the preset score, and screening out a first category of suspected persons from the test subject according to the comparison result; Setting a data update cycle, the data update cycle includes an entity update cycle and a relationship update cycle, obtaining the entity update cycle according to the change cycle of the entity in the initial knowledge graph, obtaining the relationship update cycle according to the change cycle of the relationship in the initial knowledge graph, comparing the relationship update cycle with the entity update cycle, selecting the shortest update cycle as the current update cycle, obtaining current lung cancer medical data at a corresponding time node, and constructing a current lung cancer knowledge graph according to the current lung cancer medical data, including obtaining the current lung cancer medical data at a corresponding time node according to the current update cycle; Perform entity extraction and relationship extraction on the current lung cancer medical data according to the dictionary and data properties, construct a second type of triples according to the extraction results, and construct the current lung cancer knowledge graph according to the second type of triples; The current lung cancer knowledge graph is compared with the initial lung cancer knowledge graph to determine whether to update the first type of lung cancer screening questionnaire, and the second type of suspected persons are screened out according to the update result, wherein the comparison and determination step includes: Traversing the initial lung cancer knowledge graph and the current lung cancer knowledge graph respectively, and extracting the corresponding initial entity set and current entity set; A similarity scoring function is set to perform similarity scoring calculation on the entities in the initial entity set and the current entity set to obtain a similarity scoring result, and a preset update threshold is used to compare the result with the similarity scoring result. If the update threshold is exceeded, the first type of lung cancer screening questionnaire needs to be updated. If the update threshold is not exceeded, a prompt screening message is directly sent to the first type of suspected persons; It also includes the graph update step: Comparing the phrase type of the initial entity with the current entity set to determine a current entity that matches the initial entity; Traversing the initial lung cancer knowledge graph to extract an initial path starting from any initial entity in the initial lung cancer knowledge graph; Traversing the current lung cancer knowledge graph, extracting a path starting from the current entity as an update path; Comparing the initial path with the updated path, and obtaining difference features according to the comparison result; The initial lung cancer knowledge graph is updated according to the difference features to obtain the latest lung cancer knowledge graph; The step of screening out the second category of suspected persons according to the update results includes: Combining the knowledge content of the latest lung cancer knowledge graph and the question template to construct a second type of lung cancer screening questionnaire; The second type of lung cancer screening questionnaire includes a preset number of latest questions, each of the latest questions is provided with at least two latest options, and each of the latest options is provided with a corresponding score; The second category score is obtained according to the corresponding scores of the first category of suspected persons / a new batch of subjects after completing all the latest questions, the second category score is averaged and compared with the preset score, and the second category of suspected persons is screened out from the first category of suspected persons according to the comparison result; Sending a prompt screening message to the first category of suspected persons / the second category of suspected persons; Develop a follow-up plan for the first category suspected persons / second category suspected persons.

2. A method for prompting screening and management of high-risk groups for lung cancer according to claim 1, characterized in that: The initial lung cancer medical data includes lung cancer medical literature and lung cancer patient case data before the current time point.

3. A method for screening and management of high-risk groups for lung cancer according to claim 2, characterized in that: The current lung cancer medical data includes lung cancer medical literature and lung cancer patient case data between the time point of the previous data acquisition and the current time point.

4. A method for prompting screening and management of high-risk groups for lung cancer according to claim 3, characterized in that: The step of extracting entities from the initial lung cancer medical data / the current lung cancer medical data includes: The entity extraction includes entity extraction of normalized data and entity extraction of non-normalized data; When performing entity extraction on normalized data, entities are extracted directly based on text matching rules; When performing entity extraction on non-normalized data, first mark the parts of speech and extract the nouns that match the field. Then, through text-based rule matching, obtain the entities in the field of lung cancer medicine and add them to the dictionary. Finally, use the dictionary to match the words in the sentence and extract the entities based on the matching results.

5. A method for prompting screening and management of high-risk groups for lung cancer according to claim 4, characterized in that: The step of extracting the relationship from the initial lung cancer medical data / the current lung cancer medical data includes: The relationship extraction includes structured data relationship extraction and unstructured data relationship extraction; Structured data relationship extraction includes relationship extraction between entities and attribute relationship extraction of some entities; Unstructured data relation extraction uses a word vector convolutional neural network-based extraction method to perform relation extraction.

6. A method for prompting screening and management of high-risk groups for lung cancer according to claim 5, characterized in that: The sending of prompt screening information to the first category suspected person / the second category suspected person includes: It is suggested that the CT imaging information of the first category of suspected persons / the second category of suspected persons needs to be obtained, and a subsequent assessment of the risk of lung cancer is performed based on the analysis results of the CT imaging information of the first category of suspected persons / the second category of suspected persons.

7. A method for prompting screening and management of high-risk groups for lung cancer according to claim 6, characterized in that: The subsequent assessment of the risk of lung cancer based on the analysis results of the CT image information of the first category of suspected persons / the second category of suspected persons includes: Construct a lung cancer CT image training set; Processing the lung cancer CT image training set to generate required training samples; Use training samples to train a deep learning comparison model based on CT images and generate an image comparison analysis model; Collecting CT images of the first category of suspected persons / the second category of suspected persons as an atlas to be analyzed; The atlas to be analyzed is processed by an image comparison analysis model to obtain a further assessment of the risk of lung cancer.

8. A high-risk group of lung cancer screening management system, characterized by: include: The data acquisition module is used to obtain initial lung cancer medical data and build an initial lung cancer knowledge graph, including: Manually correct the initial lung cancer medical data and extract entities and relationships according to the dictionary and data properties, construct a first type of triples according to the extraction results, and construct the initial lung cancer knowledge graph according to the first type of triples; The first grassroots institution screening module is used to construct a first type of lung cancer screening questionnaire according to the initial lung cancer knowledge graph, and screen out a first type of suspected persons according to the first type of lung cancer screening questionnaire, including: Combining the knowledge content of the initial knowledge graph and the question template to construct the first type of lung cancer screening questionnaire; Obtaining a first category score according to the corresponding scores of the test subject after completing all the initial questions, performing mean processing on the first category score and comparing it with the preset score, and screening out a first category of suspected persons from the test subject according to the comparison result; A graph update module, used to set a data update cycle, wherein the data update cycle includes an entity update cycle and a relationship update cycle, the entity update cycle is obtained according to the change cycle of the entity in the initial knowledge graph, the relationship update cycle is obtained according to the change cycle of the relationship in the initial knowledge graph, the relationship update cycle is compared with the entity update cycle, the update cycle with the shortest time is selected as the current update cycle, the current lung cancer medical data of the corresponding time node is obtained, and the current lung cancer knowledge graph is constructed according to the current lung cancer medical data, including obtaining the current lung cancer medical data of the corresponding time node according to the current update cycle; Perform entity extraction and relationship extraction on the current lung cancer medical data according to the dictionary and data properties, construct a second type of triples according to the extraction results, and construct the current lung cancer knowledge graph according to the second type of triples; The second grassroots institution screening module is used to compare the current lung cancer knowledge graph with the initial lung cancer knowledge graph to determine whether to update the first type of lung cancer screening questionnaire, and screen out the second type of suspected persons according to the update result, wherein the comparison and judgment step includes: Traversing the initial lung cancer knowledge graph and the current lung cancer knowledge graph respectively, and extracting the corresponding initial entity set and current entity set; A similarity scoring function is set to perform similarity scoring calculation on the entities in the initial entity set and the current entity set to obtain a similarity scoring result, and a preset update threshold is used to compare the result with the similarity scoring result. If the update threshold is exceeded, the first type of lung cancer screening questionnaire needs to be updated. If the update threshold is not exceeded, a prompt screening message is directly sent to the first type of suspected persons; It also includes the graph update step: Comparing the phrase type of the initial entity with the current entity set to determine a current entity that matches the initial entity; Traversing the initial lung cancer knowledge graph to extract an initial path starting from any initial entity in the initial lung cancer knowledge graph; Traversing the current lung cancer knowledge graph, extracting a path starting from the current entity as an update path; Comparing the initial path with the updated path, and obtaining difference features according to the comparison result; The initial lung cancer knowledge graph is updated according to the difference features to obtain the latest lung cancer knowledge graph; The step of screening out the second category of suspected persons according to the update results includes: Combining the knowledge content of the latest lung cancer knowledge graph and the question template to construct a second type of lung cancer screening questionnaire; The second type of lung cancer screening questionnaire includes a preset number of latest questions, each of the latest questions is provided with at least two latest options, and each of the latest options is provided with a corresponding score; The second category score is obtained according to the corresponding scores of the first category of suspected persons / a new batch of subjects after completing all the latest questions, the second category score is averaged and compared with the preset score, and the second category of suspected persons is screened out from the first category of suspected persons according to the comparison result; A screening and analysis module, used for sending prompt screening information to the first category suspected persons / second category suspected persons; A plan-making module is used to make a follow-up plan for the first category suspected person / the second category suspected person.

9. A computer device, comprising: A processor, wherein the processor is configured to implement the steps of the method according to any one of claims 1 to 7 when executing a computer program stored in a memory.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images