Application of Knowledge of TCM Based on the work of the previous 4 chapters, we have obtained important structured knowledge of TCM, and then we will try some related applications based on this knowledge. First of all, in the process of studying clinical experience of well-known doctors of TCM, we propose and construct the basic ontology of TCM to solve the problems of standardization, sharing and acquisition of domain knowledge. Then, we try to find a treatment method based on TCM syndrome differentiation and treatment by combining medical records with the knowledge graph. Next, the 4 classification methods are combined to construct an ensemble model to provide technical support for the diagnosis of hypertension. Finally, we use reinforcement learning methods to infer the paths among entities. According to the diagnostic principles of TCM, the combined reasoning method improves the diagnostic performance. 5.1 Fuzzy Ontology Constructing and its Application in TCM The key to the inheritance and development of TCM is the standardization and normalization of TCM terms. The individuation and complexity of TCM diagnosis and treatment, and the complex structure of TCM knowledge system, obstruct the inheritance and development of TCM in sharing, exchanging and fusion of knowledge. Therefore, during the study of academic thought and clinic experience of famous veteran doctors of TCM, TCM basic ontology has already been proposed and constructed to solve the problems of domain knowledge standardization, sharing and acquisition. 5.1.1 Structure of Fuzzy Ontology In ontology, domain knowledge is formally expressed by concepts that both human and machine can understand. These concepts include entity, attribute, relation and truth, etc. It is used as a standard knowledge representation for the Semantic Web. Nowadays, most of ontology is used to describe concepts and relations quantitatively. However, the concepts and relations are fuzzy and uncertain in many application areas. The classical ontology could not represent information effectively. A feasible solution of this problem is to introduce Fuzzy Logic into ontology to process uncertain information. ·144 · 智能中医信息处理技术与应用（英文版） Domain ontology is a formal expression of domain concepts, which needs both semantics of concepts and relations between concepts. Fuzzy sets are used to describe concepts-terms of TCM. Fuzzy relations describe relations between concepts. By referring to their characteristics in TCM, the following relations are defined: kind-of, attribute-of, and similar-to, etc. Each relation is quantitatively described. Related Definitions is afunction mapping arbitrary μA isuniverse, . : Fuzzy Set. Suppose Definition 5-1 to a value in[0,1], ...... 0, 1.. .... .. . ................ : . ..................Aof, set . is defined as a membership function defined in ... Fuzzy set .. to.......... . , . is named as fuzzy set in can be represented as Eq. (5-1). . ...... . Then is named as a membership degree of μ Au μ Au μ Au ( 1 ) ( 2 ) ( n) A=++.. + (5-1) u1 u2 un All TCM terms (concepts) are described by fuzzy set, i.e. each term corresponds to a fuzzy set. Membership functions are given by domain experts. For example, suppose that the value range of universe U is [30,42], U is body temperature of patients, and Ais fuzzy set of "High Fever". Then A= {0/35+0/35.6+0/36.5+0.1/37+0.5/37.5+0.6/37.8+0.8/38+0.9/38.5+1/39} The above membership degrees are usually determined by expert estimation, or acquired by statistics of mass medical records. is named a fuzzyrelation ......inR: Fuzzy Relation. Fuzzy relationDefinition 5-2 ....... For arbitrary ......Rand S are two fuzzy sets in . Suppose .... . , denoted as .. , if...... . from Uto V , R(u, v)≥S(u, v), then fuzzy relation R includes S . Definition 5-3: Similar-to Relation. There are synonyms and near-synonyms in one domain. The relations between these semantic concepts are the similar to relations of concepts, represented by values in [0,1], where 0 represents no relations between concepts, while 1 represents the same concepts. For instance, " 尿黄" and "尿赤" are the same pathological manifestations. They are the two expressions of one concept. Values between 0 and 1 represent the similarity degree between concepts. For instance, "肝气抑郁" and " 肝脾不调" are two disease types with some similar symptoms. The similarity degree value of the two concepts is 0.6, as shown in Figure 5-1. In order to improve the process speed, the synonymous relation in 5 Application of Knowledge of TCM ·145 · ontology is not marked, and all synonyms are collected into a synonym storehouse. Definition 5-4: Attribute-of Relation. It represents that one concept is an attribute of the other concept. Value in [0,1] is used to describe the importance of the attribute. In TCM ontology, symptom is the attribute of the corresponding disease types. For instance, " 嗳气", "呃逆", "泛酸" and " 胀痛" are attributes of " 肝气抑郁". As shown in Figure 5-3, values are the measure of importance in disease diagnosis and treatment, given by TCM experts. Definition 5-5: Kind-of Relation. It represents membership relation and describes the hierarchy of concepts in ontology. The containing and being contained relations, similar to relations between parent class and child class, can all be viewed as relations between abstraction and concrete of concepts. Values in [0,1] can be used to describe the containing degree of membership. For instance, "肝气抑郁"(0.88), " 肝胃不和"(0.9), "肝脾不调"(0.9), "气滞不瘀"(0.86) are different kinds of liver disease of different degrees. " 肝胃不和" is a kind of " 肝病", and it is also a kind of " 胃病" ,as shown in Figure 5-1. Figure 5-1 Similar-to relation Definition 5-6: Fuzzy Intersection (AND). The intersection of fuzzy sets A and B is represented as relations of more than one attribute. Each attribute is described by a characteristic of the concept. Its value in [0,1] evaluates its importance to the concepts. The relations of different attributes of the concept " 肝气抑郁" are shown as Figure 5-3, including "孤僻", "抑郁" and " 嗳气", etc. Definition 5-7: Fuzzy Union (OR). The union of fuzzy sets A and B represents the relations among many attributes, the maximum value in [0,1] is used in reasoning. Description of Fuzzy Ontology Fuzzy ontology is a marked fuzzy directed graph G, ....,...... .. Where G is a marked fuzzy directed graph, which represents fuzzy ontology here; V is the .. , set of nodes in fuzzy ontology. Suppose ·146 · 智能中医信息处理技术与应用（英文版） .......... ....,..., .. .., .. F is a fuzzy set with universe of V, suppose ..../..,...,.. ../../........ ...... 0≤μi≤1, i=1,2,…,n .... is the membership degree of node vi, representing the fuzzy degree and importance , described bythe following fuzzymatrix: ......is a fuzzy relation in Edegree, etc. ........ … .. . ........….. . ...... .. .. ........….... ........ ........ , . .. .. .... .. .. .. .. Where Nj is the name of node Vj; is membership degree of Vj, representing certainty and fuzzy degree of Vj, etc., as shown in Figure 5-1. The connection between node Vi and Vj can be described as follow: ......,,....,....,.. .... .. .. Where ei,j is the semantic description of the connection between node Vi and Vj; ....,.. and Vj, representing relevant is the membership degree of the connection between node Vi semantic relationship strength, as shown in Figure 5-2. Figure 5-2 Fuzzy ontology Building of Fuzzy Ontology (1) Important terms in domain ontology. List all terms needed when we are defining domain ontology. For instance, list all TCM 5 Application of Knowledge of TCM ·147 · terms that are the concepts described in ontology. (2) Define concepts of domain and hierarchy of concepts. According to description of fuzzy ontology, the terms listed are classified into three levels, respectively describing concepts of the class, the objects in the class, and the attributes of the object. For instance, " 肝病" is a concept of the class describing a kind of disease. The liver disease can be divided into some types, including " 肝气抑郁", "肝胃不和", "肝脾不调", "气 滞血瘀", etc. All of them are concepts of the object. There are different symptoms for each disease type. And these symptoms are the attributes of the object. (3) Establishment of relations between concepts. When constructing domain ontology, we should consider both semantics of concepts and relations between concepts. Fuzzy relations between concepts are described by the relations defined above, and fuzzy values are given by experts. The building process of TCM domain fuzzy ontology is presented, using liver disease as an example. First, according to the requirement analysis and the collection and analysis of information of liver disease in TCM, concepts of liver disease in TCM can be divided into three levels. The first level is the concept of disease class: " 肝病". The second level is the concepts of disease types in the class, including " 肝气抑郁", "肝胃不和", "肝脾不调", "气滞 血瘀", etc. The third level is concepts of some typical symptoms of each disease type. For instance, disease type " 肝胃不和" contains some typical symptoms, including " 嗳气", "呃逆", "泛酸", "胀痛", etc. According to the relations defined above and the collection and analysis of plenty of information of liver disease, it can be concluded that relation eij between disease class " 肝病" and its typical disease " 肝气抑郁", "肝胃不和", "肝脾不调", "气滞血瘀" is similar-to; relation eij between typical disease and symptoms is attribute-of; relation eij between symptoms is OR. Meanwhile, relation of different symptoms of the same disease type is AND. All the membership degree .... . of these relations is given by experience. Using to represent liver disease class, to represent typical disease, □ to represent typical symptoms, and → to represent relations. The words on the line represent relations between the two nodes that the line connects. The value given represents the degree of the relation. The fuzzy ontology of liver disease is shown as Figure 5-3. It is stored in knowledge base. 5.1.2 Application of Fuzzy Ontology The structural description and building process of fuzzy ontology and the reasoning method based on fuzzy ontology are proposed. The application of this method in TCM field is also given. ·148 · 智能中医信息处理技术与应用（英文版） Figure 5-3 Fuzzy ontology of liver disease Fuzzy Ontology-Based Reasoning The reasoning method on fuzzy ontology is described as follows. (1) Constructing fuzzy ontology from problems and known facts. This fuzzy ontology is called the problem ontology. Matching the problem ontology with fuzzy ontology in knowledge base, searching some related nodes Vi, and then calculating their attribute similarity degree SIM, SIM........1.. |...... .... is the certainty degree of nodes in fuzzy ontology in knowledge base, | .... .... (5-2) .. . .. Where is .... the certainty degree of nodes in the problem ontology. .. (2) The certainty of node Vj, the upper-level node of node Vi. When the relation between the two nodes is AND, the calculation formula of the certainty is .. .... .. Σ[SIM( Vi ) ......,...., .. .. 1,2,… , .. (5-3) When the relation between the two nodes is OR, the calculation formula of the certainty is .. .... ......SIM....max ......,...... (5-4) For instance, N7 is the upper-level node of attribute node N2, N3, and N4, the certainty 5 Application of Knowledge of TCM ·149 · . .... ..SIM .. .. (5-5) degree of the node N7 in Figure 5-2 is ....SIM....max..,.......... .... ....IM,S ......,.... ......,.... .. (3) Sort by value of certainty and present the reasoning result. Fuzzy Ontology-Based Reasoning The following knowledge can be derived based on fuzzy ontology as shown in Figure 5-3. The fuzzy reasoning is as follows: (1) IF one patient has symptoms of " 孤僻"(0.3) AND " 抑郁"(0.5) AND " 嗳气"(0.2), THEN "肝气抑郁"(0.96). (2) IF one patient has symptoms of " 嗳气"(0.2) AND " 呃逆"(0.3) AND " 泛酸"(0.3) AND "胀痛"(0.2) OR " 腹胀"l*(0.2),THEN " 肝胃不和"(0.95). (3) IF one patient has symptoms of " 胀痛"(0.3) AND " 腹泻"(0.4) AND " 水谷不化"(0.3), THEN "肝脾不调"(0.98). (4) IF one patient has symptoms of " 肿块"(0.2) AND " 胸胁疼痛"(0.8) AND " 水谷不化 "(0.8), THEN " 气滞血瘀"(0.96). The symptoms in a TCM medical record are " 嗳气" "呃逆" "泛酸" "胀痛" "腹痛". The certainties of the symptoms are 0.88, 0.9, 0.92, 086, 0.75, respectively. The procedure of reasoning is shown as follows: (1) Constructing fuzzy ontology from problems and known facts. The problem ontology of symptoms in a TCM medical record is shown in Figure 5-4. Matching the problem ontology in Figure 5-4 with fuzzy ontology in knowledge base in Figure 5-3, several related attribute nodes are found: " 嗳气" "呃逆" "泛酸" "胀痛" "腹痛". Then the attribute similarity of these nodes are calculated respectively: Similarity of " 呃逆": SIM(呃逆)=1.|0.96.0.9|=0.94; Similarity of " 泛酸": SIM(泛酸)=1.|0.96.0.92|=0.96. Figure 5-4 Symptoms in a TCM medical record (2) Calculating the certainties of nodes " 肝气抑郁" "肝胃不和" "肝脾不调" "气滞血瘀", using attribute nodes " 嗳气" "呃逆" "泛酸" "胀痛" "腹痛". μ(肝气抑郁) = 0×0.3+0×0.5+0.93×0.3=0.28; μ(肝胃不和) = 0.93×0.2+0.94×0.3+0.96×0.3+max(0.88×0.2 ,0.79×0.2)=0.93; ·150 · 智能中医信息处理技术与应用（英文版） μ(肝脾不调) = 0.88×0.3+0×0.4+0×0.3=0.26; μ(气滞血瘀) = 0. (3) Ordering the certainties by descending indicates that the most likely disease type of the patient is " 肝胃不和", and the most unlikely is " 气滞血瘀". 5.1.3 Conclusions The fuzzy ontology building method of liver disease in TCM is discussed. The representation model of fuzzy ontology, and the reasoning based on fuzzy ontology are proposed. An simple application example is also presented. In the future, we will build a relatively complete fuzzy ontology system of liver disease in TCM. TCM, as national culture quintessence of China, plays a significant role in medical and health services in China nation and even all around of the world. But the knowledge acquisition of TCM liver disease meets great difficulties. To solve these difficulties, it needs both profound understanding of liver disease knowledge in TCM, and comprehensive knowledge in computer science. Therefore, the work of exploring better method for TCM knowledge acquisition has great potential and prospect. 5.2 Personalized Diagnostic Modal Discovery of TCM Knowledge Graph Knowledge graph is a new research hotspot in the field of artificial intelligence. TCM knowledge graph can well describe the relationship between symptoms, syndromes, etiology, treatment, prescriptions and so on. 5.2.1 Access to Medical Data and Normalization Most of the experience of the famous TCM doctor is a kind of implicit knowledge. It is found out by the continuous social practice, and very individualized. It is difficult to summarize and refine the experience, if this part of implicit knowledge is explicit, it will play a great role in knowledge inheritance. Medical records are the data records of the practice of knowledge application in TCM. It contains the results of solving practical problems in the application of TCM. We try to find the pattern of treatment according to syndrome differentiation of TCM by combining the medical records with the knowledge graph of the TCM. Through the national 10th and 11th Five-Year plan, we have a large number of high quality medical records of TCM doctors. First, we 5 Application of Knowledge of TCM ·151 · standardize the symptoms, syndromes, medications and other information of each medical record according to the language standard of the TCM knowledge graph, and store the standard medical records. Standardized Medical Record According to the description of each symptom, syndrome, treatment method and TCM in the medical record, and combining the method of semantic similarity calculation, the corresponding terms in the basic knowledge graph of TCM determine its standardized expression and further standardize the medical record words into a set of standard words. For example, a medical record is normalized, symptoms are like "stomach fullness, dull pain, recurrent attacks, exacerbation in the past two weeks." Pain in hunger and postprandial, dull pain, occasionally tingling. The tongue is dark red, yellow and white tongue coat, and the roots are greasy and the veins are deep. It is standardized as "stomachache, belch, dry mouth, swallowing bitterness, gastric anorexia, and dry", and further dismantling as a collection of words {stomach pain, belch, dry mouth, pharynx bitterness, gastric anorexia, and dry stool}; syndrome "interresistance of phlegm and stasis, unbalance of stomach" standardized as "stasis and stomach collateral syndrome", dismantling as a collection of words {syndrome of static blood in stomach collaterals} etc. Standardized Medical Record Storage Structure The standardized medical records are stored in accordance with the structure shown in Table 5-1. Table 5-1 Standardization of medical records storage structure Table Name Symptom in Medical Records / Syndrome in Medical Records / Treatment in Medical Records / Name of Chinese Medicine in Medical Records Data Storage Basic theory database of TCM Column name Data Type Remarks ID varchar ID of medical records Name varchar Symptom / Syndrome /Treatment/ Chinese Medicine Normalization of all medical records of an old Chinese medicine is shown in Table 5-2, Table 5-3, Table 5-4 and Table 5-5. ·152 · 智能中医信息处理技术与应用（英文版） Table 5-2 Partial the complete data basis symptoms ID Name ID Name 5498 Palpitation 5498 Five upset hot 5498 Feverish palms and soles 5498 Dry lips 5498 Dry mouth 5498 Dry mouth and throat 5498 Dry stool 5498 Thirst 5498 Dreaminess 5498 Constipation 5498 Red tongue 5504 Dizziness 5498 Yellow and greasy tongue coating 5504 Headache 5498 Thready pulse # # ID 5498 5498 5504 # Table 5-3 Complete data basis syndrome (Partial) Name Syndrome of endogenous heat due to Yin deficiency Syndrome of deficiency-heat due to the heart-Yin Syndrome of Yin deficiency of liver and kidney # Table 5-4 Complete data basis medical treatment (Partial) ID 5498 5498 5498 5504 # Name Nourishing Yin and clearing heat Soothe the nerves and diazepam palpitation Tranquilizing and sedating the Mind Calm the liver and suppress Yang #