Application Hybrid Model to Support in the Diagnosis of Posttraumatic Stress Disorder

Transcription

1 Application Hybrid Model to Support in the Diagnosis of Posttraumatic Stress Disorder Luciano Comin Nunes, Plácido Rogério Pinheiro, Tarcísio Cavalcante Pequeno, Mirian Calíope Dantas Pinheiro Graduate Program in Applied Computer Sciences, University of Fortaleza Av. Washington Soares, Bloco J, sala 30, CEP: , Fortaleza, Ceará, Brazil lcominn@bnb.gov.br, {placido, tpequeno, caliope}@unifor.br Abstract. Among the main causes of low productivity at work, absence from work hours and even disability professionals, are the psychological disorders. Which most often manifests as psychotic disorders, mood disorders, anxiety disorders, antisocial personality, multiple personality disorder and substance abuse, causing damage to the person himself, his family and to the productive system in the country One of the major problems is related to the lack of early diagnosis, which has provided late and reactive measures when the trader is affected by psychological disorder is already showing signs of inability to work. This study aims to present an alternative to help in decision making when seeking to establish early diagnosis of psychological disorders, specifically those related to the disorder of posttraumatic stress. It presents a proposal for a hybrid expert system consists of structured methodologies for decision support (Multi-Criteria Decision Analysis - MCDA) and representations of structured knowledge in production rules and probabilities (Artificial Intelligence - AI). Keywords: Expert SINTA, expert systems, psychological disorders, multicriteria, Hiview 1 Introduction Psychological disorders vary from person to person and also in the severity degree. Although psychological disorders have causes and specific symptoms, the establishment of diagnosis is not so easy to be determined by clinical and laboratory tests, but by analyzing the behavioral responses from a human being facing daily lives events. In the same way as personality, consciousness and intelligence, the term abnormal behavior is difficult to define because of subjectivity inherent to the subject. Major difficulties are encountered when trying to diagnose whether a person is really suffering from a psychological disorder or not.there are some aggravating factors, which include: first, the resistance offered by a person suffering from psychological disorder to accept the situation and second, a healthy individual pretending to be suffering the disorder in order to get some personal advantages.[1] Besides the possibility of fraud, psychological disorders have been identified among the main causes to worker removal from the workplace, taking in consideration that the disorder influences the individual immune system, causing 157

2 him/her frequent diseases, such as sleep and behavior disorders, infections and inflammations. Aiming to facilitate the understanding of the study contained in this article, it will be presented a hybrid model combining the methodology of multicriteria decision support and expert system. This model will examine the main symptoms and causes of the following types of psychological disorders that have affected the economically active population in the world, covering various industriesand professions: psychotic disorders, mood disorders, anxiety disorders. Afterwards the force relation between symptoms and causes will be established in order to create a hierarchy to set the rules that will be used as a base of knowledge into the expert system to help in the diagnosis in accordance to assigned value to variables representing the symptoms and causes. Important to note that hybrid models have been used to help in the decision-making and the search for diagnoses in the health area, as expressed in the articles referenced into [2], [3], [4] and [5]. 2 Psychological Disorders 2.1 General information Some types of psychological disorders lead to a degeneration of the nervous system of the individual, such as psychotic disorders who, at every crisis, causing the individual to the destruction and loss of nerve cells, affecting mainly the memory capacity and critical thinking, the basic resources needed the integration of the individual in social interaction and the production chain. [6] One of the most drastic consequences for a person who suffers from psychological disorders is the image of incapable made by the society in which lives. It is common to observe in the workplace how people react when a team mate suffers from a psychological disorder, even if he or she is being treated by specialists and taking medicines, most of the work team members observe this person with suspicion. A severe infection controlled by medication, or even eradicated cancer in early stage provides chances for reinstatement of a person into the social life. However, the stigma of suffering from a severe psychological disorder usually invalidates the individual anywhere in the world. One explanation for this discrimination against people suffering from psychological disorders is supported by the misinformation from the great majority who make up the society, starting into the family of this individual. 2.2 Psychological Disorders discussed in this article The DSM-IV or Mental Disorders Diagnostic and Statistical Manual, fourth edition, was decided to be used in this study, considering that it proportionate a better presentation structure and the best synthesis capacity without losing the quality and the details of the studied object. The map, described in Figure 1, generated using the software HIVIEW, version , presents the most well known psychological disorders, highlighting the disorder of posttraumatic stress. 158

3 Figure 1. Major categories of Psychological Disorders. 2.3 Considerations about the Psychological Disorders studied in this article Based on the premise that it is possible to identify the causes and symptoms to psychological problems, we seek to diagnose these disorders quickly. Therefore, in this study were highlighted some of the main control events that are correlated with symptoms and causes that show posttraumatic stress disorder. For each of these control events was assigned a confidence factor that indicates the weight on the outcome of the diagnosis. In Figure 2, generated by the software Hiview can be observed these events with their respective control reliability factors, in decreasing order of importance of each of the disorders in this study. Note that these control events and reliability factors will be submitted to the methodology to support decision making with multiple criteria, described in item 3 below, the purpose of constructing an array of sense of values representing the differences in attractiveness between the events. Figure 2. Main events for the control of posttraumatic stress disorder. 3 Model of decision making support proposed In order to establish a diagnosis in Psychology and Psychiatry, studies, experiments 159

4 and careful observations of the symptoms that lead to the evident disorder cause are necessary. A major difficulty for reaching a correct diagnosis is the complexity of highlighted factors, as well as the vast amount of information, that the expert must take into account including cultural, biological, psychosocial issues, information quality, signs and symptoms common to many diseases. Because of this the process gets complicated and it is transformed into a difficult modeling. The methodology to support multi-criteria decision-making has much to add to the Psychology of Health diagnosis processes. 3.1 Description of methodology to support the decision making in multi-criteria There are three main stages in the decision support [7]: Structure: the problem formulation and the goals identification. Seeks to identify, characterize and organize the relevant factors considered in the decision support process; Assessment: this stage can be divided into a phase of actions (alternatives) partial evaluation under each point of view (criteria) and an evaluation phase, considering the overall number of partial evaluations; Recommendation: sensitivity and robustness analysis are made to check whether changes in the model parameters affect the final assessment outcome. It is a crucial step that helps to generate knowledge about the problem and thereby increase confidence in the decision outcome. The MACBETH method (Measuring Attractiveness by a Categorical Based Evaluation Technique) is an approach for multi-criteria decision support. The research that initiated the method was performed by CA Bana e Costa and J.C. Vansnick in the beginning of the 90s [8]. This methodology was in response to the question: How to build a wide range of preferences from a range of options without forcing policy makers to produce their direct numerically preferences? This approach allows assigning notes to each alternative through a paired comparison. Given two alternatives, the decision which must be the most attractive in the case has greater degree of confidence, and in which degree of attractiveness on a scale that has semantic correlation with ordinal scale. The program itself is the analysis of cardinal consistency (transitivity) and semantics (relations between the differences), suggesting, in case of inconsistency how to resolve it. Linear programming is suggested a range of notes and intervals as they may change without making the problem inconsistent (PPL unfeasible). It also possible adjusts the decision graphically the value of the awarded marks, within the ranges allowed. According to the pioneers of this method, only after this adjustment, with the expert knowledge introduction, it gets characterized as the construction of a cardinal scale of values. The difference in attractiveness is very important in this methodology. For the MACBETH method is important the following reasoning: Given the impacts ij (a) and ij (b) of two potential actions a and b, according to a fundamental point of view FPVj, being judged a more attractive than b, the difference in attractiveness between a and b is judged as "null", "very weak", "weak," "moderate," "strong", "very strong" or "extreme." It introduced a scale formed by different semantic categories in attractiveness; the size not necessarily equal, to facilitate the interaction between the decision maker and analyst. The semantic categories, C k, k = , are represented as follows [9]: C 1 very weak difference of attractiveness (or between zero and weak) C 1 = [s 1, s 2 ] and s 1 = 0 160

5 C 2 weak difference of attractiveness C 2 =] s 2, s 3 ] C 3 moderate difference of attractiveness (or between weak and strong) C 3 =] s 3, s 4 ] C 4 strong difference of attractiveness C 4 =] s 4, s 5 ] C 5 strong difference of attractiveness (or between strong and extreme) C 5 =] s 5, s 6 ] C 6 extreme difference of attractiveness C 6 =] s 6, +[ To facilitate the expression of absolute judgments of difference in attractiveness between the pairs of alternatives, it is useful to the construction of arrays of value judgments [10] Results after submitting the events of control to the support decision making methodology After submitting the events to the methodology to support decision making, implemented through software HIVIEW, which also runs MACBETH method, the control events cited in Figure 2, were adjusted in order to compose the matrix of the main constant value in the left side of the Figure 3, which allows the visualization of the degree of attraction between the events and the Current Scale of confidence factors, and indicate that the results are consistent appearing "Consistent Judgments." The new degrees of attractiveness indicated on the ruler are reflected in the matrix current scale, as well as the adjustments in the confidence factors of the control events, as can be seen in the right side of the Figure 3. The proposed model in this article suggests that the control events and confidence factors listed in Figure 2 be exported to the Expert SINTA tool, to compose the knowledge base of an expert system to help the posttraumatic stress disorder diagnosis. Figure 3. Matrix of value-judgment and degrees of attractiveness varying within permited intervals. 4 Expert systems Expert systems are computer programs that offer solutions to certain problems in the same way that human experts offer solutions under the same conditions. The software Expert 161

6 SINTA is a computational tool that uses artificial intelligence techniques for automatic generation of expert systems [11]. It uses a knowledge representation model based on production rules and probabilities, with the main objective of simplifying the construction work of expert systems through the use of a shared machine inference, the automatic construction of screens and menus, treatment of probabilistic rules production and use of sensitive explanations to the context of the knowledge base modeled. The user responds to a sequence of menus, and the system provides answers that fit the framework identified by the user. As previously mentioned, the control events and confidence factors in a posttraumatic disorder are initially presented in Figure 2, and subsequently analyzed and compared with each other, to calculation the difference in the degree of attractiveness for the construction and trial of the value matrix, generated by the MACBETH method, as shown in Figure 3. After that, the information is compiled again, for finally feeding the Expert SINTA to the construction of expert system. Thus, it is possible to see in Figures 4 and 5, some steps of the Expert System s construction which will assist in posttraumatic disorder diagnosis. Figure 4. Building a new base of knowledge. Figure 5. List of logical rules and analysis of the results obtained after the diagnosis. 5 Conclusion and Future Works Because the hybrid models are not the state of the art, the following questions arise: What is the real contribution of the model presented in this article? What are its advantages? The answers these questions are in the facts of this article to propose an integration of methodologies scientifically structured, to assist areas of human knowledge, providing greater consistency in solving complex problems, which involve subjectivities. Many intelligent techniques have been incorporated into the decision-making process. Despite these technological advances much still needs to be done, especially when it involves the multicriteria analysis. The test realized in this article attempts to show a hybrid model, using a connection manually. So it tries to present the feasibility of integration between the technologies mentioned. With this proposal, we hope to help in the quality of automated diagnostics. 162

7 Acknowledgments. The second author is thankful to the National Counsel of Technological and Scientific Development (CNPq) for the support received. References [1] K. Huffman, M.Vernoy, J. Vernoy, Psychology in Action, Psychology in Action, Chapter 14, Publisher John Wiley & Sons, Inc., [2] Luciano C. Nunes, Plácido R. Pinheiro and Tarcisio C. Pequeno, An Expert System Applied to the Diagnosis of Psychological Disorders, IEEE International Conference on Intelligent Computing and Intelligent Systems (ICIS), p , [3] Luciano C. Nunes; Plácido Rogério Pinheiro; Pequeno, Tarcísio Cavalcante, Pinheiro, Mirian Caliope Dantas. Support Tool in the Diagnosis of Major Depressive Disorder. Communications in Computer and Information Science (Print), v.112, p , [4] A.K.A.de Castro, P.R. Pinheiro, M.C.D.Pinheiro, An Approach for the Neuropsychological Diagnosis of Alzheimer s Disease: A Hybrid Model in Decision Making, Lecture Notes in Computer Science, Rough Sets and Knowledge Technology, Springer Berlin / Heidelberg, Volume 5589/2009, Pages [5] A.K.A.de Castro, P.R. Pinheiro, M.C.D.Pinheiro, A Hybrid Model for Aiding in Decision Making for the Neuropsychological Diagnosis of Alzheimer s Disease, Lecture Notes in Computer Science, Rough Sets and Current Trends in Computing, Volume 5306/2008, Pages [6] R.O.Straub, Helth Psychology, University of Michigan, Desrborn,Worth Publishers, [7] C.A.Bana e Costa, L.Ensslin, E.C.Correa, J.C.Vansnick, Decision support systems in action: integrates application in a multicriteria decision aid process. European Journal of Operational Research, v. 133, p , [8] C.A.Bana e Costa, J. M.D.Corte; J.C.Vansnick, "MACBETH", LSE-OR Working Paper, 56, [9] C.A.Bana e Costa, J.C.,Lourenço, M.P.Chagas, J.C.Bana e Costa, Development of Reusable Bid Evaluation Models for the Portuguese Electric Transmission, Company. Decision Analysis, 5(1):22 42, [10] C.A.Bana e Costa, M.P.Chagas, A career choice problem: An example of how to use MACBETH to build a quantitative value model based on qualitative value judgments, European Journal of Operational Research, 153: , [11] J.H.M.Nogueira, J.F.L.Alcântara, R.C.de Andrade, R.B.de A.e Silva, R.S.Silvestre, Grupo SINTA - SistemasINTeligentes Aplicados, Federal University of Ceará (UFC) and State University of Ceará (UECE),