Computer-Based Medical Consultations: MYCIN

PrefaceForewordChapter 1. Introduction 1.1 Computer Applications in Medicine 1.1.1 Problems and Promises 1.1.2 Medical Computing Application Areas 1.2 Artificial Intelligence 1.2.1 Areas of Application 1.2.2 AI Methodologies and Techniques 1.3 Computer-Assisted Medical Decision Making 1.3.1 Major Problem Area 1.3.2 Data Retrieval as a Decision Aid 1.3.3 Decisions Based on Numerical Computations 1.3.4 Probabilistic Approaches to Decision Making 1.3.5 Artificial Intelligence and Medical Decisions 1.3.6 Philosophical Observations 1.4 Antimicrobial Selection 1.4.1 Nature of the Decision Problem 1.4.2 Evidence that Assistance is Needed 1.5 MYCIN System 1.5.1 System's Organization 1.5.2 Sample Consultation Session Chapter 2. Design Considerations for Mycin 2.1 Introduction 2.2 Design Considerations for Consultation Programs 2.2.1 Program Should be Useful 2.2.2 Program Should be Educational when Appropriate 2.2.3 Program Should be Able to Explain Its Advice 2.2.4 Program Should be Able to Understand Questions 2.2.5 Program Should be Able to Acquire New Knowledge 2.2.6 Program's Knowledge-Base Should be Easily Modified 2.3 MYCIN and Acceptability Criteria 2.3.1 Modularity to Insure Straightforward Modification 2.3.2 Ability to Acquire New Knowledge from Experts 2.3.3 Ability to Understand Questions 2.3.4 Ability to Explain Decisions 2.3.5 Educational Capabilities 2.3.6 General UsefulnessChapter 3. Consultation System 3.1 Introduction 3.2 System Knowledge 3.2.1 Decision Rules 3.2.2 Categorization of Rules by Context 3.2.3 Clinical Parameters 3.2.4 Certainty Factors 3.2.5 Functions for Evaluation of Premise Conditions 3.2.6 Static Knowledge Structures 3.2.7 Translation of Rules into English 3.3 Use of Rule s to Give Advice 3.3.1 Previous Goal-Oriented Problem Solving Systems 3.3.2 MYCIN'S Control Structure 3.3.3 Creation of Dynamic Data Base 3.3.4 Self-Referencing Rules 3.3.5 Preventing Reasoning Loops 3.4 Propagation of Context Tree 3.4.1 Data Structures Used for Sprouting Branches 3.4.2 Explicit Mechanisms for Branching 3.4.3 Implicit Mechanisms for Branching 3.5 Selection of Therapy 3.5.1 Creation of Potential Therapy List 3.5.2 Selecting Preferred Drug from List 3.6 Mechanism s for Storage of Patient Data 3.6.1 Changing Answers to Questions 3.6.2 Remembering Patients for Future Reference 3.7 Future Extension s 3.7.1 Dynamic Ordering of Rules 3.7.2 Dynamic Ordering of Conditions within Rules 3.7.3 Pre-Screening of Rules 3.7.4 Placing all Knowledge in Rules 3.7.5 Need for Context Graph 3.8 Advantage s of MYCIN Approach 3.8.1 Modularity of Knowledge 3.8.2 Dynamic Reasoning Chain 3.8.3 Domain-Independent Control Structure 3.8.4 Reasoning with Judgmental KnowledgeChapter 4. Model of Inexact Reasoning in Medicine 4.1 Introduction 4.2 Problem Formulation 4.3 MYCIN'S Rule-Based Approach 4.4 Theoretical Background 4.5 Proposed Model of Evidential Strength 4.6 Model as Approximation Technique 4.7 MYCIN'S Use of ModelChapter 5. Explanation System 5.1 Introduction 5.2 Using Question-Answering System 5.2.1 Rule-Retrieval Questions 5.2.2 Questions Regarding Dynamic Base Data 5.2.3 Additional Options 5.3 Future ExtensionsChapter 6. Future Directions for Mycin 6.1 Introduction 6.2 Plans for Immediate Future 6.3 Knowledge Acquisition 6.3.1 Current Status of Rule-Acquisition 6.3.