Encyclopedia of E-Health and Telemedicine

Encyclopedia of E-Health and Telemedicine Maria Manuela Cruz-Cunha Polytechnic Institute of Cávado and Ave, Portugal & Algoritmi Research Centre, Portugal Isabel Maria Miranda Câmara Municipal de Guimarães, Portugal Ricardo Martinho Polytechnic Institute of Leiria, Portugal & CINTESIS - Center for Research in Health Technologies and Information Systems, Portugal Rui Rijo Polytechnic Institute of Leiria, Portugal & INESCC - Institute for Systems and Computers Engineering at Coimbra, Portugal & CINTESIS - Center for Research in Health Technologies and Information Systems, Portugal

Published in the United States of America by Medical Information Science Reference (an imprint of IGI Global) 701 E. Chocolate Avenue Hershey PA, USA 17033 Tel: 717-533-8845 Fax: 717-533-8661 E-mail: [email protected] Web site: http://www.igi-global.com Copyright © 2016 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher. Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark. Library of Congress Cataloging-in-Publication Data Names: Cruz-Cunha, Maria Manuela, 1964- editor. | Miranda, Isabel Maria, 1954- editor. | Martinho, Ricardo, 1974- editor. | Rijo, Rui, editor. Title: Encyclopedia of E-health and telemedicine / Maria Manuela Cruz-Cunha, Isabel Maria Miranda, Ricardo Martinho, and Rui Rijo, editors. Description: Hershey, PA : Medical Information Science Reference, 2016. | Includes bibliographical references and index. Identifiers: LCCN 2015051069| ISBN 9781466699786 (hardcover) | ISBN 9781466699793 (ebook) Subjects: LCSH: Medical care--Technological innovations--Encyclopedias. | Medical informatics--Encyclopedias. Classification: LCC R858 .E518 2016 | DDC 610.28503--dc23 LC record available at http://lccn.loc.gov/2015051069

British Cataloguing in Publication Data A Cataloguing in Publication record for this book is available from the British Library. All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the authors, but not necessarily of the publisher. For electronic access to this publication, please contact: [email protected].

Category: Emerging Technologies

827

Simulation for Medical Training Cecilia Dias Flores Federal University Health Science of Porto Alegre, Brazil Ana Respício Universidade de Lisboa, Portugal Helder Coelho Universidade de Lisboa, Portugal Marta Rosecler Bez Feevale University, Brazil João Marcelo Fonseca Federal University Health Science of Porto Alegre, Brazil

INTRODUCTION General Perspective In a recent meeting held in London in April, Wired Health 2014, it was discussed the future of medicine, along the fusion of healthcare with technology and under the motto “What gets measured gets done”. Three elements are key now for envisaging artifacts, namely data, technology and design. Sensors, algorithms, big data, machine learning, nanotechnology, neurosciences, behavioral psychology and economics, are now adequate triggers for changing radically health and putting it under new tracks. All these topics are within the so-called Social Computation area where several disciplines come together to support aggressive applications for education, entertainment, business or healthcare. The goal is to build social systems, kind of artificial structures, designed and transformed by human action. In what concerns health, these systems may be very complex covering behaviors (predictions and explanations) and artifacts (e.g. for policies, methodologies for organization change, or transition management projects). The aim is to boost efficiency in the services (monitoring vital signs remotely to detect impending problems) and, at the same time, to transform patient experiences with innovative tools capable to predict dis-functions before they happen (the data uploaded to distance servers where it is run through preprogrammed rules that flag up early signs of trouble). The idea is taking earlier decisions before things have actually gone wrong, and builds interventions we have never had the opportunity to consider before, tailored to a person´s profile. The vision of a connected and intelligent approach covers the ability to deal with illness, aging and fitness, by articulating detect with intervene and prevent.

Objectives Part of the illness around us may be mitigated by education and changes of our own behaviours. But it is also necessary help patients to move from physical to digital (and connected) healthcare, by getting them to take their medicine when alarms are activated on account of simple symptoms. A policy of DOI: 10.4018/978-1-4666-9978-6.ch064 Copyright © 2016, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

E

Simulation for Medical Training

early indicators (disrupted patterns) with tacking technology insures the follow-up of new diagnostic and therapeutic approaches. The innovation drive consists of moving from conservative and traditional social information processes toward emphasizing social intelligence, and by inventing new roles for information, Internet and mobile technology. Social intelligence and technology improves also our understanding about human behave and social interactions in human society at the individual, interpersonal and community levels. This chapter focus on simulation for medical training. The literature review examines simulators in the area of healthcare and medical simulation. The chapter describes SimDeCS, the Intelligent Simulator for Decision Making in Health Care Services (in Portuguese, Simulador Inteligente para a Tomada de Decisão em Cuidados de Saúde) which is an end result of a large project for medical learning (Flores, Fonseca, Bez, Respício, & Coelho, 2014). Special focus is given to its architecture and the methodology employed in building clinical cases. SimDeCS plays the role of a virtual patient (Orton & Mulhausen, 2008; McLaughlin et al., 2008) and has been extensively evaluated (Barros, Cazella, & Flores, 2015; Flores et al., 2014; Maroni, Flores, Cazella, Bez, & Dahmer, 2013). Examples of clinical cases are presented. In addition, the chapter proposes future research directions in simulation for medical training and draws final conclusions.

BACKGROUND Many studies have confirmed the effectiveness of simulation in the teaching of medicine and clinical knowledge as well as in the assessment at the undergraduate and graduate medical education levels (Okuda et. al., 2009). Several currently existing simulators propose to offer students safe virtual environments, where they can test and consolidate recently acquired theoretical knowledge in simulated clinical situations (Brookfield, 2005; Botezatu et al., 2010; Holzinger et al., 2009). Table 1 presents examples of several types of simulators for healthcare from the literature. Simulators provide learning environments for the playful application of acquired knowledge and eventually the evaluation of that process. The simulator can also be an open space where students exercise the decision making process in a more realistic framework; not only achieving a goal – such as making a diagnosis or choosing a therapy – but also understanding that different decisions will imply different financial costs, risks to the virtual patient and time expenditures. A simulation can, therefore, also show the student that although excessive research can lead to the correct final result, shorter and cheaper strategies may also lead to adequate results. The development of simulators for healthcare is to a large extent on the use and refinement of Artificial Intelligence (AI). Simulators integrate AI in the form of algorithms that can handle concepts, heuristics use, knowledge representation, support for computation with inaccurate data, multiple solutions, and integrate machine learning mechanisms. According to Bourg and Seemann (2004), AI techniques can be divided into two groups: deterministic and non-deterministic. The first are predictable, easy and quick to implement, however, predictability restricts the simulation, after a few iterations the users realizes what the next states and events. Non-deterministic techniques facilitate learning by providing an unpredictable end to the simulation. Their difficulty lies in the implementation, computational tests, and validation of specific events. The types of simulators using AI are divided by Machado et al. (2009) according to their performance on two levels: the upper level control, referring to the decisions related to the course

828


Table 1. Examples of simulators for healthcare Focus/Type

Software/Subject

Language/Environment

E

Validation

Motivational serious game

• Happy Farm (Gamberini, Marchetti, Martino, & Spagnolli, 2009) • Intensive glycemic control (Thompson et al., 2010)

Two-dimensional, side-scrolling, “platform” games and 3D environment

Skills/Artifacts

• Videolaparoscopic simulator (Seymour et al., 2002) • The Emory Neuro Anatomy Carotid Training (ENACT)(Nicholson et al., 2006) • Personal patient simulation (Gibson, Grimson, Kanade, & Kikinis, 2000) – Simulator patented • Pressure Ulcer Simulator and Related Methods (Sparks, Sanger, & Conner-Kerr, 2012) • Patient Specific Planning and Simulation of Ablative Procedures (Mansi et al., 2014)

Virtual Reality

Strategies

• Cognitive forcing strategies in Emergency Room (Bond et al., 2004) • Training triage of incidents (Huizinga, 1971)

3D Simulators

• JDoc (Sliney & Murphy, 2008) • HAEMOdynamics SIMulator (Holzinger et al., 2009)

• C++ • Several light-weighted JAVA2Applets

Usability Questionnaire

Training

Cardiopulmonary resuscitation (Creutzfeldt, Hedman, Medin, Heinrichs, & Felländer-Tsai, 2010)

Massively multiplayer virtual world

A test-retest (6-month interval between sessions)

Educational Software

• Online virtual patients (Dewhurst, Borgstein, Grant, & Begg, 2009) • Web-based Simulation of Patients application (Web-SP) (Botezatu, Hult, Tessma, & Fors, 2010) • The Clinical Health Economics System Simulation (CHESS) (Voss, Nadkarni, & Schectman, 2005)

• Computerized team-based quasicompetitive simulator • Explorative linear-interactive virtual patient simulation

Usability Questionnaire

Best performance/ Error minimizations; Quantification of the learning curve

Assessment exercise based on tagging accuracy

of action of the simulation, and the lower level of control, referring to the decentralized decisions in the course of the internal simulator decision making process. These systems, also known as knowledge-based systems have rules that replicate the knowledge of human experts, and are used to solve problems in specific domains. The main characteristics of simulators using AI are listed by (Plemenos & Miaoulis, 2009) and include: manipulation of concepts far beyond numerical data; use of heuristic methods to solve problems for which there are no known exact solutions; ability to representing knowledge explicitly; capability to manipulate inaccurate or incomplete data; possibility to obtain multiple solutions; ability to learn, including machine learning mechanisms that imitate human reasoning Bayesian networks (BN), an approach in AI to emulate clinical reasoning, are an option to provide the structure of a medical simulator (Vicari et al., 2003). Achieving the most adequate diagnosis (and treatment) for a given scenario and changing choices, based on new evidence, make BN be similar to several features of medical cognition (Flores et al., 2005; Simel, 2007; Schwartz and Elstein, 2008; Niedermayer, 2008; Pearl, 2009). This advantage has been explored with other clinical decision support tools. A widely known experiment is the Quick Medical Reference – Decision Theoretic, which consists of a model wherein 600 diseases are related with approximately 4000 symptoms (Jaakkola & Jordan, 1999). Over the last two decades, the use of BN, as a basis for algorithms applied in medicine, was reinvigorated, including the addition of learning resources and automatic calibration based on existing clinical records (Ananthaswamy, 2011). The certification examination in family practice of the American Board of Family Practice was enlarged to consider computer-based case simulations (Hagen et. al., 2003).

829


INTELLIGENT SIMULATOR FOR DECISION MAKING IN HEALTH CARE SERVICES: SIMDECS SimDeCS is a multi-agent computational environment that simulates patient care in a Basic Health Unit or during a house call, where the patient is a virtual character and the player is a medical student. During the process, the player receives interventions from an instructor whose goal is to drive the performance. Clinical cases are modeled based on clinical protocols and therapeutic guidelines that are formulated with rigorous quality parameters and represent the technical-scientific consensus. The development of this virtual environment supports the medical teaching and learning processes, both at the graduate and specialization level. Trainee doctors in groups led by more experienced doctors discuss patient information, obtained by anamnesis and physical and subsidiary exams. Scientific literature is also used as a support source. This volume of data is analyzed, mainly by a teacher (or a group of teachers) who indicates which information should receive greatest weight in the decision process and which subsequent research and treatment stages are necessary. Medical doctors, based on the information collected from the above described sources, attempt to establish a differential diagnosis from among the pathologies that may have led to the clinical findings found in the patient. Often new exams are necessary to supply supplementary information in order to narrow the scope of the diagnostic alternatives.

SimDeCS Architecture The SimDeCS architecture (see Figure 1) consists of three layers. The lowest level is the content layer, which includes the techniques applied for knowledge representation and that supply the knowledge data base. Above the content layer is the communication layer that manages the interaction between the different system agents. The upper layer is the presentation one, developed in Flash, able to exchange information with the user and the communication layer through a Java Servlet. This servlet is a component, like a server, that manages HTML and XML data for the presentation layer of a Web application, a class in the programming language Java that dynamically processes requisitions and responses, thereby affording new resources to the multi-agent environment. The Domain Agent represents the domain of the specialist’s knowledge, represented by the clinical cases in the Presentation layer, while the Learner Agent represents the knowledge of the student. If the decisions taken by the student are different from the case information, the Mediator Agent – representing the tutor/instructor – attempts to motivate the learner to review their decision or obtain additional information on the case. The Mediator Agent guides the student using pedagogical strategies selected by an Influence Diagram (ID), explained below. The content layer represents the storage of information about the clinical cases, the logs (records) of student navigation in the simulation as well as the dialog records of the characters and the processing of the ID (Bez et al., 2012) and BN (Flores et al., 2012). The exchange of information among agents is essential for the simulator performance. The SimDeCS agents communicate on a Foundation for Intelligent Physical Agents (FIPA) platform. BN and ID are representations, for communication purposes, in an XML (XBN) based format. In order to establish communication among agents a common structure of references and shared ontologies is necessary. This will determine how a specific message should be interpreted. From an implementation point of view, the format to encode probabilistic knowledge like BN and ID was the HUGIN1 ’s BN representation format.

830


Figure 1. SimDeCS architecture

E

Communication along SimDeCS is established using a JADE2 (JAVA Agent Development) framework that supplies alternatives for the development of agent-based technologies, accelerating the development process.

The Method for Building Simulations in SimDeCS SimDeCS relies on three steps to build a simulation, as presented in Figure 2 and detailed in the following. Step 1: Modeling the Knowledge Specialist. In Step 1, the specialist practitioner structures the medical knowledge into a BN, using the Clinical Practice Guidelines as a basic resource. Some of the Brazilian Society of Family and Community Medicine guidelines were adopted for modeling by BN within the framework of the SimDeCS project. The clinical guideline for Headache, used as a testbed in the project, was written by Family Medicine doctors and offers the perspective of the specialty for the diverse ambulatory issues of the Headache problem (Pinto et al., 2012). The elaboration of the BN is summarized as having three main moments (Pearl, 1988): 1. Identification of the pertinent and relevant variables;

831


Figure 2. Steps of building a simulation in SimDeCS Bez et al., 2012.

2. Organization of these variables into the structure of a directed acyclic graph that represents their relationships; and 3. For each variable (node), obtaining the possible conditional probability values and prior probabilities for model propagation. Guided by the presented semantic issues and with the three tasks in mind, one can define a method for building BN. The database, the literature and the knowledge of the domain specialists are the usually accepted sources for the attribution of numeric values for the probabilities of each network node (Druzdzer & Van Der Gaag, 2000). The method used to generate the BN begins by reading the text and isolating the clinical information that can influence the medical diagnosis or is related with the steps to be adopted. This bibliography search and choice of variables is performed by a domain specialist, using their knowledge as a reference for the choice of variables and the future connection among them. For the clinical guidelines about Headache, 30 variables were used to compose the nodes of the BN – some of which are described in Table 2. The BN built according to this clinical guideline is presented in Figure 3. After separating important variables, nodes corresponding to clinical evidence reinforcing a diagnosis are connected to the corresponding diagnosis node (such the ones presented in Table 2). The calibration of probabilities is also done by the domain specialist doctor, using a specific tool for building the BN. As a diagnosis emerges as predominant over others, a node also emerges as the most adequate. Note that analgesics (“Analgesia” in Figure 3), for instance, are the pertinent treatment for several diagnoses in the headache domain. On the other side, the frequency of crises can, within the diagnosis of migraine, determines the pertinence of a prophylactic treatment. Once the network is built, the domain specialist simulates several typical and non-typical presentations of the diagnoses in questions. That allows for the adjustment of node probabilities so that the most probable diagnosis emerges, according to specialist clinical experience.

832


Table 2. Examples of nodes of the BN Node

E

Description

Aneurysmal history

Family history of central nervous system aneurysm in more than one first degree relative

Facial pain

Pain in face not related with mastication

Fever

Presence of the symptom in patient anamnesis

Prophylaxy *

Prescription of prophylactic medication to decrease recurrence

Specialist *

Decision by the family doctor to solicit consultation by a specialist

Temporomandibular #

Dysfunction of the temporomandibular articulation

Tension-type Headache #

Tension headache

# Represents a diagnosis node. * Represents a clinical procedure node. Remaining nodes represent anamnesis findings.

Figure 3. Bayesian Network built based upon the headache clinical guideline

Each node of the BN has a colloquial phrase associated. Table 3 includes some examples of these sentences. The interface in which the player interacts is composed of characters in a medical office or a home environment. As the player inquires the virtual patient about their symptoms, using structured phrases, the player receives a response, in the form of a colloquial term, as modeled in the clinical case. The BN is inferred by the system and the probability of each node is converted, using a mapping of the answers into probabilities, as follows “always” and “almost always” are mapped into [90%-100%], “most times” and “part of the time” into [50% - 90%]; “few times” and “sometimes” into [10-50%]; and “never”, “almost never”, and “rarely” into [0-10%].

833


When the clinical interview is concluded, the player can formulate their diagnostic hypothesis given the options present in the game – these are highlighted as the most prevalent in the clinical guideline. Subsequently, the player is led to take decisions about supplementary research, leading to a specialist or prescription of medication. Step 2: Modeling clinical cases. In Step 2, clinical cases are created by the teacher based on BN previously built by the domain specialist. By freely including symptoms and signs available on the network, the teacher propagates the probabilities and causes one or more diagnoses and their respective treatments to emerge, thus modeling the case that will be simulated by the students. The clinical cases are stored in the Knowledge Base. They are composed of the nodes selected by the teacher, and make up all the phases of the game (simulation): research, diagnosis and treatment. Additional information about the clinical case like exam images, auscultations and biological signs can be stored at Knowledge Base. Some examples of clinical cases:

Case 1: Maria José This case was to create a female character with complaints suggestive of a migraine. By adding Hemicranial Pain, as present in the BN, the default value of Migraine increases. By adding the Pulsating Pain symptom, Migraine will increase to a probability of 89%, becoming the predominant diagnosis, in contrast with others. The expected treatment is prophylaxis (given the frequency of crisis), followed by analgesics and, secondarily, a follow-up with a specialist (neurologist) given the magnitude of the symptoms. If the student directs his diagnostic hypothesis to the second most probable diagnosis, the decision system of the ID will consider the distance between the probabilities in the network according to the symptoms and signs modeled by the teacher. If the diagnoses have close probabilities (alternative diagnoses or differential diagnoses), the pedagogical strategies will take that similarity into consideration. If they are very discrepant, the strategies chosen to redirect the student will tend to be more incisive. The Figure 4 presents four screens from the SimDeCS simulator interface, representing the interface Table 3. Some of the colloquial phrases associated with nodes Node

Questions and Associated Answers

High frequency

Do you have more than four headache crises per month? _______I have more than four headache crises per month.

Tension-type Headache

“I think your headache is a type called tension headache.”

Analgesia

“I’m going to prescribe analgesics. This is a medication that will relieve pain when it arises.”

Uncommon episodes

Do your headaches occur approximately four times per month or less? ________ I have headaches four time per month or more.

Holocranial Pain

Does your whole hear hurt? Is the pain spread throughout the head? Does it hurt on both sides of your head? _________ all my head hurts. I feel pain _________ throughout my head.

Painful mastication

When you eat, do you feel pain or discomfort in your masticating articulation? _______ it hurts to chew. ______ feel discomfort or pain when I chew.

Specialist

“Considering your symptoms, I think it is necessary to refer you to a specialist for examination and treatment.”

Nuchal rigidity

Is your nape hard, such that you can’t touch your chest with your chin? ______ I feel my nape hard.

834


module of simulation construction, the research process, which is separated into groups: medical history (anamnesis), physical, and requesting (or not) complementary exams; the viewing of medical records of the patient; and the intervention of the simulator (Mediator Agent).

Case 2: Dirceu Cruz When modeling, in the BN editor, a virtual patient called Dirceu Cruz, the intent was to create a young male patient with complaint of a Tension-type cephalalgia. The Holocranial pain and Nasal obstruction nodes were selected as present. When these two symptoms are propagated in the network, Tension-type cephalalgia arises as the most probable diagnosis (98%), followed at a distance (15%) by Sinusitis. Step 3: Use of SimDeCS by students. Step 3 corresponds to the execution of clinical cases by the final user (medical student). In this stage the Learner Agent interacts with the student through a kind of game with two players (student and simulator). The student selects the clinical case to be solved and the simulator presents a summary of the case, the patient file and the interaction possibilities in the three phases of the simulation (Research, Diagnosis and Treatment). In the diagnosis phase, when the student asks the virtual patient a question, the simulaFigure 4. Interface of the SimDeCS

835

E


tor checks the network propagated by the teacher and obtains an answer that expresses colloquially the node probability referred to by the question. Under the dialog form, the student has the possibility of formulating his diagnostic hypotheses by selecting the questions asked of the virtual patient, thus reinforcing or refuting them. The Learner Agent collects all the concrete evidence about the status of its learning process. Based on this evidence, the Learner Agent elaborates and updates the student mode, inferring the credibility (expectation) the system may have about the student, and also records the level of self-confidence declared by the student. A question concerning student self-confidence is asked at the beginning of the clinical case and at the end of the research, diagnosis and treatment phases. Credibility is obtained throughout the whole simulation process. The Domain and Mediator Agent also interact reinforcing the role of the teacher in SimDeCS. The Domain Agent evaluates the student decisions. The result is sent to the Mediator Agent, in order to coordinate the whole interaction process. The interactions between the student and SimDeCS are viewed as a process of pedagogical negotiation (PN), wherein the Mediator Agent resolves differences using several pedagogical strategies. The role of the Mediator Agent is to measure the interactions between the student (Learner Agent) and the teacher (Domain Agent) in each phase. This agent uses an ID to choose the strategy that will demonstrate the best usefulness in each moment. The parameters used are the level of confidence declared by the student and the credibility (inferred by the Learner Agent given the actions of the student during a Simulation). The ID, pedagogical strategies and messages sent to the student are presented in Bez et al. (2012) as shown in Figure 5. The ID is a variation on the BN (Flores et al., 2012; Pearl 1988; An et al., 2007). Its objective is to monitor the actions of the student and what is expected. From this comparison, pedagogical strategies emerge (Bez et al., 2012; Flores et al., 2005) that are sent to the student to positively reinforce him during the research phase, suggest corrections and manage the performance report. Between each phase of research, diagnosis and treatment the student is asked to declare his degree of self-confidence. This confidence modulates the interventions of the ID, through the pedagogical strategies (Table 4). According to DePaola (DePaola, 2008), a way of improving metacognitive skills is to provide students with formative assessments, resources and opportunities that allow them to reflect on their learning through adequate feedback. Several authors of Virtual Patient simulators as Botezatu et al. (2010) and Orton and Mulhausen (2008) show the importance of feedback in the evolution of the clinical case. In

Figure 5. Influence diagram for pedagogical strategy selection

836


Table 4. Available strategies in the influence diagram

E

Credibility

Confidence

High

Medium

Low

High

Extension

Contestation

Proof

Medium

Proof

Contestation

Orientation

Low

Support

Support

Orientation

Bez et al., 2012.

SimDeCS, feedback is provided by an intelligent agent, which monitors the progress during the simulation and adopts pedagogical strategies. The simulator provides interventions of the Mediator Agent between in each phase of the medical interview, physical exams, diagnostic hypothesis, conduct (comprised of prescription medication, new exams or forwarding to specialist).

Pilot Experiment In a preliminary analysis of SimDeCS, we used an ongoing distance education environment directed towards health professionals in the field of Family and Community Medicine. A pilot analysis of the simulator compared with the traditional testing method. Nineteen family physicians received individual usernames and passwords to access the virtual learning environment. They were invited to make a formal evaluation or to play with the simulator. The formal evaluation was composed of 15 multiple-choice questions based exclusively on the content of the clinical guideline. The simulation was composed of several clinical cases of headache in the context of community medicine. Of the 19 invited practicing family doctors only 12 completed both stages of the procedure. The scores of the formal test and the simulator were compared. We did not obtain sufficient participation to allow for a Gaussian statistical evaluation of results. The small number of participants was insufficient for a comparison of performance using the simulator and in the formal test. In spite of this, the experiment could evaluate the pedagogical agent that selects and issue pedagogical strategies for students to undergo simulations of complex clinical cases in the SimDeCS. This evaluation got some interesting results published by Maroni (Maroni et al., 2013). Another evaluation done from this pilot experiment was the evaluating software based on the ISO / IEC 9126 standards and the ten golden rules for software in medical education published in (Barros et al., 2015).

FUTURE RESEARCH DIRECTIONS The importance of (educational) technology in a digital world is greater than a decade ago. For example in education, MOOC´s (Massive Online Open Courses) increased their relevance, from 2012 on, and they present now advantages in what concerns traditional courses because they may be watched every time the students desire, playing an entirely different game at each time (see YouTube for a variety of these courses). MOOC´s expect that their participants will be motivated and will have learned how to learn. It is reasonable to assume they will learn to take responsibility when they are given responsibilities. Also simulation tools seem suitable for medical education by creating virtual worlds easily switched on in several situations to face a young doctor and push him to think about the best options. BN´s, and other tools with intelligent agents, one of the powerful technologies around, may create environments

837


where students can practice their skills by interacting with each other. In such virtual environments, students can learn by watching and joining in, and so to learn to be better reasoners, motivators and arguers. Each individual is unique. Yet, we are cared as we were all identical. Today, any illness requires fast and acute diagnostics, in order to keep patients in a safer environment. A new look of healthcare requires new kinds of personalized medicines, towards precision and anticipation. For example, more tests are needed to reveal cancer presence (biological signature), drug resistance or bacteria transmission. Also, a new generation of tools will be capable to support sharp ways (earlier prescriptions, prognostic evaluation, therapeutic targeting, controllable and dynamic devices to detect malfunctions and infections in the body) to foresee what comes next. A different line of research for medical diagnosis is to follow inspiration, and no longer causality. A doctor is always looking for ideas (ideation and action) when he is facing a complex case. Therefore, a different tool able to foster creativity is needed to draw ideas from a stream by generating opportunities to connect him to inspiration anytime and anywhere. The app Pinterest (2014) is very suggestive to help build another sort of tools, where discovery is articulated with guided search. The aim is to facilitate a kind of fluid creativity, and a short path is the curating ability of handling personalized streams of images. Collecting is a part of the creative process because collectors often recognize patterns, and this tool may spot patterns and make connections.

CONCLUSION This chapter presents our perspective on simulation for medical training based on our experience on the development of SimDeCS, an educational tool that simulates providing medical care to a patient using a virtual character. This experience promoted the study whether the use of BNs as a pedagogical resource would be feasible. In addition, we investigated whether BNs would enable the student to model properly his knowledge, follow the student’s actions during the learning process, make inferences through a probabilistic agent, and, also, select pedagogical actions that have the maximum utility for each student at each moment of his knowledge construction process. All these processes involve all the complexity and dynamics of a human agent learning process, but with the possibility of being followed by artificial agents and, therefore, are assumed to be probabilistic. In addition, our research has been supported by our own vision on how to analyse, interpret and model the complex phenomena that occurs in the whole teaching-learning process, through modelling the student and even the process of pedagogical negotiation (Flores et al., 2005). The temporality of the relationships among the nodes in BN generates problems in assembling the networks, on account of the false causal nexuses (Pearl, 2009). This problem is especially relevant for networks that express this type of causal relationship among variables (nodes) through time. Time dimension can be explored along two directions, one toward the past (to check the spectrum of alternatives), and another for the future (to advance predictions). Nevertheless, multiple causality can arise from contexts where each one and several different nodes have influence over a child node. The structural contingencies that can modify these influences, if modified through time, make it difficult to adequately represent this knowledge in the form of a BN (Pearl, 2009). Many of the correct medical practices have an ideal moment to be applied, behaving as if there were an adequacy time gradient that can be increasing, decreasing or vary irregularly. A procedure may benefit a patient admitted to the emergency room if done in good time, yet not benefit the patient (or potentially even be harmful) if performed later.

838


The current SimDeCS architecture does not integrate the time factor in the analysis of student performances. We envisage an improvement by relying on the Fuzzy Logic concept association. Many clinical decisions have a temporal relation, because adequate measures in the beginning stages of pathology can be inadequate at later stages, or vice-versa. In addition, often used concepts are context dependent and may have imprecise and variable meanings in different scenarios. To sum up, in our perspective there still are challenges and opportunities to increase the trustworthiness of the simulator as a learning environment, namely, by widening it for clinical situations even closer to real ones which encompass complex decision-making processes.

ACKNOWLEDGMENT During 2010, the Brazilian educational agency CAPES through an open Call (Edital 024/2010) invited all the universities to propose projects in the area of healthcare education. This research around a simulator with probabilistic networks (SimDeCS) was accomplished in the following where a methodology to build up clinical cases with this simulator was further developed. We would like to acknowledge CAPES for all the supplied aid. We acknowledge FCT funding under project UID/MAT/04561/2013. We also acknowledge the anonymous reviewers of a previous version of the chapter.

REFERENCES An, N., Liu, J., & Bai, Y. (2007, August). Fuzzy Influence Diagrams: An Approach to Customer Satisfaction Measurement. In Fuzzy Systems and Knowledge Discovery, 2007. FSKD 2007. Fourth International Conference on (Vol. 4, pp. 493-497). IEEE. Ananthaswamy, A. (2011). I, algorithm: A new dawn for AI. New Scientist, 209(2797), 28–31. doi:10.1016/ S0262-4079(11)60216-9 Barros, P. R. M., Cazella, S. C., & Flores, C. D. (2015). Analyzing Softwares in Medical Education Focusing on Quality Standards. Academic Press. Bez, M. R., Flores, C. D., Fonseca, J. M., Maroni, V., Barros, P. R., & Vicari, R. M. (2012). Influence Diagram for selection of pedagogical strategies in a multi-agent system learning. In. Lecture Notes in Computer Science: Vol. 7637. Advances in Artificial Intelligence–IBERAMIA 2012 (pp. 621–630). Springer Berlin Heidelberg. doi:10.1007/978-3-642-34654-5_63 Bond, W. F., Deitrick, L. M., Arnold, D. C., Kostenbader, M., Barr, G. C., Kimmel, S. R., & Worrilow, C. C. (2004). Using simulation to instruct emergency medicine residents in cognitive forcing strategies. Academic Medicine, 79(5), 438–446. doi:10.1097/00001888-200405000-00014 PMID:15107283 Botezatu, M., Hult, H., Tessma, M. K., & Fors, U. (2010). Virtual patient simulation: Knowledge gain or knowledge loss? Medical Teacher, 32(7), 562–568. doi:10.3109/01421590903514630 PMID:20653378 Brookfield, S. (2005). The Power of Critical Theory: Liberating Adult Learning and Teaching (1st ed.). San Francisco, CA: Jossey-Bass. Ceccim, R. B., & Feuerwerker, L. (2004). A four-way approach to training in the health field: Teaching, management, care, and social control. Physis (Rio de Janeiro, Brazil), 14(1), 41–65.

839

E


Creutzfeldt, J., Hedman, L., Medin, C., Heinrichs, W. L., & Felländer-Tsai, L. (2010). Exploring virtual worlds for scenario-based repeated team training of cardiopulmonary resuscitation in medical students. Journal of Medical Internet Research, 12(3), e38. doi:10.2196/jmir.1426 PMID:20813717 DePaola, D. P. (2008). The revitalization of U.S. dental education. Journal of Dental Education, 72(2Suppl), 28–42. PMID:18250375 Dewhurst, D., Borgstein, E., Grant, M. E., & Begg, M. (2009). Online virtual patients - A driver for change in medical and healthcare professional education in developing countries? Medical Teacher, 31(8), 721–724. doi:10.1080/01421590903124732 PMID:19811208 Druzdel, M. J., & Van Der Gaag, L. C. (2000). Building probabilistic networks:” Where do the numbers come from? IEEE Transactions on Knowledge and Data Engineering, 12(4), 481–486. doi:10.1109/ TKDE.2000.868901 Flores, C. D., Bez, M. R., Respício, A., & Fonseca, J. M. (2012). Training Clinical Decision-Making through Simulation. In Hernandez, J et al., Decision Support Systems–Collaborative Models and Approaches in Real Environments, Lecture Notes in Business Information Processing (vol. 121, pp. 59-73). Springer Berlin Heidelberg. doi:10.1007/978-3-642-32191-7_5 Flores, C. D., Fonseca, J. M., Bez, M. R., Respício, A., & Coelho, H. (2014). Method for Building a Medical Training Simulator with Bayesian Networks: SimDeCS. Studies in Health Technology and Informatics, 207, 102–114. PMID:25488216 Flores, C. D., Seixas, L. J., Gluz, J. C., & Vicari, R. M. (2005). A model of pedagogical negotiation. In. Lecture Notes in Computer Science: Vol. 3808. Progress in Artificial Intelligence (pp. 488–499). Springer Berlin Heidelberg. doi:10.1007/11595014_49 Frenk, J., Chen, L., Butta, Z., & Cohen, J. et al.. (2010). Health professionals for a new century: Transforming education to strengthen health systems in an interdependent world. Lancet, 376(9756), 1923–1958. doi:10.1016/S0140-6736(10)61854-5 PMID:21112623 Gamberini, L., Marchetti, F., Martino, F., & Spagnolli, A. (2009). Designing a serious game for young users: The case of happy farm. Studies in Health Technology and Informatics, 144, 77–81. PMID:19592735 Gibson, S. F. F., Grimson, W. E. L., Kanade, T., & Kikinis, R. (2000). Personal patient simulation. Google Patents. Hagen, M. D., Sumner, W., Roussel, G., Rovinelli, R., & Xu, J. (2003). Computer-based testing in family practice certification and recertification. The Journal of the American Board of Family Practice, 16(3), 227–232. doi:10.3122/jabfm.16.3.227 PMID:12755250 Holzinger, A., Kickmeier-Rust, M. D., Wassertheurer, S., & Hessinger, M. (2009). Learning performance with interactive simulations in medical education: Lessons learned from results of learning complex physiological models with the HAEMOdynamics SIMulator. Computers & Education, 52(2), 292–301. doi:10.1016/j.compedu.2008.08.008 Huizinga, J. (1971). Homo ludens: o jôgo como elemento da cultura. Editora da Universidade de S. Paulo, Editora Perspectiva.

840


Jaakkola, T. S., & Jordan, M. I. (1999). Variational Probabilistic Inference and the QMR-DT Network. Journal of Artificial Intelligence Research, 10, 291–322. Machado, L. S., Moraes, R. M., & Nunes, F. (2009). Serious Games para Saúde e Treinamento Imersivo. In Abordagens Práticas de Realidade Virtual e Aumentada. Porto Alegre: SBC. Mansi, T., Sharma, P., Mihalef, V., Kamen, A., Rapaka, S., & Comaniciu, D. (2014). Patient Specific Planning and Simulation of Ablative Procedures. Google Patents. Maroni, V., Flores, C. D., Cazella, S. C., Bez, M. R., & Dahmer, A. (2013). Development and Evaluation of an Intelligent Pedagogical Agent for the SimDeCS Software. Procedia Technology, 9, 1217–1226. doi:10.1016/j.protcy.2013.12.136 McLaughlin, S., Fitch, M. T., Goyal, D. G., Hayden, E., Kauh, C. Y., Laack, T. A., & Gordon, J. A. (2008). Simulation in graduate medical education 2008: A review for emergency medicine. Academic Emergency Medicine, 15(11), 1117–1129. doi:10.1111/j.1553-2712.2008.00188.x PMID:18638028 Nicholson, W. J., Cates, C. U., Patel, A. D., Niazi, K., Palmer, S., Helmy, T., & Gallagher, A. G. (2006). Face and content validation of virtual reality simulation for carotid angiography: Results from the first 100 physicians attending the Emory NeuroAnatomy Carotid Training (ENACT) program. Simulation in Healthcare, 1(3), 147–150. doi:10.1097/01.SIH.0000244457.30080.fc PMID:19088583 Niedermayer, D. (2008). An Introduction to Bayesian Networks and Their Contemporary Applications. In D. E. Holmes & L. C. Jain (Eds.), Innovations in Bayesian Networks (Vol. 156, pp. 117–130). Springer Berlin Heidelberg. doi:10.1007/978-3-540-85066-3_5 Okuda, Y., Bryson, E. O., DeMaria, S., Jacobson, L., Quinones, J., Shen, B., & Levine, A. I. (2009). The utility of simulation in medical education: what is the evidence?. Mount Sinai Journal of Medicine: A Journal of Translational and Personalized Medicine, 76(4), 330-343. Orton, E., & Mulhausen, P. (2008). E-learning virtual patients for geriatric education. Gerontology & Geriatrics Education, 28(3), 73–88. doi:10.1300/J021v28n03_06 PMID:18215989 Pearl, J. (1988). Probabilistic Reasoning in Intelligent Systems. San Mateo, CA: Morgan Kaufmann. Pearl, J. (2009). Causality: Models, Reasoning and Inference (Vol. 29). Cambridge University Press. doi:10.1017/CBO9780511803161 Pinterest. (2014). Retrieved September 9, 2014 from https://www.pinterest.com/ Pinto, M. E. B., Wagner, H. L., Klafke, A., Ramos, A., Stein, A. T., & de Castro Filho, E. D. (2007). Diagnóstico e tratamento das cefaléias em adultos na Atenção Primária à Saúde. Sociedade Brasileira de Medicina de Família e Comunidade, 17. Schwartz, A., & Elstein, A. S. (2008). Clinical reasoning in medicine. In J. Higgs, M. A. Jones, S. Loftus, & N. Christensen (Eds.), Clinical reasoning in the health professions (3rd ed.; pp. 223–234). Edinburgh, UK: Elsevier. Seymour, N. E., Gallagher, A. G., Roman, S. A., O’Brien, M. K., Bansal, V. K., Andersen, D. K., & Satava, R. M. (2002). Virtual reality training improves operating room performance: results of a randomized, double-blinded study. Ann Surg, 236(4), 458-463; discussion 463-454.

841

E


Simel, D. L. (2011). Approach to the patient: history and physical examination. In L. Goldman & A. I. Schafer (Eds.), Goldman’s Cecil Medicine (24th ed.). Philadelphia, PA: Elsevier Saunders. Sliney, A., & Murphy, D. (2008). JDoc: A Serious Game for Medical Learning. Paper presented at the Advances in Computer-Human Interaction, 2008 First International Conference on. Thompson, D., Baranowski, T., Buday, R., Baranowski, J., Thompson, V., Jago, R., & Griffith, M. J. (2010). Serious Video Games for Health How Behavioral Science Guided the Development of a Serious Video Game. Simulation & Gaming, 41(4), 587–606. doi:10.1177/1046878108328087 PMID:20711522 Vicari, R. M., Flores, C. D., Silvestre, A. M., Seixas, L. J., Ladeira, M., & Coelho, H. (2003). A multiagent intelligent environment for medical knowledge. Artificial Intelligence in Medicine, 27(3), 335–366. doi:10.1016/S0933-3657(03)00009-5 PMID:12667742 Voss, J. D., Nadkarni, M. M., & Schectman, J. M. (2005). The Clinical Health Economics System Simulation (CHESS): A teaching tool for systems- and practice-based learning. Academic Medicine, 80(2), 129–134. doi:10.1097/00001888-200502000-00004 PMID:15671315

ENDNOTES

1 2

842

HuginR, avaliable at http://www.hugin.com JAVA Agent Development available at http://jade.tilab.com/