Neural networks and artificial intelligence for biomedical engineering

Using examples drawn from biomedicine and biomedical engineering, this essential reference book brings you comprehensive coverage of all the major techniques currently available to build computer-assisted decision support systems. You will find practical solutions for biomedicine based on current theory and applications of neural networks, artificial intelligence, and other methods for the development of decision aids, including hybrid systems.

Neural Networks and Artificial Intelligence for Biomedical Engineering offers students and scientists of biomedical engineering, biomedical informatics, and medical artificial intelligence a deeper understanding of the powerful techniques now in use with a wide range of biomedical applications.

Highlighted topics include:

Types of neural networks and neural network algorithms Knowledge representation, knowledge acquisition, and reasoning methodologies Chaotic analysis of biomedical time series Genetic algorithms Probability-based systems and fuzzy systems Evaluation and validation of decision support aids

Author Notes

Donna L. Hudson is professor of Family and Community Medicine at the University of California, San Francisco (UCSF), and Director of Medical Information Resources at the UCSF Fresno Medical Education Program. She is also a member of the executive committee for the Medical Information Sciences Program at UCSF and a member of the Bioengineering Graduate Group at UC Berkeley and UCSF. In 1987 Dr. Hudson received the Faculty Research Award at UCSF. Dr. Hudson is a senior member of the Institute of Electrical and Electronics Engineers (IEEE), a member of the Administrative Committee of IEEE Engineering in Medicine and Biology Society, a Fellow of the American Institute of Medical and Biological Engineering, and president of the International Society for Computers and Their Applications.
Maurice E. Cohen is professor of Radiology at the University of California, San Francisco; professor of Mathematics at California State University, Fresno; professor and Founding Member of the Graduate Group in Medical Information Science, and a member of the Bioengineering Graduate Group at UC Berkeley and UCSF. In 1977 Dr. Cohen solved a problem involving Jacobi functions that was believed to be impossible. He received the 1985 American Medical Informatics Association Best Paper Award, the 1991 Outstanding Professor Award at CSUF, and the 1996 Faculty Research Award at UCSF. Dr. Cohen is a Fellow of the American Institute for Medical and Biological Engineering and treasurer of the International Society for Computers and Their Applications.

Reviews 1

Choice Review

Biomedical engineering evokes the idea of bionic people, but on a more mundane and practical level it treats the technology needed to make decisions about medical diagnosis and treatment. Hudson and Cohen (Univ. of California, San Francisco) broadly survey the computational methodologies for such decision support systems. The book's three major sections discuss neural networks, artificial intelligence, and alternate approaches. The first section treats the various kinds of neural nets, their learning algorithms, and the problems involved in actually using such models; the section on artificial intelligence discusses more traditional AI systems. While some of these systems have been somewhat successful in diagnosis, the authors point out the many difficulties that still must be faced to make such systems practical. The section on alternative approaches throws together statistics, genetic algorithms, and fuzzy systems. The authors also briefly describe their system, which has features from all of the approaches discussed. This broad survey is an admirable attempt to give practitioners an overview of the available approaches to medical decision making. Unfortunately, this book suffers from the problems of breadth--coverage is too shallow; examples are too limited; systems are mentioned rather than described; there are errors in details; and many of the formulas contain typos or undefined symbols. Graduate students; faculty. P. Cull; Oregon State University

Preface	p. xxi
Acknowledgments	p. xxiii
Overview	p. 1
0.1 Early Biomedical Systems	p. 1
0.1.1 History	p. 1
0.1.2 Medical Records	p. 2
0.1.3 Drawbacks of Traditional Approaches	p. 3
0.1.4 Numerical versus Symbolic Approaches	p. 3
0.2 Medical and Biological Data	p. 3
0.2.1 Binary data	p. 4
0.2.2 Categorical data	p. 4
0.2.3 Integer Data	p. 4
0.2.4 Continuous Data	p. 4
0.2.5 Fuzzy Data	p. 4
0.2.6 Temporal Data	p. 5
0.2.7 Time Series Data	p. 6
0.2.8 Image Data	p. 8
0.3 Organization of the Book	p. 8
Part I Neural Networks
Chapter 1 Foundations of Neural Networks	p. 13
1.1 Objectives of Neural Networks	p. 13
1.1.1 Modeling Biomedical Systems	p. 13
1.1.2 Establishment of Decision-Making Systems	p. 14
1.2 Biological Foundations of Neural Networks	p. 14
1.2.1 Structure of the Neuron	p. 14
1.2.2 Structure of the Central Nervous System	p. 16
1.3 Early Neural Models	p. 16
1.3.1 McCulloch and Pitts Neuron	p. 16
1.3.2 Hebbian Learning	p. 17
1.3.3 ADALINE	p. 17
1.3.4 Rosenblatt Perceptron	p. 17
1.3.5 Problems with Early Systems	p. 18
1.4 Precursor to Current Models: Pattern Classification	p. 19
1.4.1 Feature Extraction	p. 19
1.4.2 Supervised Learning	p. 21
1.4.3 Unsupervised Learning	p. 21
1.4.4 Learning Algorithms	p. 22
1.5 Resurgence of the Neural Network Approach	p. 24
1.6 Basic Concepts	p. 25
1.6.1 Artificial Neurons	p. 25
1.6.2 Selection of Input Nodes	p. 25
1.6.3 Network Structure	p. 25
1.6.4 Learning Mechanism	p. 26
1.6.5 Output	p. 26
1.7 Summary	p. 26
Chapter 2 Classes of Neural Networks	p. 29
2.1 Basic Network Properties	p. 29
2.1.1 Terminology	p. 29
2.1.2 Structure of Networks	p. 29
2.1.3 Computational Properties of Nodes	p. 30
2.1.4 Algorithm Design	p. 31
2.2 Classification Models	p. 32
2.3 Association Models	p. 32
2.3.1 Hopfield Nets	p. 33
2.3.2 Other Associative Memory Approaches	p. 34
2.3.3 Hamming Net	p. 36
2.3.4 Applications of Association Models	p. 37
2.4 Optimization Models	p. 38
2.4.1 Hopfield Net	p. 38
2.4.2 Boltzmann Machines	p. 39
2.4.3 Applications of Optimization Models	p. 40
2.5 Self-Organization Models	p. 40
2.6 Radial Basis Functions (RBFs)	p. 41
2.6.1 Theoretical Basis	p. 41
2.6.2 Applications of Radial Basis Functions	p. 42
2.7 Summary	p. 43
Chapter 3 Classification Networks and Learning	p. 45
3.1 Network Structure	p. 45
3.1.1 Layer Definition	p. 45
3.1.2 Input Layer	p. 45
3.1.3 Hidden Layer	p. 45
3.1.4 Output Layer	p. 45
3.2 Feature Selection	p. 46
3.2.1 Types of Variables	p. 46
3.2.2 Feature Vectors	p. 46
3.2.3 Image Data	p. 47
3.2.4 Time Series Data	p. 47
3.2.5 Issues of Dimensionality	p. 50
3.3 Types of Learning	p. 51
3.3.1 Supervised Learning	p. 51
3.3.2 Unsupervised Learning	p. 52
3.3.3 Causal Models	p. 55
3.4 Interpretation of Output	p. 55
3.5 Summary	p. 55
Chapter 4 Supervised Learning	p. 59
4.1 Decision Surfaces	p. 59
4.2 Two-Category Separation, Linearly Separable Sets	p. 60
4.2.1 Fisher's Linear Discriminant	p. 60
4.2.2 Gradient Descent Procedures	p. 61
4.2.3 Perceptron Algorithm	p. 62
4.2.4 Relaxation Procedures	p. 62
4.2.5 Potential Functions	p. 63
4.3 Nonlinearly Separable Sets	p. 64
4.3.1 Nonlinear Discriminant Functions	p. 64
4.3.2 Hypernet, A Nonlinear Potential Function Algorithm	p. 64
4.3.3 Categorization of Nonlinearly Separable Sets	p. 65
4.4 Multiple Category Classification Problems	p. 66
4.4.1 Extension of Fisher Discriminant	p. 66
4.4.2 Kesler Construction	p. 67
4.4.3 Backpropagation	p. 68
4.5 Relationship to Neural Network Models	p. 69
4.6 Comparison of Methods	p. 70
4.6.1 Convergence and Stability	p. 70
4.6.2 Training Time	p. 70
4.6.3 Predictive Power	p. 70
4.7 Applications	p. 71
4.7.1 Single-Category Classification	p. 71
4.7.2 Multicategory Classification	p. 72
4.7.3 Reduction of Nodes	p. 74
4.8 Summary	p. 74
Chapter 5 Unsupervised Learning	p. 79
5.1 Background	p. 79
5.2 Clustering	p. 79
5.2.1 Basic Isodata	p. 79
5.2.2 Similarity Measures	p. 80
5.2.3 Criterion Functions	p. 80
5.2.4 Hierarchical Clustering	p. 82
5.2.5 Metrics	p. 82
5.3 Kohonen Networks and Competitive Learning	p. 83
5.4 Hebbian Learning	p. 85
5.5 Adaptive Resonance Theory (ART)	p. 86
5.6 Applications	p. 88
5.6.1 Dimensionality Reduction	p. 88
5.6.2 Biomedical Applications	p. 89
5.6.3 Diagnosis of CAD as a Clustering Problem	p. 89
5.6.4 Other Biomedical Applications	p. 90
5.7 Summary	p. 91
Chapter 6 Design Issues	p. 95
6.1 Introduction	p. 95
6.1.1 Objective of the Model	p. 95
6.1.2 Information Sources	p. 95
6.2 Input Data Types	p. 96
6.2.1 Extracting Information from the Medical Record	p. 96
6.2.2 Using Information from Data Collection Sheets	p. 97
6.2.3 Time Series Data	p. 100
6.2.4 Image Data	p. 101
6.3 Structure of Networks	p. 101
6.3.1 Number of Layers	p. 101
6.3.2 Connectivity	p. 101
6.4 Implications of Network Structures	p. 101
6.4.1 Classification Potential	p. 101
6.4.2 Training	p. 103
6.4.3 Reduction in Number of Nodes	p. 104
6.4.4 Output	p. 105
6.5 Choice of Learning Algorithm	p. 105
6.6 Summary	p. 106
Chapter 7 Comparative Analysis	p. 109
7.1 Introduction	p. 109
7.2 Input Data Considerations	p. 109
7.3 Supervised Learning Algorithms	p. 110
7.3.1 Gradient Descent Procedures	p. 111
7.3.2 Extensions to Nonlinear Decision Functions	p. 111
7.3.3 Extensions to Multiple Categories	p. 113
7.4 Unsupervised Learning	p. 114
7.4.1 Clustering Methods	p. 114
7.4.2 Self-Organization Networks	p. 115
7.5 Network Structures	p. 115
7.5.1 Number of Categories	p. 115
7.5.2 Connectivity	p. 115
7.6 Interpretation of Results	p. 116
7.6.1 Supervised Learning	p. 116
7.6.2 Unsupervised Learning	p. 117
7.6.3 Data Scaling and Normalization	p. 118
7.6.4 Dependence on Training Data	p. 118
7.7 Summary	p. 118
Chapter 8 Validation and Evaluation	p. 121
8.1 Introduction	p. 121
8.2 Data Checking	p. 121
8.2.1 Verification of Accuracy of Data for Training	p. 121
8.2.2 Appropriateness of Data for Training	p. 122
8.2.3 Use of Gold Standards in Supervised Learning	p. 123
8.3 Validation of Learning Algorithm	p. 123
8.3.1 Technical Integrity of Algorithm	p. 123
8.3.2 Appropriateness of Algorithm for Given Application	p. 123
8.4 Evaluation of Performance	p. 124
8.4.1 Supervised Learning Algorithms	p. 124
8.4.2 Unsupervised Learning	p. 124
8.5 Summary	p. 126
Part II Artificial Intelligence
Chapter 9 Foundations of Computer-Assisted Decision Making	p. 131
9.1 Motivation for Computer-Assisted Decision Making	p. 131
9.2 Databases and Medical Records	p. 131
9.2.1 The First Decade (1970-1980)	p. 131
9.2.2 The Second Decade (1980-1990)	p. 134
9.2.3 Current Approaches to Medical Databases	p. 134
9.3 Mathematical Modeling and Simulation	p. 135
9.4 Pattern Recognition	p. 136
9.5 Bayesian Analysis	p. 137
9.5.1 Early Bayesian Systems	p. 137
9.5.2 Bayesian Belief Networks	p. 138
9.6 Decision Theory	p. 138
9.7 Symbolic Reasoning Techniques	p. 139
9.7.1 Early Expert Systems	p. 139
9.7.2 Second-Generation Expert Systems	p. 142
9.8 Summary	p. 144
Chapter 10 Knowledge Representation	p. 151
10.1 Production Rules	p. 151
10.1.1 General Structure	p. 151
10.1.2 Methods of Confirming Conditions	p. 152
10.1.3 Rule Searching Strategies	p. 153
10.1.4 Expanded Production Rule Systems	p. 155
10.1.5 Certainty Factors	p. 156
10.1.6 Advantages of Production Systems	p. 158
10.1.7 Disadvantages of Production Systems	p. 158
10.1.8 Areas of Applications of Production Rule Systems	p. 160
10.2 Frames	p. 160
10.2.1 General Structure	p. 160
10.2.2 Inheritance	p. 161
10.2.3 Example of a Frame-Based Decision-Support System	p. 161
10.3 Databases	p. 162
10.3.1 Relational Databases	p. 163
10.3.2 Query Languages	p. 165
10.3.3 Object-Oriented Databases	p. 165
10.4 Predicate Calculus and Semantic Nets	p. 166
10.4.1 Syntax	p. 166
10.4.2 Method of Proof	p. 167
10.4.3 Semantic Trees	p. 167
10.4.4 Semantic Interpretation	p. 167
10.4.5 Applications	p. 167
10.5 Temporal Data Representations	p. 168
10.5.1 Knowledge-Based Systems	p. 168
10.5.2 Data-Based Systems	p. 169
10.6 Summary	p. 169
Chapter 11 Knowledge Acquisition	p. 173
11.1 Introduction	p. 173
11.2 Expert Input	p. 173
11.2.1 Acquisition for a New System	p. 173
11.2.2 Updating of Information	p. 177
11.3 Learned Knowledge	p. 178
11.3.1 Rote Learning	p. 178
11.3.2 Learning by Being Told	p. 178
11.3.3 Learning from Examples	p. 178
11.4 Meta-Knowledge	p. 181
11.5 Knowledge Base Maintenance	p. 182
11.5.1 Assuring Accuracy	p. 182
11.5.2 Maintaining Consistency	p. 183
11.6 Summary	p. 183
Chapter 12 Reasoning Methodologies	p. 185
12.1 Introduction	p. 185
12.2 Problem Representations	p. 185
12.2.1 Graphs and Trees	p. 185
12.2.2 Binary Trees	p. 187
12.2.3 State-Space Representations	p. 188
12.2.4 Problem Reduction Representations	p. 189
12.2.5 Game Trees	p. 190
12.3 Blind Searching	p. 190
12.3.1 Depth-First Search	p. 190
12.3.2 Breadth-First Search	p. 191
12.4 Ordered Search	p. 191
12.4.1 Uniform Cost Method	p. 192
12.4.2 Using Heuristic Information	p. 192
12.5 AND/OR Trees	p. 193
12.5.1 Breadth-First Search and Depth-First of AND/OR Tree	p. 194
12.5.2 Costs of Solution Trees	p. 194
12.5.3 Ordered Searching of AND/OR Trees	p. 195
12.6 Searching Game Trees	p. 196
12.6.1 Minimax	p. 196
12.6.2 Alpha-Beta	p. 196
12.7 Searching Graphs	p. 196
12.7.1 Breadth-First Graph Searching	p. 197
12.7.2 Uniform Cost Algorithm	p. 197
12.7.3 AND/OR Graphs	p. 197
12.8 Rule Base Searching	p. 197
12.8.1 Backward-Chaining	p. 197
12.8.2 Forward-Chaining	p. 198
12.9 Higher-Level Reasoning Methodologies	p. 198
12.9.1 Inference Engines	p. 198
12.9.2 Cognitive Models	p. 200
12.9.3 Automatic Deduction	p. 200
12.9.4 Natural Language Processing	p. 200
12.10 Examples in Biomedical Expert Systems	p. 201
12.11 Summary	p. 201
Chapter 13 Validation and Evaluation	p. 205
13.1 Introduction	p. 205
13.2 Algorithmic Evaluation	p. 205
13.2.1 Searching Algorithm	p. 205
13.2.2 Inference Engine	p. 206
13.2.3 Aggregation of Evidence	p. 207
13.2.4 Use of Meta Knowledge	p. 207
13.3 Knowledge Base Evaluation	p. 207
13.3.1 Expert-Derived Information	p. 207
13.3.2 Learned Information	p. 208
13.3.3 Meta Knowledge	p. 208
13.4 System Evaluation	p. 208
13.4.1 Evaluation with Original Knowledge Base	p. 209
13.4.2 Evaluation with Updated Knowledge Base	p. 209
13.4.3 Evaluation of User Interaction Parameters	p. 209
13.4.4 Validation with Clinical Data	p. 210
13.5 Summary	p. 212
Part III Alternative Approaches
Chapter 14 Genetic Algorithms	p. 217
14.1 Foundations	p. 217
14.2 Representation Schemes	p. 218
14.3 Evaluation Functions	p. 218
14.4 Genetic Operators	p. 218
14.4.1 Mutation	p. 218
14.4.2 Crossover	p. 218
14.5 Evolution Strategies	p. 219
14.5.1 Genetic Algorithms	p. 219
14.5.2 Optimization Strategies	p. 219
14.5.3 Genetic Search	p. 220
14.6 Biomedical Examples	p. 221
14.6.1 Literature References	p. 221
14.7 Summary	p. 222
Chapter 15 Probabilistic Systems	p. 225
15.1 Introduction	p. 225
15.2 Bayesian Approaches	p. 225
15.2.1 Bayes' Rule	p. 225
15.2.2 Bayes' Decision Theory	p. 226
15.2.3 Risk Analysis	p. 226
15.2.4 Supervised Bayesian Learning	p. 228
15.2.5 Decision Trees	p. 229
15.3 Parameter Estimation	p. 231
15.3.1 Maximum Likelihood Estimation	p. 231
15.3.2 Bayesian Estimation	p. 232
15.4 Discriminant Analysis	p. 233
15.5 Statistical Pattern Classification	p. 234
15.6 Unsupervised Learning	p. 235
15.6.1 Parzen Windows	p. 235
15.6.2 Nearest Neighbor Algorithms	p. 236
15.6.3 Mixture Densities and Maximum Likelihood Estimates	p. 236
15.6.4 Unsupervised Bayesian Learning	p. 237
15.7 Regression Analysis	p. 238
15.8 Biomedical Applications	p. 240
15.9 Summary	p. 240
Chapter 16 Fuzzy Systems	p. 243
16.1 Introduction	p. 243
16.2 Fuzzy Information	p. 243
16.2.1 Input Data	p. 243
16.2.2 Fuzzy Logic and Fuzzy Set Theory	p. 244
16.2.3 Representation of Fuzzy Variables	p. 244
16.2.4 Membership Functions	p. 245
16.3 Fuzzy Neural Networks	p. 245
16.4 Fuzzy Approaches for Supervised Learning Networks	p. 246
16.4.1 Pre-Processing of Fuzzy Input	p. 246
16.4.2 Propagation of Results	p. 247
16.5 Fuzzy Generalizations of Unsupervised Learning Methods	p. 248
16.5.1 Fuzzy Associative Memories	p. 248
16.5.2 Fuzzy Clustering	p. 248
16.6 Reasoning with Uncertain Information	p. 249
16.6.1 Uncertainty in Input Data	p. 250
16.6.2 Uncertainty in Knowledge Base	p. 250
16.6.3 Inference Engines for Uncertain Information	p. 251
16.6.4 Evidential Reasoning	p. 252
16.6.5 Compatibility Indices	p. 254
16.6.6 Approximate Reasoning	p. 255
16.7 Pre-Processing and Post-Processing Using Fuzzy Techniques	p. 256
16.8 Applications in Biomedical Engineering	p. 257
16.9 Summary	p. 258
Chapter 17 Hybrid Systems	p. 261
17.1 Hybrid Systems Approaches	p. 261
17.2 Components of Hybrid Systems	p. 261
17.2.1 Knowledge-Based Approaches	p. 262
17.2.2 Data-Based Approaches	p. 264
17.2.3 General Methodologies	p. 266
17.3 Use of Complex Data Structures	p. 267
17.3.1 Time Series Data	p. 267
17.3.2 Image Data	p. 268
17.4 Design Methodologies	p. 268
17.4.1 System Structure	p. 268
17.4.2 User Interfaces	p. 269
17.4.3 Pre-Processing	p. 269
17.4.4 Post-Processing	p. 269
17.4.5 Presentation of Results	p. 270
17.5 Summary	p. 270
Chapter 18 HyperMerge, a Hybrid Expert System	p. 273
18.1 Introduction	p. 273
18.2 Knowledge-Based Component	p. 273
18.2.1 Crisp Implementation	p. 273
18.2.2 Partial Substantiation of Antecedents	p. 274
18.2.3 Weighted Antecedents and Partial Substantiation of Rules	p. 275
18.2.4 Handling of Temporal Data	p. 276
18.3 Neural Network Component	p. 276
18.3.1 Learning Algorithm	p. 276
18.3.2 Special Data Types	p. 277
18.4 Analysis of Time Series Data	p. 278
18.4.1 Chaos Theory	p. 278
18.4.2 Continuous versus Discrete Chaotic Modeling	p. 279
18.4.3 Difference Equations and Graphs	p. 280
18.4.4 Central Tendency Measure	p. 281
18.5 Combined System	p. 282
18.5.1 Weighting of Antecedents	p. 282
18.5.2 Determination of Thresholds	p. 282
18.5.3 Neural Network with Symbolic Layer	p. 283
18.6 Application: Diagnosis of Heart Disease	p. 284
18.6.1 Categories of Heart Disease	p. 284
18.6.2 Knowledge-Based Information	p. 285
18.6.3 Data-Based Information	p. 285
18.6.4 Chaotic Data	p. 285
18.6.5 Sample System	p. 288
18.7 Summary	p. 289
Chapter 19 Future Perspectives	p. 291
19.1 Introduction	p. 291
19.2 Effects of Hardware Advances	p. 291
19.2.1 Faster Computing Speeds	p. 291
19.2.2 Increased Memory	p. 292
19.2.3 Parallel Machines	p. 292
19.2.4 Miniaturization	p. 292
19.2.5 Organic Semiconductors	p. 292
19.3 Effects of Increase in Knowledge	p. 293
19.3.1 Information Explosion	p. 293
19.3.2 Human Genome Project	p. 293
19.3.3 Proliferation of Databases	p. 293
19.3.4 Communication of Information	p. 294
19.4 The Future of Software	p. 294
19.4.1 Hybrid Systems	p. 294
19.4.2 Parallel Systems	p. 294
19.4.3 Nontextual Data	p. 295
19.4.4 Neural Network Models	p. 295
19.4.5 Artificial Intelligence Approaches	p. 295
Index	p. 297
About the Authors	p. 305

Available:*

On Order

Summary

Summary

Author Notes

Reviews 1

Choice Review

Table of Contents