Disclosure of Invention
In view of this, the application provides a disease cognition system based on main symptoms and accompanying symptom words, which aims to solve the technical problem that the accuracy of disease cognition cannot be improved by extracting feature words and Jaccard coefficient similarity theorem for the second time in the prior art.
The technical scheme of the application is realized as follows:
in one aspect, the present application provides a disease cognitive system based on a dominant symptom and an accompanying symptom-like word, the disease cognitive system based on a dominant symptom and an accompanying symptom-like word comprising:
the data acquisition module is used for acquiring disease characteristic information and corresponding symptom characteristic information and establishing a disease knowledge database according to the disease characteristic information and the corresponding symptom characteristic information;
the feature extraction module is used for acquiring clinical symptom feature information to be identified, extracting disease feature words and corresponding symptom feature words from the clinical symptom feature information to be identified through TF-IDF, and acquiring the disease feature words to be identified and the corresponding symptom feature words to be identified according to the disease feature words and the corresponding symptom feature words;
the matching module is used for matching the disease characteristic words to be identified with the disease characteristic information, obtaining the matching similarity between the disease characteristic words to be identified and the disease characteristic information, and searching symptom characteristic information corresponding to the disease characteristic information from the disease knowledge database according to the matching similarity;
the cognition module is used for establishing a Jaccard coefficient similarity algorithm, calculating the similarity between the symptom characteristic words to be identified and the symptom characteristic information through the Jaccard coefficient similarity algorithm, and generating a corresponding cognition report.
On the basis of the above technical solution, preferably, the data acquisition module includes a data set establishment module, configured to acquire disease characteristic information and corresponding symptom characteristic information, where the disease characteristic information includes: disease feature words, symptom feature information includes: the main symptom characteristic words and the accompanying symptom characteristic words, different characteristic word sets are established according to the disease characteristic information and the symptom characteristic information, and the method comprises the following steps: disease feature word set, main symptom feature word set, and accompanying symptom feature word set.
On the basis of the above technical solution, preferably, the data acquisition module further includes a database establishment module, configured to establish a relationship table of disease feature information and corresponding symptom feature information according to a correspondence between the disease feature information and the corresponding symptom feature information, and combine the feature word set with the relationship table to serve as a disease knowledge database.
On the basis of the above technical solution, preferably, the feature extraction module includes a feature word extraction module, configured to obtain feature information of clinical symptoms to be identified, where the feature information of clinical symptoms to be identified includes: the method comprises the steps of extracting appearance frequency data of each term from clinical symptom characteristic information to be identified by using TF-IDF, setting an appearance frequency threshold, comparing the appearance frequency data of each term with the appearance frequency threshold, and taking the term with the appearance frequency data larger than the appearance frequency threshold as characteristic word segmentation.
On the basis of the technical scheme, preferably, the feature extraction module comprises a feature word extraction module, wherein the feature extraction module is used for setting a common word stock, screening feature words according to the common word stock, deleting the screened common words from the corresponding feature words, and reserving the rest feature words as the feature words of the diseases to be identified.
On the basis of the technical scheme, preferably, the matching module comprises a matching calculation module, which is used for setting a matching similarity threshold, carrying out matching calculation on each feature word in the disease feature word set and the disease feature word to be identified, calculating corresponding matching similarity, comparing the matching similarity with the matching similarity threshold, marking the feature word in the disease feature word set when the matching similarity is larger than the matching similarity threshold, and inquiring corresponding symptom feature information from the disease knowledge database through the feature word.
On the basis of the technical scheme, preferably, the cognitive module comprises a calculation cognitive module, wherein the calculation cognitive module is used for establishing a Jaccard coefficient similarity algorithm, setting a similarity threshold value, calculating the similarity between the symptom characteristic words to be identified and the symptom characteristic information through the Jaccard coefficient similarity algorithm, comparing the similarity with the similarity threshold value, and generating a corresponding cognitive report when the similarity is larger than the similarity threshold value.
Still further preferably, the disease-recognizing device based on the dominant symptom and the companion symptom works includes:
the data acquisition unit is used for acquiring disease characteristic information and corresponding symptom characteristic information, and establishing a disease knowledge database according to the disease characteristic information and the corresponding symptom characteristic information;
the feature extraction unit is used for acquiring the feature information of the clinical symptoms to be identified, extracting disease feature words and corresponding symptom feature words from the feature information of the clinical symptoms to be identified through TF-IDF, and acquiring the feature words of the disease to be identified and the corresponding feature words of the symptoms to be identified according to the disease feature words and the corresponding symptom feature words;
the matching unit is used for matching the disease characteristic words to be identified with the disease characteristic information, obtaining the matching similarity between the disease characteristic words to be identified and the disease characteristic information, and searching symptom characteristic information corresponding to the disease characteristic information from the disease knowledge database according to the matching similarity;
the cognition unit is used for establishing a Jaccard coefficient similarity algorithm, calculating the similarity between the symptom characteristic words to be identified and the symptom characteristic information through the Jaccard coefficient similarity algorithm, and generating a corresponding cognition report.
Compared with the prior art, the disease cognitive system based on the main symptoms and the accompanying symptom words has the following beneficial effects:
(1) The method has the advantages that a joint cognition model is constructed through a keyword matching method and the Jaccard coefficient similarity theorem, and the relation between symptom characteristic information to be diagnosed and possible diseases can be predicted or estimated relatively accurately through semantic similarity calculation between clinical symptom amblyopia, so that the cognition accuracy of the system is improved, and meanwhile, the user experience is improved;
(2) The symptom characteristics are self-learned from the existing symptom characteristic information base by adopting an unsupervised mode, semantic similarity calculation between symptom characteristic information word pairs is intelligently carried out, diseases are predicted, workload is reduced, manual interference is avoided, and the method is flexible to realize and high in practicability.
Detailed Description
The following description of the embodiments of the present application will clearly and fully describe the technical aspects of the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, are intended to fall within the scope of the present application.
Referring to fig. 1, fig. 1 is a block diagram showing a first embodiment of a disease cognitive system based on dominant symptoms and accompanying symptomatic terms according to the present application. Wherein the disease cognitive system based on the dominant symptom and the companion symptom idiom comprises: a data acquisition module 10, a feature extraction module 20, a matching module 30 and a cognition module 40.
The data acquisition module 10 is configured to acquire disease characteristic information and corresponding symptom characteristic information, and establish a disease knowledge database according to the disease characteristic information and the corresponding symptom characteristic information;
the feature extraction module 20 is configured to obtain feature information of clinical symptoms to be identified, extract feature words of a disease and corresponding feature words of symptoms from the feature information of clinical symptoms to be identified through TF-IDF, and obtain feature words of the disease to be identified and corresponding feature words of the symptoms to be identified according to the feature words of the disease and the corresponding feature words of the symptoms;
the matching module 30 is configured to match the disease feature word to be identified with the disease feature information, obtain a matching similarity between the disease feature word to be identified and the disease feature information, and search symptom feature information corresponding to the disease feature information from the disease knowledge database according to the matching similarity;
the cognition module 40 is configured to establish a Jaccard coefficient similarity algorithm, calculate similarity between the symptom feature words to be identified and the symptom feature information according to the Jaccard coefficient similarity algorithm, and generate a corresponding cognition report.
It should be understood that the execution subject of the present embodiment may be a processor or controller in a patient or doctor cognitive state processing terminal, or the like.
It should be understood that the scheme of this embodiment is as follows: firstly, according to the related types of diseases, a database of clinical main symptoms and accompanying symptoms characteristic information of a certain type of diseases and specific diseases under the type of diseases and corresponding knowledge of the diseases is established. The library respectively comprises the type of diseases, the vector sets (or first word pairs) of the characteristic information of the clinical main symptoms and the accompanying symptoms of the specific diseases under the type of diseases, the corresponding specific disease knowledge and the like;
secondly, collecting the characteristic information of the clinical symptoms to be identified of the patient by selecting a characteristic item through TF-IDF, extracting a symptom characteristic information vector set (or a second pair of words, wherein the second pair of words possibly contain clinical main symptoms and accompanying symptom characteristic information vector sets) of the clinical symptoms to be identified, and establishing the clinical symptom characteristic information vector set to be identified;
thirdly, performing similarity matching on the main characteristic information vector set (namely the main characteristic keyword set) in the first pair of words and all characteristic information vector sets (namely the characteristic keyword set) in the second pair of words to be identified by using a keyword matching method; if the second word pair inside feature information keywords are similar to or have high probability of being similar to the main feature information vector set (namely the main feature keyword set) inside the first word pair, classifying the second word pair inside feature information keywords into the category of diseases, and performing next cognition; otherwise, not classifying the disease into the category, and stopping the next cognition;
fourth, on the basis of clearly distinguishing the above-mentioned category of diseases, the similarity of the first word to the relationship between the accompanying symptom characteristic information vector set and the second word to be identified accompanying symptom characteristic information set is calculated by using the Jaccard coefficient similarity theorem. If similar, giving the possible disease names and knowledge, and vice versa; finally, weighting is carried out according to the age, sex, constitution and other factors of the patient, a relatively accurate result is calculated and deduced, and a solution is given.
Further, as shown in fig. 2, a block diagram of a second embodiment of the disease cognitive system based on the main symptoms and the accompanying symptom words according to the present application is provided based on the above embodiments, and in this embodiment, the data acquisition module 10 further includes:
a data set establishing module 101, configured to obtain disease characteristic information and corresponding symptom characteristic information, where the disease characteristic information includes: disease feature words, symptom feature information includes: the main symptom characteristic words and the accompanying symptom characteristic words, different characteristic word sets are established according to the disease characteristic information and the symptom characteristic information, and the method comprises the following steps: a set of disease feature words, a set of chief complaint feature words, and a set of concomitant complaint feature words;
the database establishing module 102 is configured to establish a relationship table of disease feature information and corresponding symptom feature information according to the corresponding relationship between the disease feature information and the corresponding symptom feature information, and combine the feature word set and the relationship table to serve as a disease knowledge database;
it should be appreciated that the system will obtain disease profile information, including: disease feature words, symptom feature information includes: the main symptom characteristic words and the accompanying symptom characteristic words, different characteristic word sets are established according to the disease characteristic information and the symptom characteristic information, and the method comprises the following steps: disease feature word set, main symptom feature word set, and accompanying symptom feature word set.
It should be understood that the above steps can also be interpreted as building a database of clinical principal symptom characteristic information and accompanying symptom characteristic information of a certain class of diseases and specific diseases under that class, and corresponding knowledge of the diseases, according to the relevant class of diseases. The library contains the type of disease and the set of feature information vectors (or first word pair) of the clinical main symptoms and accompanying symptoms of the specific disease under the type, the corresponding specific disease knowledge, and the like, respectively.
By way of example, the principal symptom characteristic information of diseases of the stomach (generally referred to as stomach, abdomen, upper digestive tract, etc.) system is mainly: abdominal pain, abdominal distention, belch, acid regurgitation, anorexia, nausea, emesis, abdominal discomfort, stomach pain, inappetence, etc.; if erosive gastritis is said, in addition to the symptom characteristic information described above, accompanying symptom characteristic information is: gastric mucosa multiple punctiform or diffuse hyperemia, erosion, black manure, syncope, shock, etc.; in the case of acute suppurative gastritis, in addition to the symptom characteristic information, accompanying symptom characteristic information is: acute onset, bacterial infection of stomach wall, frequent chill and high fever, pain in upper abdomen, chill and fever, vomit as pus blood sample, hematochezia, etc.; in the case of a pharmaceutical gastric disease, besides the symptom characteristic information, there are misuse of the drug, gastric symptoms occurring during the drug administration, and the like accompanying the symptom characteristic information. For another example, in addition to the symptom characteristic information, accompanying symptom characteristic information is: excessive eating, eating uncooked food, etc.
It should be understood that, after that, the system establishes a relationship table of disease feature information and corresponding symptom feature information according to the corresponding relationship between the disease feature information and the corresponding symptom feature information, combines the feature word set with the relationship table, and the main combination mode is a simple association relationship, and can find the corresponding feature word from the relationship table through the feature words in the feature word set, as a disease knowledge database.
Further, as shown in fig. 3, a block diagram of a third embodiment of the disease cognitive system based on the main symptoms and the accompanying symptom words according to the present application is provided based on the above embodiments, and in this embodiment, the feature extraction module 20 further includes:
the feature word extraction module 201 is configured to obtain feature information of clinical symptoms to be identified, where the feature information of clinical symptoms to be identified includes: the method comprises the steps of extracting appearance frequency data of each term from clinical symptom characteristic information to be identified by using TF-IDF, setting an appearance frequency threshold, comparing the appearance frequency data of each term with the appearance frequency threshold, and taking the term with the appearance frequency data larger than the appearance frequency threshold as characteristic word segmentation.
The feature word extraction module 202 is configured to set a common word stock, screen feature words according to the common word stock, delete the screened common words from the corresponding feature words, and reserve the remaining feature words as feature words of the disease to be identified.
It should be appreciated that the system will obtain clinical symptom characteristic information to be identified, including: and extracting the appearance frequency data of each entry from the clinical symptom characteristic information to be identified by using the TF-IDF, setting an appearance frequency threshold value, comparing the appearance frequency data of each entry with the appearance frequency threshold value, and taking the entry with the appearance frequency data larger than the appearance frequency threshold value as a characteristic word segmentation.
It should be understood that this step is to collect the characteristic information of the clinical symptoms to be identified by using TF-IDF to select the characteristic item, extract the characteristic information vector set (or second pair of words, which may contain the characteristic information vector set of the clinical main symptoms and the accompanying symptoms) of the clinical symptoms to be identified, and establish the characteristic information vector set of the clinical symptoms to be identified. The main ideas of TF-IDF are: if a word appears in one article with a high frequency TF and in other articles with few occurrences, the word or phrase is considered to have good category discrimination and is suitable for classification. The Term Frequency (TF) represents the frequency with which terms (keywords) appear in text. This number will typically be normalized (typically word frequency divided by the total number of articles) to prevent it from biasing toward long documents.
It should be understood that, because the articles and the number of terms in the embodiment are fewer, there may be a certain error in the obtained feature word, so in this embodiment, a common word stock is set, the feature word is screened according to the common word stock, the screened common word is deleted from the corresponding feature word, and the remaining feature word is reserved as the feature word of the disease to be identified, so that the feature word is further extracted in this way, and an accurate feature word is obtained for subsequent calculation, so that the accuracy of the system can be improved.
For example, a short text is described as follows: itching of the skin is usually followed by pimples, wheal, bright red or pale skin, and oedema erythema in a few patients. The skin is subject to a large, volt-like patch, which occurs rapidly and resolves, and a large, rubella-like mass, which resolves rapidly with severe itching. Rashes repeatedly occur in batches and are seen by the evening authors. The feature information vector set extracted and de-duplicated through word segmentation technology is as follows: erythema, papules, wheal, associated with intense itching; rapid onset and rapid regression; repeated batch occurrences; the evening authors see more, these are feature words, and then the feature words obtained by further extracting these feature words are: erythema, pimples, wheal, severe itching, rapid onset, rapid regression, repeated, batch onset, and frequent evening authors.
Further, as shown in fig. 4, a block diagram of a fourth embodiment of the disease cognitive system based on the main symptoms and the accompanying symptom words according to the present application is proposed based on the above embodiments, and in this embodiment, the matching module 30 includes:
the matching calculation module 301 is configured to set a matching similarity threshold, perform matching calculation on each feature word in the disease feature word set and the feature word of the disease to be identified, calculate a corresponding matching similarity, compare the matching similarity with the matching similarity threshold, mark the feature word in the disease feature word set when the matching similarity is greater than the matching similarity threshold, and query corresponding symptom feature information from the disease knowledge database through the feature word.
It should be understood that, finally, the system will use a keyword matching method to perform similarity matching on the main feature information vector set (i.e. the main feature keyword set, because the clinical main symptom feature information often represents the main feature information of a certain disease) in the first pair of words and all feature information vector sets (i.e. feature keyword sets) in the second pair of words to be identified; for convenience of comparison, the system sets a matching similarity threshold, judges the similarity probability through the similarity threshold, and if the feature information keywords in the second word pair are similar to or have high similarity probability with the main feature information vector set (i.e. the main feature keyword set) in the first word pair, classifies the second word pair as the category of diseases, and carries out the next cognition; otherwise, the disease is not classified as the disease, and the next cognition is stopped.
For example, there are many categories of dermatological disorders: (1) dermatological disorders with erythema, papules, and cellulite as the primary cause; (2) dermatological disorders with proliferative papules and nodules as the primary; (3) dermatological disorders with non-neoplastic nodules and plaque as a primary component; (4) skin disorders, mainly papular herpes, herpes simplex; (5) pustule-based dermatological diseases, and the like. The method comprises the steps that after a database of main feature information vector sets (namely main feature keyword sets) in a first pair of words is taken, and similarity matching is carried out on all feature information vector sets (namely feature keyword sets) in a second word to be identified, the method can find that: their common main characteristic information: erythema, papules, wheal, etc., then, of course, the characteristic information to be identified is assigned to the category of dermatological disorders in which erythema, papules, wheal predominate.
It should be understood that the main methods of word or phrase recognition and word or phrase tagging in the text description herein are: matching the word in the word or phrase with the word or phrase in the disease word lexicon, if the matching is successful, marking the word or phrase as the disease word or phrase, otherwise, marking the word or phrase as the descriptive word or phrase; the descriptive word or phrase herein refers to a disease symptom descriptive word or phrase; the disease word stock comprises words or phrases of each category and each disease symptom (or sign) in the motion system, the digestive system, the respiratory system, the urinary system, the reproductive system, the endocrine system, the immune system, the nervous system and the circulatory system of the human body.
Further, as shown in fig. 5, a block diagram of a fifth embodiment of the disease cognitive system based on the main symptoms and the accompanying symptom words according to the present application is proposed based on the above embodiments, in which the cognitive module 40 includes:
the computing and cognition module 401 is configured to establish a Jaccard coefficient similarity algorithm, set a similarity threshold, calculate similarity between the symptom feature word to be identified and the symptom feature information through the Jaccard coefficient similarity algorithm, compare the similarity with the similarity threshold, and generate a corresponding cognition report when the similarity is greater than the similarity threshold.
It should be understood that, finally, on the basis of definitely distinguishing the above-mentioned diseases, the similarity of the first word to the relationship between the feature information vector set of the accompanying symptoms and the feature information set of the second word to be identified is calculated by using the Jaccard coefficient similarity theorem. If similar, the possible disease names and knowledge are given, and vice versa. For example, the above-mentioned identification information has been judged to be classified into the skin diseases mainly including erythema, pimple, and wind mass. Then, only "severe itching" is calculated; rapid onset and rapid regression; repeated batch occurrences; the similarity of accompanying symptoms such as "frequent by evening authors" is sufficient. Here, what are the specific diseases of the skin with erythema, papules, wind-mass predominance? Such as urticaria, contact dermatitis, drug eruptions, erythema multiforme, erythema annulare, erythema multiforme, lupus erythematosus, erysipelas, and the like. The result of calculation using the Jaccard coefficient similarity theorem is: the skin disease is urticaria. What is then urticaria? Urticaria is commonly known as rubella. Urticaria is a skin rash like a wheal, which is a large plaque of skin, which is caused by skin allergy, and the rash is rapid to resolve and massive like a rubella block, which is caused by skin allergy and is accompanied by severe pruritus. And then the system sets a similarity threshold, calculates the similarity between the symptom characteristic words to be identified and the symptom characteristic information through a Jaccard coefficient similarity algorithm, compares the similarity with the similarity threshold, and generates a corresponding cognitive report when the similarity is larger than the similarity threshold.
It should be understood that the Jaccard coefficient is the proportion of the intersection elements of two sets a and B at A, B and collectively, referred to as the Jaccard coefficient for both sets. The Jaccard coefficient is equal to the ratio of the number of sample set intersections to the number of sample set union, denoted by J (a, B). Given two sets a, B, each contains n common attributes, each of which takes a value of 0 or 1. Specifically, the method comprises the following steps: given two sets A, B, the Jaccard coefficient is defined as the ratio of the size of the A and B intersection to the size of the A and B union, J (A, B) ∈ [0,1], defined as follows:
when both sets A, B are empty, J (A, B) is defined as 1. The index related to the Jaccard coefficient is called the Jaccard distance, and is used to describe dissimilarity between sets. The greater the Jaccard distance, the lower the sample similarity. The formula is defined as follows:
wherein, for spread (symmetric difference): aΔb= |a u b| -a n b|, such as: m is M 11 The attribute values of the sets A and B are 1; m is M 01 A number that is 1 for the attribute value of sample a and 0 for the attribute value of sample B; m is M 10 The attribute value of the sample A is 0, and the attribute value of the sample B is 1; m is M 00 The attribute values for samples a and B are all 0. The following conclusion is drawn:
M 11 +M 01 +M 10 +M 00 =n,J=M 11 /(M 01 +M 10 +M 00 );
if sets a, B are empty sets, J (a, B) =1 is defined. Obviously, J (A, B) is more than or equal to 0 and less than or equal to 1. The larger the J value, the greater the similarity of the two samples. The N-dimensional vector refers to the N-dimensional characteristics of the samples, forming a set. And the set is composed of elements, and if the sample has the characteristic in the corresponding characteristic position, the value of the position set is 1 to indicate that the element is contained; otherwise, take 0, indicating that the element is not included. As can be seen, element = feature.
It should be noted that the foregoing is merely illustrative, and does not limit the technical solution of the present application in any way.
As can be seen from the above description, the present embodiment proposes a disease cognitive system based on main symptoms and accompanying symptomatic words, including: the data acquisition module is used for acquiring disease characteristic information and corresponding symptom characteristic information and establishing a disease knowledge database according to the disease characteristic information and the corresponding symptom characteristic information; the feature extraction module is used for acquiring clinical symptom feature information to be identified, extracting disease feature words and corresponding symptom feature words from the clinical symptom feature information to be identified through TF-IDF, and acquiring the disease feature words to be identified and the corresponding symptom feature words to be identified according to the disease feature words and the corresponding symptom feature words; the matching module is used for matching the disease characteristic words to be identified with the disease characteristic information, obtaining the matching similarity between the disease characteristic words to be identified and the disease characteristic information, and searching symptom characteristic information corresponding to the disease characteristic information from the disease knowledge database according to the matching similarity; the cognition module is used for establishing a Jaccard coefficient similarity algorithm, calculating the similarity between the symptom characteristic words to be identified and the symptom characteristic information through the Jaccard coefficient similarity algorithm, and generating a corresponding cognition report. According to the method, the joint cognition model is constructed through the keyword matching method and the Jaccard coefficient similarity theorem, the diseases to be identified are accurately cognized through the joint cognition model, the identification accuracy of the system is improved, and the user experience is improved.
In addition, the embodiment of the application also provides disease cognitive equipment based on the main symptoms and the accompanying symptom scholaryngitis. As shown in fig. 6, the disease cognitive device based on the dominant symptom and the companion symptom works includes: a data acquisition unit 10, a feature extraction unit 20, a matching unit 30, and a recognition unit 40.
A data acquisition unit 10, configured to acquire disease feature information and corresponding symptom feature information, and build a disease knowledge database according to the disease feature information and the corresponding symptom feature information;
the feature extraction unit 20 is configured to obtain feature information of clinical symptoms to be identified, extract feature words of a disease and corresponding feature words of symptoms from the feature information of clinical symptoms to be identified through TF-IDF, and obtain feature words of the disease to be identified and corresponding feature words of the symptoms to be identified according to the feature words of the disease and the corresponding feature words of the symptoms;
a matching unit 30, configured to match the disease feature word to be identified with the disease feature information, obtain a matching similarity between the disease feature word to be identified and the disease feature information, and search symptom feature information corresponding to the disease feature information from the disease knowledge database according to the matching similarity;
and the cognition unit 40 is configured to establish a Jaccard coefficient similarity algorithm, calculate similarity between the symptom feature words to be identified and the symptom feature information through the Jaccard coefficient similarity algorithm, and generate a corresponding cognition report.
In addition, it should be noted that the above embodiment of the apparatus is merely illustrative, and does not limit the scope of the present application, and in practical application, a person skilled in the art may select some or all modules according to actual needs to achieve the purpose of the embodiment, which is not limited herein.
In addition, technical details not described in detail in this embodiment may be referred to the disease cognitive system based on the dominant symptom and the accompanying symptom like words provided in any embodiment of the present application, and will not be described herein.
The foregoing description of the preferred embodiments of the application is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the application.