[go: up one dir, main page]

EP1433130A2 - Method for creating a knowledge-based causal network - Google Patents

Method for creating a knowledge-based causal network

Info

Publication number
EP1433130A2
EP1433130A2 EP02752975A EP02752975A EP1433130A2 EP 1433130 A2 EP1433130 A2 EP 1433130A2 EP 02752975 A EP02752975 A EP 02752975A EP 02752975 A EP02752975 A EP 02752975A EP 1433130 A2 EP1433130 A2 EP 1433130A2
Authority
EP
European Patent Office
Prior art keywords
knowledge
causal network
collected
software tool
diseases
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP02752975A
Other languages
German (de)
French (fr)
Inventor
Joachim Horn
Marco Pellegrino
Ruxandra Scheiterer
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Siemens AG
Siemens Corp
Original Assignee
Siemens AG
Siemens Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Siemens AG, Siemens Corp filed Critical Siemens AG
Publication of EP1433130A2 publication Critical patent/EP1433130A2/en
Withdrawn legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G06N5/022—Knowledge engineering; Knowledge acquisition

Definitions

  • the invention relates to a method for creating a causal network (Bayesian Network) based on a knowledge acquisition.
  • the invention therefore lies in the field of decision theory.
  • Causal networks also known as causal or Bayesian networks, represent graphical representations of causal relationships in a domain, and a large number of probability calculations already exist for these networks.
  • Causal networks e.g. described in F.V. Jensen: An Introduction to Bayesian Networks, UCL Press, 1996) provide a precise and efficient framework, e.g. for calculating the probability of each stochastic variable for a given set of observations.
  • causal network is still a complex undertaking when applied to complex systems, such as medical diagnosis.
  • a particular difficulty in creating causal networks is to design the knowledge acquisition in such a way that it is carried out sufficiently completely by a non-mathematic layperson, such as a doctor, to make a causal network meaningful.
  • the invention provides for the method in question to carry out the knowledge acquisition separately from the creation of the causal network.
  • the knowledge acquisition provides for the gathering of relevant knowledge by structuring the collected knowledge into a structured representation that is so complete that the causal network can be created automatically by means of a computer.
  • the invention accordingly takes a new approach to knowledge acquisition and generation of a causal network, a subset being generated from the collected knowledge, preferably using a mathematical method, in such a way that the resulting representation is complete.
  • the relevant knowledge is collected using a software tool.
  • This collection by means of the software tool is preferably carried out in dialogue on a display device, for example on the monitor of a computer in which the software tool is implemented.
  • An interesting field of application of the method according to the invention relates to the possible support of a medical decision.
  • the software tool for specifying diseases and findings, for relationships between diseases and findings and for specific marginal probabilities and conditional probabilities is designed to ensure that the knowledge gathered is so complete, that the causal network can be created automatically using a compiler.
  • the software tool uses the diseases and the findings as a stochastic variable.
  • unit, its marginal probability and additional information are displayed on the display device.
  • the additional information contains factors which promote and inhibit the selected disease. In order to quantify the effects of the promoting and inhibiting factors, provision is advantageously made to specify conditional probabilities.
  • the support for medical decision-making explained above provides that the symptoms of a selected disease are displayed on a computer monitor, for example, together with the conditional probability that this disease causes the symptom.
  • the inventors of the present application developed the above-described application of the method according to the invention to support medical decision-making as part of the so-called HealthMan project (T. Birkhölzer, M. Haft, R. Hofmann, J. Hörn, M. Pellegrino, V. Tresp: "Intelligent Communication in Medical Gare”. Proceedings of the Joint European Conference on Artificial Intelligence in Medicine and Medical Decision Making (AIMDM 99), Aalborg, Denmark, June 1999, p. 4). Knowledge is first collected and converted into a structured representation using a software tool that is tailored to medical use. This software tool is also referred to here as MedKnow.
  • FIG. 1 shows an embodiment of a surface (monitor display) of the HealthMan dialog and advice system
  • Fig. 2 is a monitor representation of the software tool MedKnow
  • 3 shows a causal network for infections that was generated automatically by a knowledge compiler.
  • the HealthMan project mentioned above and shown in FIG. 1 in the form of an exemplary anamnesis process provides a self-diagnosis service which is used as a health guide in a dialog-guided manner, for example with the patient, and thus significantly relieves the medical practitioner with regard to the diagnosis.
  • the HealthMan project aims to emulate the medical doctor's anamnesis process, i.e. to carry out an interactive process that is dynamically driven by medical knowledge and analyzes the information already available.
  • Causal networks have proven to be a suitable technique for this application because they guarantee knowledge acquisition in the medically relevant direction, i.e. from diseases to symptoms, and by taking into account the previous disposition for special diseases.
  • causal networks (Bayesian Networks) represent a correct means of calculation for the uncertainty underlying the medical history in particular.
  • the HUGIN library is used for inference within the HealthMan project.
  • the inventors used the scenario "initial evaluation of the seriousness of common childhood diseases" as an example for testing the method according to the invention.
  • networks for several subdomains were developed (e.g. infections, respiratory system, skin, abdomen, eyes, ears).
  • the system was tested by a professional usability laboratory and received by users (mothers of young children) as well as by the accompanying doctors.
  • the MedKnow software tool mentioned above is designed so that on the one hand medical experts can formulate their medical knowledge without having to bring in special knowledge of causal networks and probability theory, and on the other hand it is guaranteed that the acquired knowledge in this sense it is complete that the causal network can be generated automatically or on its own.
  • the MedKnow software tool uses two classes of stochastic variables: diseases and findings.
  • a finding can play the role of a symptom, or the role of a disease-promoting or inhibitory factor.
  • An example of the acquisition of the required knowledge is shown in FIG. 2.
  • All diseases and findings are listed in the left part of the window displayed on a computer monitor.
  • the selected disease or finding is shown in the main part of the window.
  • the medical area of infections is shown here as a model and the disease "measles" is selected.
  • the upper part of the main window shows the promoting and inhibiting factors, in this case contact with infected people and immunity.
  • the necessary probabilities must also be specified in order to quantify the effect of the promoting and inhibiting factors. The meaning of these required probabilities and the assumptions on which they are based are discussed in the appendix to the present description.
  • the central part of the main window in FIG. 2 shows the selected disease, its marginal probability and additional information used in the HealthMan project, for example the urgency to consult a doctor.
  • the lower part of the main window shows the symptoms the disease along with the required probability that the disease will actually cause the symptom.
  • FIG. 3 shows the graphical representation of a causal network for infections, generated by a knowledge compiler in accordance with the method according to the invention.
  • the generation (in this case the automatic generation) of a causal network using the knowledge acquired as explained above can be divided into two subtasks: the generation of the graph (shown in FIG. 3) and the calculation of the required probability tables.
  • each disease and each finding is represented by a node and additional nodes are created separately for the collection of promoting factors and for the collection of inhibitory factors of each individual disease.
  • Arrows are drawn from the diseases to the respective symptoms, from supporting factors to the respective collection nodes and from the inhibitory factors to the respective collection nodes and from the collection nodes to the respective diseases (see FIG. 3).
  • the calculation of the necessary probability tables of the causal network is based on the specified probabilities and the gate type.
  • the inventors used gates for findings such as the so-called noisysyOR (FV Jensen: An Introduction to Bayesian Networks, UCL Press, 1996), noisysyMAX and noisysyELENI (R. Lupas Scheiterer: Heal thMan Bayesian Network Description: Disease to Symptom Layer, Siemens AG, ZT IK 4, Internal Report, 1999). Diseases were modeled as a promoting / inhibiting gate (J. Hörn: Heal thMan Bayesian Network Description: Enhancing and Inhibiting Factors of Diseases. Siemens AG, ZT IK 4, internal report, 1999).

Landscapes

  • Engineering & Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a method for creating a causal network on the basis of knowledge acquisition. According to the invention, said knowledge acquisition is separated from the creation of said causal network and comprises the following steps: relevant knowledge is collected, and the knowledge in its entirety is structured to form a structured and complete representation thereof enabling the causal network to be created automatically by a computer.

Description

Beschreibungdescription

Verfahren zum Erstellen eines Kausalen Netzes auf Grundlage einer WissensakquisitionProcess for creating a causal network based on knowledge acquisition

Die Erfindung betrifft ein Verfahren zum Erstellen eines Kausalen Netzes (Bayesian Network) auf Grundlage einer Wissensakquisition.The invention relates to a method for creating a causal network (Bayesian Network) based on a knowledge acquisition.

Die Erfindung liegt demnach auf dem Gebiet der Entscheidungstheorie. Im Rahmen dieser Theorie wurde die klassische Wahrscheinlichkeitstheorie auf einen äußerst präzisen mathematischen Rahmen erweitert, um rationelle Entscheidungen mit Unterstützung von Computern treffen zu können. Kausale Netze, auch als Causal oder Bayesian Networks bezeichnet, stellen graphische Darstellungen von kausalen Beziehungen in einer Domäne dar, und für diese Netze existiert bereits eine große Anzahl an Wahrscheinlichkeitsberechnungen. Kausale Netze (beispielsweise beschrieben in F.V. Jensen: An Introduction to Bayesian Networks, UCL Press, 1996) stellen einen genauen und effizienten Rahmen beispielsweise zur Berechnung der Wahrscheinlichkeit von jeder stochastischen Variablen bei einem vorgegebenen Satz von Beobachtungen dar.The invention therefore lies in the field of decision theory. As part of this theory, the classical probability theory was expanded to an extremely precise mathematical framework in order to be able to make rational decisions with the support of computers. Causal networks, also known as causal or Bayesian networks, represent graphical representations of causal relationships in a domain, and a large number of probability calculations already exist for these networks. Causal networks (e.g. described in F.V. Jensen: An Introduction to Bayesian Networks, UCL Press, 1996) provide a precise and efficient framework, e.g. for calculating the probability of each stochastic variable for a given set of observations.

Kausale Netze kommen auf den unterschiedlichsten Gebieten zum Einsatz, beispielsweise zur Unterstützung der Entscheidung von Ärzten (s. Andreassen, M. Woldbye, B. Falck, S.K. Andersen: "MUNIN - A Causal Probabilistic Network for Interpretation of Electromyographic Findings". Proceedings of the Tenth International Joint Conference on Artificial Intelligence, Mailand, Italien, August 1987, S. 366-372; D.E. Heckerman, E.J. Horvitz, B.N. Nathwani : "Toward Normative Expert Systems: Part I. The Pathfinder Project". Methods of Information in Medicine, Band 31, 1992, S. 90-105; D.E. Heckerman, B.N. Nathwani: "Toward Normative Expert Systems: Part II. Prob- ability-Based Representations for Efficient Knowledge Acquisition and Inference". Methods of Information in Medicine, Band 31, S. 106-116; P.J.F. Lucas, H. Boot, B. Taal : "A Deci- sion-Theoretic Network Approach to Treatment Management and Prognosis". Knowledge-Based Systems, Band 11, 1998, S. 321- 330; B. Middleton, M.A. Shwe, D.E. Heckerman, M. Henrion, E.J. Horvitz, H.P. Lehmann, G.F. Cooper: "Probabilistic Diag- nosis Using a Reformulation of the INTERNIST-1/QMR Knowledge Base, II. Evaluation of Diagnostic Performance". Methods of Information in Medicine, Band 30, 1991, S. 256-267; K.G. Ole- sen, U. Kjaerulff, F. Jensen, F.V. Jensen, B. Flack, S. An- dreassen, S.K. Andersen: "A MUNIN Network for the MediänCausal networks are used in a wide variety of areas, for example to support the decision of doctors (see Andreassen, M. Woldbye, B. Falck, SK Andersen: "MUNIN - A Causal Probabilistic Network for Interpretation of Electromyographic Findings". Proceedings of the Tenth International Joint Conference on Artificial Intelligence, Milan, Italy, August 1987, pp. 366-372; DE Heckerman, EJ Horvitz, BN Nathwani: "Toward Normative Expert Systems: Part I. The Pathfinder Project". Methods of Information in Medicine, Volume 31, 1992, pp. 90-105; DE Heckerman, BN Nathwani: "Toward Normative Expert Systems: Part II. Probability-Based Representations for Efficient Knowledge Acquisition and Inference". Methods of Information in Medicine, Volume 31, pp. 106-116; PJF Lucas, H. Boot, B. Taal: "A Decision-Theoretic Network Approach to Treatment Management and Prognosis". Knowledge-Based Systems, Volume 11, 1998, pp. 321-330; B. Middleton, MA Shwe, DE Heckerman, M. Henrion, EJ Horvitz, HP Lehmann, GF Cooper: "Probabilistic Diagnosis Using a Reformulation of the INTERNIST-1 / QMR Knowledge Base, II. Evaluation of Diagnostic Performance". Methods of Information in Medicine, Volume 30, 1991, pp. 256-267; KG Olesen, U. Kjaerulff, F. Jensen, FV Jensen, B. Flack, S. Andreasreassen, SK Andersen: "A MUNIN Network for the Mediän

Nerve - A Case Study on Loops". Applied Artificial Intelli - gence, Band 3, 1989, S. 385-403; M. A. Shwe, B. Middleton, D.E. Heckerman, M. Henrion, E.J. Horvitz, H.P. Lehmann, G.F. Cooper: "Probabilistic Diagnosis Using a Reformulation of the INTERNIST-1/QMR Knowledge Base. I. The Probabilistic Model and Inference Algorithms". Methods of Information in Medi cine, Band 30, 1991, S. 241-250).Nerve - A Case Study on Loops ". Applied Artificial Intelligence - Volume 3, 1989, pp. 385-403; MA Shwe, B. Middleton, DE Heckerman, M. Henrion, EJ Horvitz, HP Lehmann, GF Cooper:" Probabilistic Diagnosis Using a Reformulation of the INTERNIST-1 / QMR Knowledge Base. I. The Probabilistic Model and Inference Algorithms ". Methods of Information in Medical, Volume 30, 1991, pp. 241-250).

Die zur Erzeugung eines Kausalen Netzes notwendige Wissens- akquisition ist jedoch nach wie vor ein aufwendiges Unterfangen bei Anwendung auf komplexe Systeme, wie etwa auf medizinische Diagnose. Eine spezielle Schwierigkeit bei der Erstellung von Kausalen Netzen besteht dabei darin, die Wissensakquisition derart zu gestalten, dass sie von einem ma- thematischen Laien, wie beispielsweise von einem Arzt, ausreichend vollständig durchgeführt wird, um ein Kausales Netz aussagekräftig zu gestalten.However, the acquisition of knowledge necessary to create a causal network is still a complex undertaking when applied to complex systems, such as medical diagnosis. A particular difficulty in creating causal networks is to design the knowledge acquisition in such a way that it is carried out sufficiently completely by a non-mathematic layperson, such as a doctor, to make a causal network meaningful.

Eine Aufgabe der Erfindung besteht darin, ein Verfahren zu schaffen, durch welches ein Nutzer in die Lage versetzt wird, ein Kausales Netz auf Grundlage einer Wissensakquisition möglichst problemlos zu erstellen.It is an object of the invention to provide a method by which a user is enabled to create a causal network based on knowledge acquisition as easily as possible.

Gelöst wird diese Aufgabe durch die Merkmale des Anspruchs 1. Vorteilhafte Weiterbildungen der Erfindung sind durch die Unteransprüche angegeben. Demnach sieht die Erfindung bei dem in Rede stehenden Verfahren vor, die Wissensakquisition getrennt vom Erstellen des Kausalen Netzes durchzuführen. Insbesondere sieht die Wissensakquisition das Sammeln von relevantem Wissen unter Strukturieren des gesammelten Wissens in eine strukturierte Darstellung vor, die so weit vollständig ist, dass das Kausale Netz mittels eines Computers automatisch erstellt werden kann.This object is achieved by the features of claim 1. Advantageous developments of the invention are specified by the subclaims. Accordingly, the invention provides for the method in question to carry out the knowledge acquisition separately from the creation of the causal network. In particular, the knowledge acquisition provides for the gathering of relevant knowledge by structuring the collected knowledge into a structured representation that is so complete that the causal network can be created automatically by means of a computer.

Die Erfindung beschreitet demnach einen neuen Ansatz zur Wissensakquisition und Erzeugung eines Kausalen Netzes, wobei aus dem gesammelten Wissen bevorzugt mittels einer mathematischen Methode eine Untermenge derart erzeugt wird, dass die daraus resultierende Darstellung vollständig ist.The invention accordingly takes a new approach to knowledge acquisition and generation of a causal network, a subset being generated from the collected knowledge, preferably using a mathematical method, in such a way that the resulting representation is complete.

Bevorzugt ist vorgesehen, das relevante Wissen mittels eines Software-Tools zu sammeln. Dieses Sammeln mittels des Software-Tools erfolgt bevorzugt unter Dialogführung auf einer Anzeigeeinrichtung, beispielsweise auf dem Monitor eines Com- puters, in welchem das Software-Tool implementiert ist.It is preferably provided that the relevant knowledge is collected using a software tool. This collection by means of the software tool is preferably carried out in dialogue on a display device, for example on the monitor of a computer in which the software tool is implemented.

Ein interessantes Anwendungsgebiet des erfindungsgemäßen Verfahrens betrifft die hierdurch mögliche Unterstützung einer medizinischen Entscheidung. In diesem Zusammenhang ist erfin- dungsgemäß bevorzugt vorgesehen, dass das Software-Tool zum Spezifizieren von Krankheiten und Befunden, von Zusammenhängen zwischen Krankheiten und Befunden und von spezifischen Randwahrscheinlichkeiten und bedingten Wahrscheinlichkeiten und dazu ausgelegt ist, sicherzustellen, dass das gesammelte Wissen derart vollständig ist, dass das Kausale Netz mittels eines Compilers automatisch erstellt werden kann. Vorteilhafterweise ist hierbei vorgesehen, dass das Softwaretool die Krankheiten und die Befunde als stochastische Variable nutzt.An interesting field of application of the method according to the invention relates to the possible support of a medical decision. In this context, it is preferably provided according to the invention that the software tool for specifying diseases and findings, for relationships between diseases and findings and for specific marginal probabilities and conditional probabilities and is designed to ensure that the knowledge gathered is so complete, that the causal network can be created automatically using a compiler. It is advantageously provided that the software tool uses the diseases and the findings as a stochastic variable.

Bei der vorstehend angesprochenen Unterstützung der medizinischen Entscheidung mittels des erfindungsgemäßen Verfahrens ist ferner bevorzugt vorgesehen, dass eine ausgewählte Krank- heit, ihre Randwahrscheinlichkeit sowie zusätzliche Information auf der Anzeigeeinrichtung angezeigt werden. Hierbei ist vorteilhafterweise vorgesehen, dass die zusätzliche Information fördernde und hemmende Faktoren der ausgewählten Krank- heit enthält. Zum Quantifizieren von Wirkungen der fördernden und hemmenden Faktoren ist vorteilhafterweise vorgesehen, bedingte Wahrscheinlichkeiten zu spezifizieren.In the case of the above-mentioned support of the medical decision by means of the method according to the invention, it is also preferably provided that unit, its marginal probability and additional information are displayed on the display device. It is advantageously provided here that the additional information contains factors which promote and inhibit the selected disease. In order to quantify the effects of the promoting and inhibiting factors, provision is advantageously made to specify conditional probabilities.

Um den Nutzer bei der Wissensakquisition zu unterstützen, ist bei dem vorstehend erläuterten Unterstützung der medizinischen Entscheidungsfindung vorgesehen, dass die Symptome einer ausgewählten Krankheit zusammen mit der bedingten Wahrscheinlichkeit, dass diese Krankheit das Symptom verursacht, auf beispielsweise einen Computermonitor angezeigt werden.In order to support the user in acquiring knowledge, the support for medical decision-making explained above provides that the symptoms of a selected disease are displayed on a computer monitor, for example, together with the conditional probability that this disease causes the symptom.

Durch die Erfinder der vorliegenden Anmeldung wurde die vorstehend im Kern erläuterte Anwendung des erfindungsgemäßen Verfahrens zur Unterstützung der medizinischen Entscheidungserfindung als Teil des sogenannten HealthMan-Projekt ent- wickelt (T. Birkhölzer, M. Haft, R. Hofmann, J. Hörn, M. Pel- legrino, V. Tresp: "Intelligent Com unication in Medical Gare". Proceedings of the Joint European Conference on Artifi cial Intelligence in Medicine and Medical Decision Making (AIMDM 99) , Aalborg, Dänemark, Juni 1999, S. 4). Dabei wird Wissen zunächst gesammelt und in eine strukturierte Darstellung unter Verwendung eines Software-Tools überführt, das auf den medizinischen Einsatz zugeschnitten ist. Dieses Software-Tool wird vorliegend auch als MedKnow bezeichnet.The inventors of the present application developed the above-described application of the method according to the invention to support medical decision-making as part of the so-called HealthMan project (T. Birkhölzer, M. Haft, R. Hofmann, J. Hörn, M. Pellegrino, V. Tresp: "Intelligent Communication in Medical Gare". Proceedings of the Joint European Conference on Artificial Intelligence in Medicine and Medical Decision Making (AIMDM 99), Aalborg, Denmark, June 1999, p. 4). Knowledge is first collected and converted into a structured representation using a software tool that is tailored to medical use. This software tool is also referred to here as MedKnow.

Um die erfindungsgemäße Wissensakquisition zum Erstellen eines Kausalen Netzes näher zu erläutern, wird erneut Bezug genommen auf die Unterstützung der medizinischen Entscheidungsfindung im Zusammenhang mit der Zeichnung; in dieser zeigen:In order to explain the knowledge acquisition according to the invention for creating a causal network in more detail, reference is again made to the support of medical decision-making in connection with the drawing; in this show:

Fig. 1 eine Ausführungsform einer Oberfläche (Monitoranzeige) des HealthMan-Dialog- und -Beratungssystems, Fig. 2 eine Monitordarstellung des Software-Tools MedKnow, und1 shows an embodiment of a surface (monitor display) of the HealthMan dialog and advice system, Fig. 2 is a monitor representation of the software tool MedKnow, and

Fig. 3 ein Kausales Netz für Infektionen, das durch einen Wissens-Compiler automatisch erzeugt wurde.3 shows a causal network for infections that was generated automatically by a knowledge compiler.

Das vorstehend angesprochene und in Fig. 1 in Gestalt eines beispielhaften Anamnese-Prozesses dargestellte HealthMan-Pro- jekt stellt eine Selbstdiagnose-Dienstleistung bereit, die dialoggeführt, beispielsweise mit dem Patienten, als Gesundheitsratgeber dient und damit den Mediziner hinsichtlich der Diagnose deutlich entlastet. Insbesondere sieht das HealthMan-Projekt vor, den Anamnese-Prozess des Mediziners zu emulieren, d.h., einen interaktiven Prozess auszuführen, der durch medizinisches Wissen dynamisch getrieben ist und die bereits vorliegende Information analysiert. Kausale Netze haben sich für diesen Einsatz als geeignete Technik erwiesen, weil sie in der medizinisch relevanten Richtung eine Wissens- akquisition gewährleisten, d.h., von Krankheiten zu Symptomen, und indem die bisherige Disposition für spezielle Krankheiten berücksichtigt wird. Insbesondere stellen Kausale Netzwerke (Bayesian Networks) eine korrektes Berechnungsmittel für die insbesondere der medizinischen Anamnese zugrundeliegende Unsicherheit dar. Zur Inferenz wird im Rahmen des HealthMan-Projekts die Bibliothek von HUGIN herangezogen.The HealthMan project mentioned above and shown in FIG. 1 in the form of an exemplary anamnesis process provides a self-diagnosis service which is used as a health guide in a dialog-guided manner, for example with the patient, and thus significantly relieves the medical practitioner with regard to the diagnosis. In particular, the HealthMan project aims to emulate the medical doctor's anamnesis process, i.e. to carry out an interactive process that is dynamically driven by medical knowledge and analyzes the information already available. Causal networks have proven to be a suitable technique for this application because they guarantee knowledge acquisition in the medically relevant direction, i.e. from diseases to symptoms, and by taking into account the previous disposition for special diseases. In particular, causal networks (Bayesian Networks) represent a correct means of calculation for the uncertainty underlying the medical history in particular. The HUGIN library is used for inference within the HealthMan project.

Die Erfinder haben beispielhaft zum Testen des erfindungs- gemäßen Verfahrens das Szenario "anfängliche Bewertung der Ernsthaftigkeit üblicher Kinderkrankheiten" herangezogen. In Zusammenarbeit mit mehreren Pädiatern wurden Netze für mehrere Subdomänen entwickelt (beispielsweise Infektionen, Atmungssystem, Haut, Bauch, Augen, Ohren) . Das System wurde durch ein professionelles Brauchbarkeitslabor getestet und durch die Nutzer (Mütter junger Kinder) ebenso positiv aufgenommen wie durch die begleitenden Ärzte. Das vorstehend bereits angesprochene Software-Tool MedKnow ist so konzipiert, dass zum einen medizinische Experten ihr medizinisches Wissen formulieren können, ohne dass sie ein Spezialwissen bezüglich Kausale Netze und Wahrscheinlichkeitstheorie einbringen müssen, und dass zum andern gewährleistet ist, dass das erfasste Wissen in dem Sinne vollständig ist, dass das Kausale Netz automatisch bzw. von sich aus erzeugt werden kann.The inventors used the scenario "initial evaluation of the seriousness of common childhood diseases" as an example for testing the method according to the invention. In cooperation with several pediatricians, networks for several subdomains were developed (e.g. infections, respiratory system, skin, abdomen, eyes, ears). The system was tested by a professional usability laboratory and received by users (mothers of young children) as well as by the accompanying doctors. The MedKnow software tool mentioned above is designed so that on the one hand medical experts can formulate their medical knowledge without having to bring in special knowledge of causal networks and probability theory, and on the other hand it is guaranteed that the acquired knowledge in this sense it is complete that the causal network can be generated automatically or on its own.

Das Software-Tool MedKnow verwendet zwei Klassen von sto- chastischen Variablen: Erkrankungen und Befunde. Ein Befund kann die Rolle eines Symptoms oder die Rolle eines fördernden oder hemmenden Faktors einer Krankheit spielen. Ein Beispiel für die Akquisition des erforderlichen Wissens ist in Fig. 2 gezeigt. Im linken Teil des auf einem Computermonitor dargestellten Fensters sind sämtliche Krankheiten und Befunde aufgelistet. Im Hauptteil des Fensters ist die ausgewählte Krankheit oder der ausgewählte Befund dargestellt. Der edi- zinische Bereich von Infektionen ist vorliegend modellhaft dargestellt und die Krankheit "Masern" ist ausgewählt.The MedKnow software tool uses two classes of stochastic variables: diseases and findings. A finding can play the role of a symptom, or the role of a disease-promoting or inhibitory factor. An example of the acquisition of the required knowledge is shown in FIG. 2. All diseases and findings are listed in the left part of the window displayed on a computer monitor. The selected disease or finding is shown in the main part of the window. The medical area of infections is shown here as a model and the disease "measles" is selected.

Der obere Teil des Hauptfensters zeigt die fördernden und hemmenden Faktoren, vorliegend den Kontakt zu infizierten Personen und die Immunität. Ferner müssen erforderliche Wahrscheinlichkeiten spezifiziert werden, um den Effekt der fördernden und hemmenden Faktoren zu quantifizieren. Die Bedeutung dieser erforderlichen Wahrscheinlichkeiten und die ihnen zugrundeliegenden Annahmen sind in der Anlage zur vorliegen- den Beschreibung diskutiert.The upper part of the main window shows the promoting and inhibiting factors, in this case contact with infected people and immunity. The necessary probabilities must also be specified in order to quantify the effect of the promoting and inhibiting factors. The meaning of these required probabilities and the assumptions on which they are based are discussed in the appendix to the present description.

Der zentrale Teil des Hauptfensters in Fig. 2 zeigt die ausgewählte Krankheit, ihre Randwahrscheinlichkeit und zusätzliche Informationen, die im HealthMan-Projekt verwendet werden, beispielsweise die Dringlichkeit, einen Arzt zu Rate zu ziehen. Der untere Teil des Hauptfensters zeigt die Symptome der Krankheit zusammen mit der erforderlichen Wahrscheinlichkeit, dass die Erkrankung das Symptom tatsächlich verursacht.The central part of the main window in FIG. 2 shows the selected disease, its marginal probability and additional information used in the HealthMan project, for example the urgency to consult a doctor. The lower part of the main window shows the symptoms the disease along with the required probability that the disease will actually cause the symptom.

Eine ähnliche Anzeige ist vorgesehen, wenn es um Befunde geht .A similar notification is provided when it comes to findings.

Fig. 3 zeigt die graphische Darstellung eines Kausalen Netzes für Infektionen, erzeugt durch einen Wissens-Compiler in Übereinstimmung mit dem erfindungsgemäßen Verfahren.3 shows the graphical representation of a causal network for infections, generated by a knowledge compiler in accordance with the method according to the invention.

Die Erzeugung (vorliegend die automatische Erzeugung) eines Kausalen Netzes unter Verwendung des wie vorstehend erläutert akquirierten Wissens kann in zwei Teilaufgaben aufgeteilt werden: In das Erzeugen des Graphen (in Fig. 3 gezeigt) und in das Berechnen der erforderlichen Wahrscheinlichkeitstabellen.The generation (in this case the automatic generation) of a causal network using the knowledge acquired as explained above can be divided into two subtasks: the generation of the graph (shown in FIG. 3) and the calculation of the required probability tables.

Die Erzeugung der Graphen gestaltet sich relativ unkompliziert: Jede Krankheit und jeder Befund wird durch einen Kno- ten wiedergegeben und zusätzliche Knoten werden getrennt für das Sammeln fördernder Faktoren und für das Sammeln hemmender Faktoren jeder einzelnen Krankheit erzeugt. Pfeile werden von den Krankheiten zu den jeweiligen Symptomen gezeichnet, von fördernden Faktoren zu den jeweiligen Sammelknoten und von den hemmenden Faktoren zu den jeweiligen Sammelknoten sowie von den Sa melknoten zu den jeweiligen Krankheiten (siehe Fig. 3) .The creation of the graphs is relatively straightforward: each disease and each finding is represented by a node and additional nodes are created separately for the collection of promoting factors and for the collection of inhibitory factors of each individual disease. Arrows are drawn from the diseases to the respective symptoms, from supporting factors to the respective collection nodes and from the inhibitory factors to the respective collection nodes and from the collection nodes to the respective diseases (see FIG. 3).

Die Berechnung der erforderlichen Wahrscheinlichkeitstabellen des Kausalen Netzes beruht auf den spezifizierten Wahrscheinlichkeiten und dem Gatter-Typ. Für Befunde haben die Erfinder Gatter, wie etwa das sogenannte NoisyOR (F.V. Jensen: An Introduction to Bayesian Networks, UCL Press, 1996) , NoisyMAX und NoisyELENI (R. Lupas Scheiterer: Heal thMan Bayesian Net- work Description : Disease to Symptom Layer, Siemens AG, ZT IK 4, Interner Bericht, 1999) verwendet. Krankheiten wurden als förderndes/hemmendes Gatter modelliert (J. Hörn: Heal thMan Bayesian Network Description : Enhancing and Inhibi ting Fac- tors of Diseases . Siemens AG, ZT IK 4, Interner Bericht, 1999) .The calculation of the necessary probability tables of the causal network is based on the specified probabilities and the gate type. The inventors used gates for findings such as the so-called NoisyOR (FV Jensen: An Introduction to Bayesian Networks, UCL Press, 1996), NoisyMAX and NoisyELENI (R. Lupas Scheiterer: Heal thMan Bayesian Network Description: Disease to Symptom Layer, Siemens AG, ZT IK 4, Internal Report, 1999). Diseases were modeled as a promoting / inhibiting gate (J. Hörn: Heal thMan Bayesian Network Description: Enhancing and Inhibiting Factors of Diseases. Siemens AG, ZT IK 4, internal report, 1999).

Die Berechnung der erforderlichen Wahrscheinlichkeiten bzw. der diesbezüglichen Tabellen ist der Anlage zu dieser Figurenbeschreibung zu entnehmen. The calculation of the required probabilities and the related tables can be found in the appendix to this figure description.

Claims

Patentansprüche claims 1. Verfahren zum Erstellen eines Kausalen Netzes auf Grundlage einer Wissensakquisition, d a d u r c h g e k e n n z e i c h n e t, dass die Wissensakquisition getrennt vom Erstellen des Kausalen Netzes erfolgt und folgende Schritte umfasst: Sammeln von relevantem Wissen, und1. Method for creating a causal network based on a knowledge acquisition, because the knowledge acquisition takes place separately from the creation of the causal network and comprises the following steps: collecting relevant knowledge, and Strukturieren des gesammelten Wissens in eine strukturierte Darstellung, die soweit vollständig ist, dass das Kausale Netz mittels eines Compilers automatisch erstellt werden kann.Structuring the collected knowledge into a structured representation that is so complete that the causal network can be created automatically using a compiler. 2. Verfahren nach Anspruch 1, dadurch gekennzeichnet, dass aus dem gesammelten Wissen mittels einer mathematischen Methode eine Untermenge derart erzeugt wird, dass die daraus resultierende Darstellung vollständig ist.2. The method according to claim 1, characterized in that a subset is generated from the collected knowledge by means of a mathematical method such that the resulting representation is complete. 3. Verfahren nach Anspruch 1 oder 2, dadurch gekennzeich- net, dass das relevante Wissen mittels eines Software-Tools gesammelt wird.3. The method according to claim 1 or 2, characterized in that the relevant knowledge is collected by means of a software tool. 4. Verfahren nach Anspruch 3, dadurch gekennzeichnet, dass das relevante Wissen mittels des Software-Tools unter Dialog- führung auf einer Anzeigeeinrichtung gesammelt wird.4. The method according to claim 3, characterized in that the relevant knowledge is collected by means of the software tool under the guidance of a dialog on a display device. 5. Verfahren nach Anspruch 4, bei dem Wissen auf medizinischem Gebiet gesammelt wird, dadurch gekennzeichnet, dass das Software-Tool zum Spezifizieren von Krankheiten und Be- funden, von Zusammenhängen zwischen Krankheiten und Befunden und von spezifischen Randwahrscheinlichkeiten und bedingten Wahrscheinlichkeiten und dazu ausgelegt ist, sicherzustellen, dass das gesammelte Wissen derart vollständig ist, dass das Kausale Netz mittels eines Compilers automatisch erstellt werden kann. 5. The method according to claim 4, in which knowledge is collected in the medical field, characterized in that the software tool is designed for specifying diseases and findings, for relationships between diseases and findings and for specific marginal and conditional probabilities to ensure that the knowledge gathered is so complete that the causal network can be created automatically using a compiler. 6. Verfahren nach Anspruch 5, dadurch gekennzeichnet, dass das Software-Tool die Krankheiten und die Befunde als sto- chastische Variable nutzt.6. The method according to claim 5, characterized in that the software tool uses the diseases and the findings as a stochastic variable. 7. Verfahren nach Anspruch 5 oder 6, dadurch gekennzeichnet, dass eine ausgewählte Krankheit, ihre Randwahrscheinlichkeit sowie zusätzliche Information auf der Anzeigeeinrichtung angezeigt werden.7. The method according to claim 5 or 6, characterized in that a selected disease, its marginal probability and additional information are displayed on the display device. 8. Verfahren nach Anspruch 7, dadurch gekennzeichnet, dass die zusätzliche Information fördernde und hemmende Faktoren der ausgewählten Krankheit enthält.8. The method according to claim 7, characterized in that the additional information contains promoting and inhibiting factors of the selected disease. 9. Verfahren nach Anspruch 8, dadurch gekennzeichnet, dass zum Quantifizieren von Wirkungen der fördernden und hemmenden9. The method according to claim 8, characterized in that for quantifying effects of the promoting and inhibitory Faktoren bedingte Wahrscheinlichkeiten spezifiziert werden.Factor probabilities are specified. 10. Verfahren nach einem der Ansprüche 5 bis 9, dadurch gekennzeichnet, dass die Symptome einer ausgewählten Krankheit zusammen mit der bedingten Wahrscheinlichkeit, dass diese Krankheit das Symptom verursacht, angezeigt werden. 10. The method according to any one of claims 5 to 9, characterized in that the symptoms of a selected disease are displayed together with the conditional probability that this disease causes the symptom.
EP02752975A 2001-07-03 2002-06-21 Method for creating a knowledge-based causal network Withdrawn EP1433130A2 (en)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
DE10132014 2001-07-03
DE10132014A DE10132014A1 (en) 2001-07-03 2001-07-03 Process for creating a causal network based on knowledge acquisition
PCT/DE2002/002280 WO2003005297A2 (en) 2001-07-03 2002-06-21 Method for creating a knowledge-based causal network

Publications (1)

Publication Number Publication Date
EP1433130A2 true EP1433130A2 (en) 2004-06-30

Family

ID=7690327

Family Applications (1)

Application Number Title Priority Date Filing Date
EP02752975A Withdrawn EP1433130A2 (en) 2001-07-03 2002-06-21 Method for creating a knowledge-based causal network

Country Status (4)

Country Link
US (1) US20040153429A1 (en)
EP (1) EP1433130A2 (en)
DE (1) DE10132014A1 (en)
WO (1) WO2003005297A2 (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1839382A (en) 2003-09-30 2006-09-27 英特尔公司 Most probable explanation generation for a dynamic Bayesian network
DE102004033614A1 (en) * 2004-07-12 2006-02-09 Emedics Gmbh Apparatus and method for estimating an occurrence probability of a health disorder
WO2007134495A1 (en) * 2006-05-16 2007-11-29 Zhan Zhang A method for constructing an intelligent system processing uncertain causal relationship information
JP4648484B2 (en) * 2006-12-07 2011-03-09 テレフオンアクチーボラゲット エル エム エリクソン（パブル） Apparatus and method for network management
US8135988B2 (en) * 2007-10-19 2012-03-13 Oracle International Corporation Non-intrusive gathering of diagnostic data using asynchronous mechanisms
US8417656B2 (en) * 2009-06-16 2013-04-09 Oracle International Corporation Techniques for building an aggregate model for performing diagnostics
US8171343B2 (en) 2009-06-16 2012-05-01 Oracle International Corporation Techniques for determining models for performing diagnostics
US8140898B2 (en) * 2009-06-16 2012-03-20 Oracle International Corporation Techniques for gathering evidence for performing diagnostics
US8612377B2 (en) * 2009-12-17 2013-12-17 Oracle International Corporation Techniques for generating diagnostic results
CN103745261B (en) * 2013-12-24 2015-04-15 张湛 Method for dynamic fault diagnosis by constructing three-dimensional DUCG intelligent system
US10866992B2 (en) 2016-05-14 2020-12-15 Gratiana Denisa Pol System and methods for identifying, aggregating, and visualizing tested variables and causal relationships from scientific research
CN107944562B (en) * 2017-10-17 2019-07-05 北京清睿智能科技有限公司 A kind of building method of the intelligence system of the uncertain causality category information of the processing of extension

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4771792A (en) * 1985-02-19 1988-09-20 Seale Joseph B Non-invasive determination of mechanical characteristics in the body
DE59108125D1 (en) * 1991-06-18 1996-10-02 Siemens Ag Knowledge-based diagnostic system with graphic knowledge acquisition component
US6208955B1 (en) * 1998-06-12 2001-03-27 Rockwell Science Center, Llc Distributed maintenance system based on causal networks

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See references of WO03005297A2 *

Also Published As

Publication number Publication date
WO2003005297A2 (en) 2003-01-16
WO2003005297A3 (en) 2004-04-22
US20040153429A1 (en) 2004-08-05
DE10132014A1 (en) 2003-01-23

Similar Documents

Publication Publication Date Title
DE69523588T2 (en) DEVICE AND METHOD FOR EVENT CORRELATION AND PROBLEM REPORTING
DE202018006897U1 (en) Dynamic, self-learning system for medical images
DE112019001136T5 (en) ANALYSIS OF ADVERSE MEDICINE EFFECTS
WO2003005297A2 (en) Method for creating a knowledge-based causal network
DE112019000747T5 (en) AUGMENTED REALITY TEMPLATE ASSOCIATED WITH A DISEASE AND / OR TREATMENT OF A PATIENT
Berg Formal tools and medical practices: Getting computer-based decision techniques to work
DE3833617A1 (en) AUXILIARY DEVICE FOR DIAGNOSIS OF DISEASES OF THE HEART VESSELS AND THE LUNG Vein SYSTEM
KR20200022113A (en) Oriental medicine prescription and health coordinator service system and method
Choudhury et al. A fuzzy logic-based expert system for determination of health risk level of patient
EP3739592A1 (en) Locally controlled imaging-based acquisition of patient data
De Vries et al. An overview of medical expert systems
DE102017205048B4 (en) DEVICE AND METHOD FOR DETERMINING A STATE OF A WORK PROCESS
Nikiforidis et al. Expert system support using Bayesian belief networks in the prognosis of head-injured patients of the ICU
Sierra et al. Medical bayes networks
Derni et al. An Advanced Heuristic Approach for the Optimization of Patient Flow in Hospital Emergency Department
Hosseinzadeh A rule-based system for vital sign monitoring in intensive care
Zagorecki et al. Online diagnostic system based on Bayesian networks
Viana et al. Do hybrid simulation models always increase flexibility to handle parametric and structural changes?
CN110752017B (en) Community doctor scheduling method and system based on deep learning
DE102007014970B3 (en) Process e.g. medical treatment, configuration method, involves configuring process step, where process includes parallel running process steps, and configuration of process steps determines sequence of process steps
Marakakis et al. Meta-rules and uncertain reasoning for diagnosis of epilepsy in childhood
Ivanović et al. Role of case-based reasoning in neurology decision support
DE112019003187T5 (en) COGNITIVE ANALYSIS AND DISAMBIGUATION OF ELECTRONIC PATIENT RECORDS TO REPRESENT RELEVANT INFORMATION FOR A MEDICAL TREATMENT PLAN
Harahap et al. Monitoring patient health based on medical records using fuzzy logic method
Rodas et al. Knowledge discovery in repeated very short serial measurements with a blocking factor. Application to a psychiatric domain

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20031223

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE TR

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20060214