An Ontology-Based Decision Support System for Interventions based on Monitoring Medical Conditions on Patients in Hospital Wards

Please download to get full document.

View again

of 124
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Games & Puzzles

Published:

Views: 0 | Pages: 124

Extension: PDF | Download: 0

Share
Related documents
Description
An Ontology-Based Decision Support System for Interventions based on Monitoring Medical Conditions on Patients in Hospital Wards By Tian Zhao Supervisors: Rune Fensli, Jan Pettersen Nytun Master Thesis
Transcript
An Ontology-Based Decision Support System for Interventions based on Monitoring Medical Conditions on Patients in Hospital Wards By Tian Zhao Supervisors: Rune Fensli, Jan Pettersen Nytun Master Thesis in Information and Communication Technology IKT590, Spring 2014 Faculty of Engineering and Science University of Agder Grimstad, 2 June 2014 Status: Final Abstract: In this project we present our research and implementation of an ontology-based clinical decision support system, which is supported by Sørlandet Sykehus Kristiansand. We discuss the generic technology of designing decision support systems as well as the practical implementation in our project. Firstly, we design the system structure using UML in Eclipse, based on which, the system is built in Protégéusing ontology techniques. And then patients information and clinical rules are added in the system as the form of individuals. Finally, SPARQL query is used to query for the decisions based on the calculation of patients information and the clinical rules. Our system can continuously monitor vital signs parameters of patients and calculate a risk triage at several levels. In collaboration with valuable experiences from medical expertise, our system helps medical personnel at hospital wards to improve patient care and therefore is of great values in clinics. Keywords: clinical decision support system (CDSS), ontology, OWL, Protégé, patients information, SPARQL query Page 2 of Preface This report is the result of the master thesis IKT 590 (30 ECTS) which is part of our fourth semester MSc study at the Faculty of Engineering and Science, University of Agder (UiA) in Grimstad, Norway. The work on this project started from 1 January 2014 and ended on 2 June We have completed the main goal of our project An Ontology-Based Decision Support System for Interventions based on Monitoring Medical Conditions on Patients in. This project is supported by a hospital in Kristiansand, Norway, which is dealing with patients with a table called TILT (Tidlig Identifisering av Livstruende Tilstander, which means Early Identification of Life-Threatening Conditions in English). We would like to thank our project supervisors Professor Rune Fensli and Jan Pettersen Nytun for the guidance in giving feedback on technical and content of report throughout this project. Through this thesis work, we learnt a lot about project content and technical report writing. Grimstad 2 June 2014 Tian Zhao 1.0 Page 3 of 125 Abbreviations CDSS OWL OPCS NICE CPG CP COMET CHF AF TILT Clinical Decision Support System Web Ontology Language Office of Population Censuses and Surveys National Institute for Clinical Excellence Clinical Practice Guidelines Clinical Pathways Co-morbidity Ontological Modeling & ExecuTion Chronic Heart Failure Atrial Fibrillation Tidlig Identifisering av Livstruende Tilstander Page 4 of Contents 1 Introduction Background and Motivation Problem Statement Literature Review Preoperative Clinical Decision Support System Clinical Guidelines Based Comorbid Decision Support Problem Solution Thesis Outline Clinical Techniques Respiration Frequency Pulse Oximetry Blood Pressure Body Temperature CNS (Central Nervous System) TILT Score System Technology Background Clinical Decision Support System (CDSS) Decision Support System (DSS) Clinical Decision Support System The Key Technology of CDSS The Challenges of CDSS Ontology Techniques Components of an Ontology Protégé OWL Language SPARQL Query Implementation of System Functionalities Requirements and Preparation Technique Solutions Knowledge Arrangement Ontology-Based Modelling of Decision Support System Validation of the Ontology-Based Decision Support System Ontology Graph of Decision Support System Query Retrieval Process Discussion and Evaluation Conclusion and Outlook Conclusion Outlook References Appendix A the System Background Codes Appendix B Graphs of the CDSS Appendix C SPARQL Query for the Eight Clinical Rules Page 5 of 125 Figures Figure 1 Overview of knowledge-based preoperative decision support system [1]...11 Figure 2 NICE preoperative guidelines based on patients ages [1]...12 Figure 3 Methodological Approach [2]...13 Figure 4 CHF Diagnosis Algorithm [2]...14 Figure 5 Classes and properties modelling medication dose uptitration [2]...14 Figure 6 Aligning CHF and AF plan [2]...15 Figure 7 Pulse Oximetry [7]...18 Figure 8 the Whole Structure of the Decision Support System...34 Figure 9 Classes of the System...34 Figure 10 Object Properties of the system...35 Figure 11 Usage of filledby...35 Figure 12 Inverse, Domain and Range of filledby...35 Figure 13 filledby Related Individuals...36 Figure 14 Usage of haschosen...37 Figure 15 Usage of hascollectedanswer...38 Figure 16 Inverse, Domain and Range of hasoption...39 Figure 17 Usage of hasoption (i)...39 Figure 18 Usage of hasoption (ii)...40 Figure 19 Usage of hasvitalsign...41 Figure 20 Data Properties of the System...42 Figure 21 Usage of hasage...42 Figure 22 Usage of hasgender...43 Figure 23 Usage of haslocation...44 Figure 24 Usage of hasdescription...45 Figure 25 Usage of hasfilleddate...46 Figure 26 Usage of hassuggestion...47 Figure 27 Usage of hasvalue...48 Figure 28 Graph of All Classes...49 Figure 29 Graph of All Patients...50 Figure 30 Graph of All Results...50 Figure 31 Query 1: patient name, filled table and filled date...51 Figure 32 Query 2: vital signs and selected options of one specific filled TILT table...52 Figure 33 Query 3: patient name, vital signs and values of vital signs...52 Figure 34 Query 4: patients who has chosen options that has value Figure 35 Query 5: calculate total score of one filled TILT table...54 Figure 36 Query 6: total scores of all filled TILT tables...55 Figure 37 Query 7: the filled TILT tables with total scores larger than Figure 38 Query 8: the vital sign with the largest value in filled TILT table Figure 39 Query 8: the vital sign with the largest value in filled TILT table Figure 40 Query 9: the items with maximum values of all the filled TILT tables...59 Figure 41 Query 10: the description based on the total score of the filled TILT table...60 Figure 42 Query 11: doctor s suggestion of one specific clinical rule...61 Figure 43 Query 12: show the suggestions of all filled TILT tables with total score larger than three...63 Figure 44 Query 13: show the suggestions, descriptions and total scores of all the filled TILT tables...65 Figure 45 Query 14: show the filled TILT table with more than one suggestions...66 Figure 46 Ontology Graph of Filled Tables Figure 47 Ontology Graph of Options Figure 48 Ontology Graph of Suggestions Figure 49 Ontology Graph of Vital Sign Figure 50 SPARQL Query for Rule Page 6 of Figure 51 SPARQL Query for Rule Figure 52 SPARQL Query for Rule Figure 53 SPARQL Query for Rule Figure 54 SPARQL Query for Rule Figure 55 SPARQL Query for Rule Figure 56 SPARQL Query for Rule Page 7 of 125 Tables Table 1 normal body temperature [13]...19 Table 2 TILT Score System [15]...20 Table 3 Rules of TILT Score System...21 Table 4 Examples of TILT Score System...21 Table 5 Grand Challenges in CDSSs [29]...28 Page 8 of 1 Introduction In this project we will design a clinical decision support system, which could output decisions and suggestions automatically according to the patients information and their testing results. In Section 1.1, the motivation of this thesis is introduced. Section 1.2 illustrates the problems need to be solved while Section 1.4 reviews literature with similar problems. Then we get our solutions in Section 1.4. At last, Section 1.5 gives the outline of this thesis. 1.1 Background and Motivation Patients in hospital wards always have long term treatment, so clinical personnel should also work for a long term, including recording patients history data, dealing with multiple tasks simultaneously, and concern the condition changes by testing regularly. Ideally clinical experts should make optimum, efficient and professional judgement according to their knowledge and experience, which could be used as good decisions for patients. However, in practice errors usually happen, for the reasons of negligence, misjudgement, insufficient time to send message, lack of attention or fail to exchange information at shift handovers. Even a small mistake may lead to an awful consequence, due to the large number of people that involved in the patient care progress, including patients, clinical professionals and their families and friends,. Thus clinical service centres are trying hard to find efficient methods to ensure and improve patient safety during treatment. Some professionals proposed the decision support systems, which could deal with patients information and manage test data, as well as output appropriate decisions or suggestions as results[1]. By this motivation, we try to use ontology techniques to build a clinical decision support system, which can generate early identification of life-threatening conditions of patients in hospital wards. This project is supported by Sørlandet Sykehus Kristiansand - one hospital in Kristiansand, Norway. This project can improve automated medical diagnosis and therefore benifit both clinical professionals and patients for saving time and human resources. This project makes patients in hospital wards access to their disease condition diagnosis and clinical suggestions automatically. In addition, this project can also help the personnel in hospital significantly in saving time and serving personalised and tailored care for patients in serious conditions. 1.2 Problem Statement For patients in a labile phase of treatment in hospital wards, clinical personnel need a continuous and follow-up care, in order to improve patients safety and stablize the conditions. Thus it is an important breakthrough in medical field to design the decision support system, which can support continuous monitoring of vital sign parameters, calculate a risk triage at several levels, and give expert based advices for interventions. The clinical decision support system we are going to design in this project may implement the following functions: 1.0 Page 9 of 125 1. Having a clear and general structure of vital sign testing table, whose elements can be easily changed and updated; 2. Storing patients information as examples, which could be accessed directly by the system; 3. Formulating some clinical rules for decision making and inserting them into the system; 4. Dealing with patients information and tes data based on inserted clinical rules, and making proper decisions. 1.3 Literature Review In this part we will introduce some similar designed clinical decision support systems, from which we could learn the importance of using OWL ontology in medical diagnosis, as well as how to implement such kind of systems Preoperative Clinical Decision Support System In 2011, Matt-Mouley Bouamrane, Alan Rector and Martin Hurrell used OWL ontology to build a clinical decision support system, in order to support patient information modelling and preoperative clinical decision making. Semantic web technology was used to design and implement this system, which is knowledge-based. This system is ontology-based, which help to develop the model, the user interface and the automated logic reasoned. This system is also expert-based, which help to avoid doctors misjudgement. Large amount of patients information could be stored into the system, which help to make correct decisions for both new and existing patients, as well as patients in specific situations. The basic process that the system work is shown in Figure 1. Page 10 of Figure 1 Overview of knowledge-based preoperative decision support system [1] This project focuses on the assessment of ontology-based preoperative decision support system. some important knowledge inserted in the system includes classification of morbidities using the ICD-10 International Classification of Diseases, classification of surgical procedures based on OPCS (Office of Population Censuses and Surveys) and other relevant evidence-based preoperative assessment medical knowledge such as the NHS NICE (National Institute for Clinical Excellence) routine preoperative tests guidelines. This system using NICE guidelines shows the ontology techniques could help to implement some pragmatic and useful functionalities, and provide a good example of clinical ontology-based reasoner, which is beyond the capabilities of a traditional rule engine. In this system, decision making is based on five factors: age, ASA, comorbidities, type of surgical procedure and risk grade of surgical procedure. For each test, there are three possible results: test recommended, test not recommended or consider test. In Figure 2 a small part of NICE guidelines are shown. 1.0 Page 11 of 125 Figure 2 NICE preoperative guidelines based on patients ages [1] As there are five factors that lead to different decisions, there are large amount of combinations of them. For example, for the age factor, we consider different tables for children under 16 years old and adults over 16 years of age. Totally, the authors declare that there are at least 1242 different possible cases. This system implement a lot of improvements based on the earlier preoperative decision support systems. Firstly, it could collect patients information according to individual medical context. Secondly, it could arrange and manage domain knowledge from a vast repository, including classification of surgical procedures and morbidities, and guidelines for routine preoperative tests [1] Clinical Guidelines Based Comorbid Decision Support In this project, the authors Samina Abidi, Jafna Cox, S. Sibte Raza Abidi and Michael Shepherd designed an ontology-based clinical decision support framework, which could deal with comorbidities in medical. The authors derive the disease-specific clinical pathways (CP) according to clinical practice guidelines (CPG) and medical synthesis knowledge. Then they abstract medical and procedural knowledge, based on which they use ontology to computerize the CP. COMET (Comorbidity Ontological Modeling & ExecuTion) system is suggested by the authors, as it could handle comorbid chronic heart failure and atrial fibrillation, and is web-accessible. The aligning CP process is at the knowledge modelling level instead of the knowledge execution level. It is well established when the common CP need to be mapped to the ontology modelled CP. As shown in Figure 3, four parts are contained in the methodological approach: knowledge identification and synthesis, knowledge modelling, knowledge alignment and knowledge execution. Page 12 of Figure 3 Methodological Approach [2] This designed decision support system is about Chronic Heart Failure (CHF), the knowledge synthesis exercise yielded algorithms for the diagnosis of CHF is shown in Figure Page 13 of 125 Figure 4 CHF Diagnosis Algorithm [2] The system is built in OWL using the ontology editor Protégé, and the flow diagram is shown in Figure 5. Figure 5 Classes and properties modelling medication dose uptitration [2] In order to align comorbid knowledge, the authors integrated the expert knowledge and experience of diagnosing CHF and AF, and made a plan of aligning CHF and AF, which is shown in Figure 6. In the Page 14 of figure, the dashed arrows indicate the alignment between the plans of CHF and AF to handle comorbid CHF+AF. Figure 6 Aligning CHF and AF plan [2] This ontology-based decision support system has lots of advantages with the comorbid CP model. First of all, it helps to avoid duplication of intervention tasks, resources and diagnostic tests. Secondly, it could reuse the results of common activities. Thirdly, it ensures patient safety when different clinical activities are crossing different CP, which are technically compatible. Finally, it makes standards that could be used in multiple institutions [2]. 1.4 Problem Solution The main task of this thesis is to develop a clinical decision support system which can be used for patients treatment. Real vital sign test table and scoring methods is provided by the hospital, and we stored them in our system. Due to the limitation of time, we do not use complex rules and real patient data. As a demo system, this project just contains a few assumed patients and rules, which are enough to test and verify the designed structure and functions. The methods of solution can be summarized as follows: 1. Firstly, according to the TILT table from hospital, the structure of a clinical decision support system is modelled by ontology techniques; 1.0 Page 15 of 125 2. Secondly, assumed clinical rules and example patients are added in the system as individuals; 3. Thirdly, logic of the designed system is evaluated by SPARQL queries according to inserted patients information and test data; 4. Finally, decisions are printed out to check the validation of the system; The implemention of these solutions will be described in detail in the fourth Chapter. 1.5 Thesis Outline The remaining thesis is structured as follows: 1. Chapter 错误! 未找到引用源 presents the clinical background, including the resource from hospital, the vital signs background and the examples and rules we use in this project. 2. In Chapter 错误! 未找到引用源, a brief introduction of technical background adopted in this project is given, including UML, ontology and SPARQL query. 3. In Chapter 错误! 未找到引用源, the whole implementing process is shown, including the whole structure, classes, properties and individuals. 4. In Chapter 错误! 未找到引用源, tests on the system utilizing SPARQL query and the executing results are presented. 5. Discussions are in Chapter 错误! 未找到引用源. 6. Chapter 错误! 未找到引用源 gives the conclusions and future works in this thesis. Page 16 of 2 Clinical Techniques As we mentioned before, this project is supported by a hospital in Kristiansand, Norway. The whole project in the hospital is called Tidlig Identifisering av Livstruende Tilstander (TILT), which means Early Identification of Life-Threatening Conditions in English. The central idea of the project is giving scores to some main factors related to patients based on their vital signs and environment, and then calculate a total score in order to make proper decisions. The decisions include an expert calling, emergency supporting and the frequency to evaluate a new score. Factors in the TILT table include respiration frequency, pulse oximetry, blood pressure, body temperature and CNS (Central Nervous System), which will be described seperately in this Chapter. 2.1 Respiration Frequency For an individual, respiration frequency (also named as respiratory rate or breathing frequency) is defined as the number of breaths taken per minute. In clinic, an individual with an increased respiratory rate can be diagnosed as tachypnea. On the contrary, a lower than normal respiratory rate should be related to the bradypnea [3]. Typically, the average respiration frequency is 12 breaths per minute for a healthy young male in his peaceful condition, i.e. he is resting in bed at sea level. This value would be impacted by position, sex, size, age, altitude, activity, fever, as well as some other illness. In most cases, compare to male adults, children and women may have higher figure of respiration frequency [4]. Respiration frequency can serve as an important clinical diagnosis and be measured in two ways. In the first method, doctors count the chest ups and downs for patients for half a minute and double times as the final respiration frequency. In addition, some medical devices, such as the optical breath rate sensor, are also widely used in clinic to monitor patients breath [5]. The value of respiration frequency is commonly investigated as an indicator of potential respiratory dysfunction. However, the respiratory rate suffers a serious limitation, which is the significant influence from inner and outer factors, such as crying, sleeping, agitation, age and so on. For this rea
Recommended
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x