CC BY
22
CLINICAL MEDICINE
УДК 004.89
SOFTWARE SYSTEM FOR DIAGNOSING SPINAL DISEASES USING CASE-BASED REASONING
Dmitry Evgenyevich PALCHUNOV12, Gulnara Erkinovna YAKHYAEVA2, Olga Vladimirovna YASINSKAYA2
1 Sobolev Institute of Mathematics
630090, Novosibirsk, Academician Koptyug av., 4
2 Novosibirsk State University 630090, Novosibirsk, Pirogov str., 2
This paper describes the Diagnostic Panel Software System designed for the domain of spinal deformity and degenerative spinal disease. The work has been based on the logical methods of processing of the data obtained from natural language medical records. The software system is based on a patient's clinical and laboratory test results that help the physicians to determine an initial diagnosis and obtain immediate information for the necessary analysis for a final diagnosis and to select the optimal treatment plan. The Diagnostic Panel Software System uses a case-based approach to presenting the information from the variety of native language sources. The model theoretic methods of domain ontology construction has been used to develop the case-based approach. The Formal Concept Analysis methodology has been implemented to process the represented information/data in the system.
Keywords: diagnosis; spine diseases; case-based model; fuzzy model; ontology; formal context; formal concept.
Degenerative-dystrophic diseases of the spine are among the most pressing problems of the present time. This is the most common chronic disease characterized by limitation of physical activity and pain, which is experienced by almost every adult. Numerous literature data indicate a steady increase in the number of patients with diseases of the spine
[1, 15].
Avoiding errors in diagnosis and timely treatment assignment require consistency in the conduct of examination of patients with diseases of the spine. Final diagnosis is made using clinical, laboratory and instrumental research methods.
Instrumental research methods play a decisive role in the diagnosis of the spine diseases [5]. For today there are many methods for the diagnosis of various deformations and degenerative diseases of the spine. However, doctors (especially outpatient care, as well as working in small towns and district hospitals) are daily confronted with the problem of diagnosis, determination of necessary diagnostic procedures and consultations of other specialists. The doctor needs to determine the preliminary (wor-
king) diagnosis based on clinical and laboratory studies and direct the patient to specific instrumental investigations.
Therefore, it is critical and economically feasible to develop such a software system that would allow doctors using the statistical data to determine a preliminary diagnosis and to quickly obtain information about the necessity of those or other instrumental diagnostic procedures for the staging the final diagnosis and selecting the optimal treatment strategy.
In this paper we describe the «Diagnostic Panel» software system designed for «spinal deformity and degenerative diseases of the spine» subject domain. The system is based on statistical processing of medical records of patients treated at the Novosibirsk Research Institute of Traumatology and Orthopedics (NRITO) n. a. Y.L. Tsivyan. The developed software system uses case-based approach for the representation of knowledge extracted from a variety of natural language texts (medical histories) [12, 17]. Case-based approach is based on model-theoretic methods of subject domain ontologies for-
Palchunov D. - doctor ofphysical mathematical sciences, professor, head of the chair for general informatics, leading researcher, e-mail: palch@math.nsc.ru
Yakhyaeva G.E. - candidate of physical mathematical sciences, docent, e-mail: gul_nara@mail.ru Yasinskaya O.V. - lecturer assistant, e-mail: yasinskaya.olga@gmail.com
malization [8]. For the processing of the knowledge represented in system the formal concept analysis methodology is used [3].
MODEL-THEORETIC FORMALIZATION OF
SUBJECT DOMAIN
With model-theoretic point of view, the description of the ontology is the description of the subject domain signature (i. e. describing the concept set of that subject domain) and setting the analytic theory of the subject domain (i.e. describing the implicit and explicit definitions of the subject domain) [9].
Seven classes of attributes were established to create the concept set of subject domain A = «Spinal deformity and degenerative diseases of the spine»:
1) Pi: «Gender»;
2) P2: «Age group»;
3) P3: «Laboratory research»;
4) P4: «Complaints on admission»;
5) P5: «Clinical research»;
6) Q: «Instrumental research»;
7) D: «Diagnoses».
Attribute class «Gender» is composed of two concepts: «Male» and «Female» (Table). Attribute class «Age group» is composed of eight concepts: «0-9 years», «10-19 years», «20-29 years», «3039 years», «40-49 years», «50-59 years», «60-69 years», «70 years and older». Class «Laboratory research» has a hierarchical structure and is divided into three subclasses: clinical tests, biochemical tests, genetic tests. Attribute set for the class «Laboratory research» was formed under the Ingerleib reference book [4]. Attribute sets for the classes «Complaints on admission» and «Clinical research» were formed as a result of parsing the medical histories of individual patients. Attribute set for the class «Instrumental research» is designed according to the National handbook of Orthopedics [6]. Attribute set for the class «Diagnoses» is selected in accordance with the 10th revision of the International Classification of Diseases (ICD-10) [14], class 13 (code M). This class also has a hierarchical structure and is
divided into subclasses: M40-M43 Deforming dor-sopathies, M45-M49 Spondylopathy, M50-M54 Other dorsopathies.
With the model-theoretic point of view, the above attributes are the one-place predicates and form a signature oa of considered in this work subject domain A. We will call them as signature predicates. Thereby,
OA = Pi u u P2 U P3 U P4 U P5 U Q U D.
Denote by S(oa) the set of all one-place predicates definable by quantifier-free formulas of signature oa. For simplicity, the elements of S(oa) will simply be called formula predicates. Note that each formula predicate is a boolean combination of signature predicates. It's obvious that cA ç S(cta).
To describe the subject domain ontology, a finite set of axioms A x(A) ç S(oa) is defined. For the given subject domain A introduce axioms of three types: general-private axioms, axioms of exclusion and the axioms of completeness.
General-private axioms. Hierarchical ordering of concept classes «Laboratory research» and «Diagnoses» should be reflected axiomatically. The scheme of such axioms is the following:
Pi(x) ^ P2(x).
For example: «If M48.0 Spinal stenosis, then M48 Stenosis».
Axioms of exclusion. The concepts of classes «Gender» and «Age group», as well as some groups of concepts from the class «Laboratory research», are mutually exclusive. The scheme of such axioms is the following:
Pi(x) ^ P2(x).
For example: «If Male, then not Female» or «If The level of hemoglobin in the blood exceeds the norm, then it is not true that The level of hemoglobin in the blood is normal».
Axioms of completeness. For the subject domain description we consider medical histories of the patients that passed the full cycle of diagnostics, for
Table
Examples of attributes for each of the above classes
Attribute class Example of attribute
Gender «Male»
Age group «0-9 years»
Laboratory research «The level of hemoglobin in the blood exceeds the norm»
Clinical research «Palpation of the spinous process is painful in the projection L4-S1»
Complaints on admission «Pain in the lumbar spine»
Instrumental research «MRI of the cervical spine»
Diagnoses «M48.0 Spinal stenosis»
which a final diagnosis is known and which treated appropriately. Therefore, we believe that at least one attribute of each of the seven classes is reflected in every considered medical history. Thus, we have seven axioms of completeness:
VP(x), where Y e{Pj, P2, P3, P4, P5, Q, D}.
PeY
Note that the ordered pair (ca, Ax(A) forms an ontology of subject domain A.
Let us now consider a finite set (e^...,en} of medical histories, i.e. a set of semi-structured texts written in natural language. Note that each medical history clearly shows all performed diagnostic studies, their results and the final diagnosis for the patient. Therefore, for each medical history ei we can describe a set of attributes (signature predicates) that is true on ei. Thus, for each medical history ei we build a singleton model ({e¿ }ca) which we call a case of subject domain A. Denote by E = (ej,..., en} the class of cases generated by a set of medical histories (ej,..., en}.
This, in turn, will allow us to build an ontologi-cal model Aa E, ca^ of subject domain A generated by the set of cases E. In ontological model Aa for each signature predicate P(x) e cA and for each case e e E we have Aa N P(e) if and only if the predicate P(e) is true on case e.
Note that if the predicate P(x) e S(oA) belongs to the axiom set of subject domain A (i. e. P(x) e Ax(A), then for any e e E holds Aa N P(e). Thus, the ontology of this subject domain is true on ontological model.
To solve the problems of statistical data processing we need both case-based and fuzzy models of considered subject domain [19]. These models are based on the ontological model.
For further consideration we need the concept of the case-based model which is a special case of a Boolean valued model.
Definition 1 [10]. Let B be a full Boolean algebra and t: S(oa) ^ B. Then the ordered triple AT = (A, c, t) is called a Boolean valued model if truth function t is closed under logical operations.
Definition 2. The ordered triple AtE ^ ^ ({a}, ca , xE) is called a case-based model of subject domain A, generated by ontological model Aa = (E, ca) , if for any predicate P(x) e S(oA) we have
x£(P(a)) = (e e E | AAP(e)}.
In case-based model each predicate is associated with set of cases for which the predicate is true. Thus, by a set of cases E we define a Boolean valued model AE. In this Boolean valued model each sentence of signature oA U (ca) is associated with the element of Boolean algebra p(E).
This description is based on the following result.
Theorem of duality [10]. Let B be a complete atomic Boolean algebra, AB - Boolean valued model, E = {Ab =\b e At (B)} and AE - case-based model. Then AB = AE.
Most methods of statistical data processing use objective and/or subjective probabilities. The objective probability refers to the relative frequency of occurrence of any event in the total number of observations, or the ratio of favorable outcomes to the total number of observations. The subjective probability refers to a measure of confidence of some expert or group of experts that this event will actually take place. In this approach the concept of fuzzy model is used to describe the objective probabilities.
Definition 3. The ordered triple A^E ^ ^ ({a}, aA, is called a fuzzy model of subject domain A, generated by ontological model Aa = = (E, ca) , if for any predicate P(x) e S(oA) we have
me<9) = II {e e E | Aa N P(e)} ||/||E||.
In the fuzzy model the truth values of the sentences (concepts) are numbers from the interval [0, 1], which reflect the objective probability that a randomly selected case has a particular concept. A more detailed description of the properties of both case-based and fuzzy models can be found in [16, 11, 12].
KNOWLEDGE PROCESSING ALGORITHMS
A formal description of the preliminary
diagnosis
Formal concept analysis technique (FCA) was used for a formal description of the preliminary diagnosis. Formal concept analysis is an applied branch of the algebraic theory of lattices. For today the FCA is one of the most powerful data mining techniques. More information on this trend can be found in the works [3, 2, 13].
The central concept of the FCA is the notion of formal context. With model-theoretic point of view, formal context is defined by the class of models K c K(o) of fixed signature o and the set of sentences S c S(o) of the same signature and is an ordered triple <K, S, N) [7].
In this paper we consider the formal context KA = <E, oA, N) generated by ontological model Aa.
Let Ae h A^ be a case-based and fuzzy models generated by ontological model Aa. Then the pair of sets (A, B), such that A c E, A c oA, is a formal concept of the context KA, if the following conditions are met:
m-e (&9(x)eB^(a)) > |xe (y(a) & (&^x)e9(a))),
for each y( x)<
s К).
B '
A = te (&9(x)es9(a)),
where {a} is a basic set of fuzzy model A^.
The set B is called the content of the formal concept (A, B). For convenience, we call the formula (&9(X)eB^(x)) (instead of set B) as the content of the formal concept (A, B).
The formal concept (A1, B1) is called a more general concept than the concept (A2, B2) (denoted (A1, B1) □ (A2, B2)) if A2 c A1. Note that if (A1, B1) □ (A2, B2), then B1 c B2.
Consider a subset of the set of signature predicates P = P1 U P2 U P3 U P4 U P5. To get a preliminary diagnosis we will use formal context KP (E, P, N), which is a subcontext of context KA.
The formal concept (A, B) of the context KP we call positive hypothesis for the diagnosis D(x) e D, if the condition is met
^E (&9(x)eB9(a) ^ D(a)) = 1.
Proposition 1. If |xE (91(a) ^ D(a)) = 1 and ^e (92(a) ^ D(a)) = 1, then ^e (91(a) v 92(a)) ^ ^ D(a)) = 1.
Proposition 2. Let (A1, B1) and (A2, B2) be a formal concepts of the context KP such that (A1, B1) □ (A2, B2). Then if concept A B1) is a positive hypothesis of a diagnosis D(x) e D, then (A2, B2) also is a positive hypothesis for the same diagnosis.
Let G(D) be the set of all positive hypotheses for the diagnosis D. We define the set Gmax(D) c c G(D) of maximal positive hypotheses for diagnosis D(x), i. e. such that for all concepts (A, B) e e Gmax(D) there is no more general concept that belongs to the set G(D).
Then the formula
FD (x) = V &<Kx)eB Mx),
A, BeGmax (D)
will be called a formula description of diagnosis D(x).
The algorithm for determining
the working diagnosis
Consider patient Pat. Suppose that a partial diagnostics of the patient Pat was made and now we need to get a preliminary diagnosis. Consequently, there is information about the truth of some, but perhaps not all, predicates of the set P. Denote by True(Pat) the set of signature predicates from the set P, whose truth is known for patient Pat. Denote by Th(Pat) closure with respect to deducibility of the set True(Pat) (i.e. the theory generated by the
set True(Pat)). According to the theory Th(Pat) we will build a model of the patient Pat.
Definition 4. The ordered triple APat = ({Pat}, oA, nPat) is called a fuzzy model of the patient Pat, if for any predicate 9(x) e S(oA) truth function nPat is defined as follows:
n(Pat) ((Pat)) =
1, ф(x) e Th(Pat); 0, — ф(х) e Th(Pat); [0,1], otherwise.
Model APat is a generalized fuzzy model of signature oA. The formal definition and description of the properties of these models can be found in [18].
Further, to determine a preliminary diagnosis (or several preliminary diagnosis), we need to check on model APat the truth of formula descriptions FD(x) of all diagnoses D of the set D. Diagnoses for which the condition nPat(FD(Pat)) = 1 is met are declared a preliminary diagnosis for the patient.
However, there may be a situation where the working diagnosis is not defined, i.e. for any D(x) e e D we have nPat(FD(Pat)) ^ 1. Then, if there is at least one diagnosis D(x) such that nPat(FD(Pat)) = = [0, 1], then the system offers to make an additional examination of the patient. If there is a situation where for any D(x) e D we have nPat(FD(Pat)) = = 0, then we are dealing with unusual situation, i.e. it is impossible to diagnose this patient using the developed system.
Algorithm of appointment of additional
diagnosing
Assume that during the initial examination of the patient Pat preliminary diagnoses D1(x), ..., Dk(x) e D were set. A further objective of the system is to select the most appropriate tools of instrumental research for setting final diagnosis.
Select a subset of cases E' from the set of cases E for which at least one of diagnoses D1(x), ..., Dk(x) was defined, i.e.
E' = {e e E | Aa N D^e) v ... v Aa N Dk(e)}.
Consider the formal context Kq = (E', Q, N) which is a subcontext of context KA. The content of this context is the set of tools of instrumental research. In this context we seek the most general formal concept (A, B). The content of the concept B will be considered as optimal set of instrumental research tools based on working diagnoses D1(x), ., Dk(x).
Note that the relation □ - «to be the most general notion» is a partial order relation. Therefore, there may be not the only one largest concept, but the several maximal concepts. Let concepts (A1, B1), ., (Al, Bl) be the maximal formal concepts of con-
text Kq with ordering □ . Then the system offers alternative solutions: B1, ..., Bl.
It is obvious that the set of maximal formal concepts (A1, Bi), (Al, Bi) has the following properties:
Ai U ... U Ai = E';
For any formal concept (A, B) of context Kq there is i = 1, ., l such that Bt c B.
Thus, by offering a set of alternative solutions B1, ..., Bl on one hand we provide coverage of all considered cases, and on the other hand we minimize the amount of instrumental research tools required for setting the final diagnosis.
DESCRIPTION OF THE «DIAGNOSTIC PANEL»
SYSTEM
In view of the anticipated program usage scenarios, it was decided to conduct the development in the form of a web application built on the ASP.NET platform. ASP.NET MVC framework was used as an architectural template of web application. Using the MS SQL Server (Express) database applies to be the best option in this case.
To describe the medical histories in the MS SQL Server 2014 database 12 tables were created and relationships between tables were organized. They describe the 7 categories of considered attributes of medical histories - Gender, Age group, Diagnosis, Complaints, Clinical research, Laboratory research, and Instrumental research tools. To replenish the database information on new medical histories and view existing in a web application the «List of medical histories» page was developed. On this page the standard operations on the data were im-
plemented such as view, create, edit, and delete. The design of the page for editing the list of medical histories is shown in Figure 1.
The interface of the main page is a form to fill in the data of a new medical history. The user fills in five categories of attributes: Gender, Age group, Complaints, Clinical research, and Laboratory research. Attributes of Gender, Age group and Complaints filled via drop-down lists. Attributes of Clinical research and Laboratory research are filled with the help of checkbox-lists. The appearance of data input form is shown in Figure 2.
The button «Find working diagnoses» starts performing the main algorithm for a given information in the form. Building a hypothesis is based on medical histories stored in the MS SQL Server database.
The result of the main algorithm work is a list of preliminary (working) diagnoses. Under the list of working diagnoses the table of instrumental research tools needed for clarification of diagnosis is shown. An example of the results of the main algorithm work is shown in Figure 3.
CONCLUSIONS
The paper describes developed methods for identifying the need for additional diagnostic tests for definitive diagnosis of the patient. These methods are based on a combination of the two methodologies: a case-based approach of knowledge representation and formal concept analysis.
For the formalization of subject domain knowledge a finite set of medical histories of patients is used, i.e. set of semi-structured texts written in natural language. For each medical history a singleton
ниито Главная Список истории болезни
Истории болезни
Создать
Диагноз Пол Возрастная Жалобы Первичный осмотр Инструментальные
категория средства
М41.1 Юношеский женский 30-39 на деформацию позвоночника Правосторонний Рентген ография Редактировать
идиопэтический сколиоз на боли в грудном отделе позвоночника грудной кифосколиоз позвоночника | Подробнее |
при физической нагрузке (спондилография) Удалить
на боли в поясничном отделе МРТ грудного отдела
позвоночника при физической нагрузке позвоночника
М41.1 Юношеский женский 20-29 на деформацию позвоночника Правосторонняя Рентген ография Редактировать
идиопатический сколиоз на боли в грудном отделе позвоночника грудная позвоночника | Подробнее |
при физической нагрузке сколиотческая дуга (спондилография) Удалить
МРТ грудного отдела
позвоночника
М41.1 Юношеский женский 20-29 на деформацию позвоночника Правосторонняя Рентгенография Редактировать
идиопатический сколиоз на боли в позвоночнике после порд оскол и отич еская позвоночника- | Подробнее |
физических нагрузок деформация грудного (спондилография) Удалить
отдела позвоночника ФЭГДС
с УЗИ ОБП
п роти вой скривленном УЗИ сердца
в поясничном отделе ФВД
Fig. 1. «List of medical histories» page
Введите известные данные о пациенте:
Пол женский
Возрастная категория
Жалобы при поступлении
□ на деформацию позвоночника
□ на боли в грудном отделе позвоночника при вертикальных нагрузках
Ы на боли в грудном отделе позвоночника при физической нагрузке
Первичный осмотр
□ Правосторонний грудной кифосколиоз
□ Правосторонняя грудная сколиотическая дуга
□ Правосторонняя лордоскопиотическая деформация грудного отдела позвоночника с прогивоискривпением в поясничном отделе
Анализы 0Биохимические исследования: общий белок в крови - норма □ Биохимические исследования: общий белок в крови - повышен а
0Биохимические исследования: мочевина в крови - норма ▼
Найти рабочие диагнозы
Fig. 2. Data input form
algebraic system - a case of subject domain is built. The class of all cases creates an ontological model of considered subject domain. Subject domain ontology is true on the ontological model.
Formal context is built on the basis of ontologi-cal model of the given subject domain. In constructed formal context a formal concepts is defined, confirming the one or the other diagnosis. Formula descriptions of diagnoses are built.
Then fuzzy model of the patient passed partial examination is constructed. Truth values of formula descriptions of various diagnoses are tested on this model, a set of preliminary diagnoses for the patient is formed. The formal context of diagnoses which is a subcontext of subject domain context is considered. In this context, the maximal formal concept, the content of which is declared as a set of needed additional instrumental research tools, is found.
The developed methods are implemented in «Diagnostic Panel» software system. The ontological model is the core of the program. The software system is tested on a «spinal deformity and degen-
erative diseases of the spine» subject domain. System returns a set of preliminary (working) diagnoses for the patient based on clinical and laboratory research of that patient with disease of the spine. On the basis of the preliminary diagnosis, the system helps the user (doctor) to select the minimum necessary set of instrumental diagnostic tools to determine the final diagnosis of the patient.
We express our gratitude to the staff of the Novosibirsk Research Institute of Traumatology and Orthopedics n. a. Y.L. Tsivyan, who kindly provided us with the necessary medical information.
ACKNOWLEDGMENTS
The reported study was supported by RFBR, research project No. 14-07-00903-a.
REFERENCES
1. Alexander J. Principles of spinal surgery. N. Y.: McGraw-Hill Comp., Health Professions Division, 1996.
Рабочие диагнозы:
• М41.1 Юношеский идиопатический сколиоз
Инструментальные средства для уточнения диагноза:
Инструментальное исследования Диагнозы
Рентгенография позвоночника (спондилография) М41.1 Юношеский идиопатический сколиоз
MPT грудного отдела позвоночника
Fig. 3. Results of the main algorithm work 102 СИБИРСКИЙ НАУЧНЫЙ МЕДИЦИНСКИЙ ЖУРНАЛ, ТОМ 36, № 1, 2016
2. Ganter B., Stumme G., Wille R. Formal concept analysis. Foundations and applications. Springer, 2005.
3. Ganter B., Wille R. Formal Concept Analysis: Mathematical Foundations // Heidelberg: Springer, 1999.
4. Ingerleyb M.B. Analyses. Full reference book. Moscow: Astral, 2011. [In Russian].
5. Maigne J., Rime B., Delignet B. Computed tomographic follow-up study of forty-eight cases of no-noperatively treated lumbar intervertebral disc herniation // Spine. 1992. 17. (9). 1071-1074.
6. Mironov S., Kotelnikov G. Orthopedics. National guidelines. Moscow: GEOTAR-Media, 2008.
7. Palchunov D. Lattices of relatively axiomatiz-able classes // Formal Concept Analysis. Eds. S. Kuz-netsov, S. Schmidt. Berlin; Heidelberg: Springer, 2007.4390. 221-239.
8. Palchunov D. Simulation of thinking and for-malization of reflection: I. Model-theoretic formaliza-tion of the ontology and reflection // Filosofiya nauki = Phylosophy of science. 2006. 31. (4). 86-114. [In Russian].
9. Palchunov D. Simulation of thinking and for-malization of reflection: II. Ontologies and formaliza-tion of concepts // Filosofiya nauki = Phylosophy of science. 2008. 37. (2). 62-99. [In Russian].
10. Palchunov D., Yakhyaeva G. Fuzzy algebraic systems // Vestnik Novosibirskogo gosudarstvennogo universiteta. Seriya: Matematika, mekhanika, infor-
matika = Herald of Novosibirsk State University. Series: Mathematics, mechanics, informatics. 2010. 10. (3). 75-92. [In Russian].
11. Palchunov D., Yakhyaeva G. Fuzzy logics and fuzzy model theory // Algebra and Logic. 2015. 54. (1). 74-80.
12. Palchunov D., Yakhyaeva G. Interval fuzzy algebraic systems // 9th Asian Logic Conference: Proc. Novosibirsk, 2005. 23-37.
13. Priss U. Formal Concept Analysis in Information Science // Annu. Rev. Inf. Sci. Technol. 2006. 40. 521-543.
14. The international statistical classification of the diseases and related health problems; 10th revision. http://www.who.int/classifications/icd/en/.
15. Umashev G., Backbone F.M. Osteochondrosis, 2nd edition. Moscow: Meditsina, 1984. [In Russian].
16. Yakhyaeva G. Fuzzy model truth values // Ap-limat': Proc. 6th Int. Conf. Bratislava, 2007.
17. Yakhyaeva G. Logic of fuzzifications // II-CAI-09: Proc. 4th Int. Conf. Tumkur, 2009.
18. Yakhyaeva G., Yasinskaya O. An algorithm to compare computer-security knowledge from different sources // ICEIS: Proc. 17th Int. Conf. Barcelona, 2015.
19. Yakhyaeva G., Yasinskaya O. Application of case-based methodology for early diagnosis of computer attacks // J. Comput. Inf. Technol. 2014. 22. (3). 145-150.
ПРОГРАММНАЯ СИСТЕМА, ОСУЩЕСТВЛЯЮЩАЯ CASE-BASED REASONING ДЛЯ ДИАГНОСТИРОВАНИЯ ЗАБОЛЕВАНИЙ ПОЗВОНОЧНИКА
Дмитрий Евгеньевич Пальчунов12, Гульнара Эркиновна Яхъяева2, Ольга Владимировна Ясинская2
1 Институт математики им. С.Л. Соболева СО РАН 630090, г. Новосибирск, пр. Академика Коптюга, 4
2 Новосибирский национальный исследовательский государственный университет 630090, г. Новосибирск, ул. Пирогова, 2
В работе описывается программная система «Diagnostic Panel», разработанная для предметной области «деформации позвоночника и дегенеративные заболевания позвоночника». Работа основана на методах статистической обработки данных, извлекаемых из медицинских документов, написанных на естественном языке. Программная система помогает врачам на основе данных клинических и лабораторных исследований пациента определять предварительный диагноз и максимально быстро получать информацию о необходимости проведения тех или иных инструментальных диагностических процедур для постановки заключительного диагноза и выбора оптимальной стратегии лечения. В программной системе «Diagnostic Panel» для представления знаний, извлечённых из различных текстов естественного языка, используется прецедентный подход к представлению знаний. Разрабатываемый прецедентный подход основан на теоретико-модельных методах формализации онтологий предметных областей. Для обработки представленных в системе знаний используется методология анализа формальных понятий.
Ключевые слова: диагноз, заболевания позвоночника, прецедентная модель, нечеткая модель, онтология, формальный контекст, формальное понятие.
Пальчунов Д.Е. - д.ф.-м.н., проф., зав. кафедрой общей информатики, в.н.с., е-шаП: palch@math.nsc.ru
Яхъяева Г.Э. - к.ф.-м.-н., доцент, е-шаП: gul_nara@mail.ru
Ясинская О.В. - ассистент преподавателя, е-шail: yasinskaya.olga@gшail.coш
