Artificial Neural Networks in Biomedicine

Tutorial and Review.- 1 The Bayesian Paradigm: Second Generation Neural Computing.- 1.1 Introduction.- 1.2 Theory.- 1.2.1 Bayesian Learning.- 1.2.2 The Evidence Framework.- 1.2.2.1 Error bars.- 1.2.2.2 Moderated outputs.- 1.2.2.3 Regularisation.- 1.2.3 Committees.- 1.3 Example Results.- 1.4 Conclusion.- 2 The Role of the Artificial Neural Network in the Characterisation of Complex Systems and the Prediction of Disease.- 2.1 Introduction.- 2.2 Diagnosis of Disease.- 2.3 Outcome Prediction.- 2.4 Conclusion.- 3 Genetic Evolution of Neural Network Architectures.- 3.1 Introduction.- 3.2 Stability: The 'Bias/Variance Problem'.- 3.3 Genetic Algorithms and Artificial Neural Networks.- 3.3.1 Description of a General Method for Evolving ANN Architecture (EANN).- 3.3.2 Prediction of Depression After Mania.- 3.3.3 EANN and the Agreement/Transparency Choice.- 3.3.4 ANN and the Stability/Specialisation Choice.- 3.4 Conclusion.- Computer Aided Diagnosis.- 4 The Application of PAPNET to Diagnostic Cytology.- 4.1 Introduction.- 4.2 First Efforts at Automation in Cytology.- 4.3 Neural Networks.- 4.4 The PAPNET System.- 4.4.1 Components of the PAPNET System.- 4.4.1.1 Technical factors affecting the performance of the machine.- 4.4.2 Performance of the PAPNET System.- 4.4.2.1 Cervicovaginal smears.- 4.4.3 Application of the PAPNET System to Smears of Sputum.- 4.4.4 Application of the PAPNET System to Smears of Urinary Sediment.- 4.4.5 Application of the PAPNET System to Oesophageal Smears.- 4.5 Comment.- 5 ProstAsure Index - A Serum-Based Neural Network-Derived Composite Index for Early Detection of Prostate Cancer.- 5.1 Introduction.- 5.2 Clinical Background of Prostate Cancer and Derivation of the ProstAsure Index Algorithm.- 5.3 Validation of PI with Independent Clinical Data.- 5.4 Issues in Developing PI.- 5.5 Conclusion.- 6 Neurometric Assessment of Adequacy of Intraoperative Anaesthetic.- 6.1 Intraoperative Awareness.- 6.2 Measuring Sensory Perception.- 6.3 Clinical Data.- 6.4 Results.- 6.5 Implementation.- 6.6 Clinical Deployment.- 6.7 Healthcare Benefit.- 6.8 Additional Studies.- 7 Classifying Spinal Measurements Using a Radial Basis Function Network.- 7.1 Introduction.- 7.2 Data.- 7.2.1 The Spines.- 7.2.2 The Measurements.- 7.2.3 Preprocessing the Data.- 7.3 Radial Basis Functions and Networks.- 7.4 Matrix Notation.- 7.5 Training RBF Networks.- 7.5.1 The Unsupervised Learning Stage.- 7.5.2 The Supervised Learning Stage.- 7.5.2.1 Regularisation as an aid to avoid over-fitting.- 7.5.2.2 Calculating the regularisation coefficients and the weights.- 7.5.2.3 Forward subset selection of RBFs.- 7.5.2.4 Input feature selection.- 7.6 Results.- 7.7 Conclusion.- 8 GEORGIA: An Overview.- 8.1 Introduction.- 8.2 The Medical Decision Support System.- 8.3 Learning Pattern Generation.- 8.4 Software and Hardware Implementation.- 8.5 Re-Training and Re-Configuring the MDSS.- 8.6 Introducing GEORGIA's Man-to-Computer Interface.- 8.7 Conclusion.- 9 Patient Monitoring Using an Artificial Neural Network.- 9.1 Overview of the Medical Context.- 9.2 Basic Statistical Appraisal of Vital Function Data.- 9.3 Neural Network Details.- 9.3.1 Default Training.- 9.4 Implementation.- 9.5 Clinical Trials.- 9.6 Clinical Practice.- 10 Benchmark of Approaches to Sequential Diagnosis.- 10.1 Introduction.- 10.2 Preliminaries.- 10.3 Methods.- 10.3.1 The Probabilistic Algorithm.- 10.3.1.1 The diagnostic algorithm for first order markov chains - the Markov I algorithm.- 10.3.1.2 The diagnostic algorithm for second order markov chains - the Markov II algorithm.- 10.3.2 The Fuzzy Methods.- 10.3.2.1 The algorithm without context - fuzzy 0.- 10.3.2.2 The algorithm with first-order context - fuzzy lA.- 10.3.2.3 The reduced algorithm with first-order context - fuzzy 1B.- 10.3.2.4 The algorithm with second-order context - fuzzy 2A.- 10.3.2.5 The reduced algorithm with second-order context - fuzzy 2B.- 10.3.3 The Neural Network Approach.- 10.4 A Practical Example - Comparative Analysis of Methods.- 10.5 Conclusion.- 11 Application of Neural Networks in the Diagnosis of Pathological Speech.- 11.1 Introduction.- 11.2 The Research Material and the Problems Considered.- 11.2.1 Dental Prosthetics.- 11.2.2 Maxillofacial Surgery.- 11.2.3 Orthodontics.- 11.2.4 Laryngology.- 11.3 The Signal Parameterisation.- 11.4 The Application of the Neural Networks and the Results.- 11.5 Conclusion.- Signal Processing.- 12 Independent Components Analysis.- 12.1 Introduction.- 12.2 Theory.- 12.2.1 The Decorrelating Manifold.- 12.2.2 The Choice of Non-Linearity.- 12.2.3 Model-Order Estimation.- 12.3 Non-Stationary ICA.- 12.3.1 Illustration.- 12.4 Applications.- 12.4.1 Source Separation.- 12.4.2 Source Number and Estimation.- 12.5 Conclusion.- 13 Rest EEG Hidden Dynamics as a Discriminant for Brain Tumour Classification.- 13.1 Introduction.- 13.2 Characterising Hidden Dynamics.- 13.3 The Clinical Study.- 13.4 The Minimum Markov Order.- 13.5 Conclusion.- 14 Artifical Neural Network Control on Functional Electrical Stimulation Assisted Gait for Persons with Spinal Cord Injury.- 14.1 Introduction.- 14.2 Methods.- 14.3 Results.- 14.4 Discussion.- 15 The Application of Neural Networks to Interpret Evoked Potential Waveforms.- 15.1 Introduction.- 15.2 The Medical Conditions Studied.- 15.3 The Evoked Potentials.- 15.4 The Relationship Between the CNV and the Medical Conditions.- 15.5 Experimental Procedures.- 15.6 Data Pre-Processing.- 15.7 Feature Extraction.- 15.8 Normalisation.- 15.9 The Artificial Neural Networks.- 15.9.1 The Simplified Fuzzy ARTMAP.- 15.9.2 The Probabilistic Simplified Fuzzy ARTMAP.- 15.9.3 ANN Training and Accuracy.- 15.9.3.1 Small numbers of training vectors.- 15.9.3.2 Simplified fuzzy ARTMAP.- 15.9.3.3 Committees of ANNs.- 15.10 Validation Issues.- 15.10.1 Technical Aspects of Validation.- 15.10.2 Clinical Aspects of Validation.- 15.11 Results.- 15.12 Implementation Considerations.- 15.13 Future Developments.- Image Processing.- 16 Intelligent Decision Support Systems in the Cytodiagnosis of Breast Carcinoma.- 16.1 Introduction.- 16.2 Previous Work on Decision Support in this Domain.- 16.3 The Data Set in this Study.- 16.3.1 Study Population.- 16.3.2 Input Variables.- 16.3.3 Partitioning of the Data.- 16.4 Human Performance.- 16.5 Logistic Regression.- 16.6 Data Derived Decision Tree.- 16.7 Multi-Layer Perceptron Neural Networks.- 16.8 Adaptive Resonance Theory Mapping (ARTMAP) Neural Networks.- 16.8.1 Potential Advantages of ARTMAP.- 16.8.2 ARTMAP Architecture and Methodology.- 16.8.3 Results from the Cascaded System.- 16.8.4 Symbolic Rule Extraction.- 16.9 Assessment of the Different Decision Support Systems.- 17 A Neural-Based System for the Automatic Classificaton and Follow-Up of Diabetic Retinopathies.- 17.1 Introduction.- 17.2 The DRA System.- 17.3 Hybrid Module.- 17.4 Committee Algorithms.- 17.4.1 New Selection Algorithms.- 17.4.1.1 Greedy selection.- 17.4.1.2 Pseudo-exhaustive selection.- 17.4.2 Sequential Cooperation.- 17.4.3 Experimental Results.- 17.5 Related Work.- 17.6 Validation of the DRA System.- 17.7 Conclusion.- 18 Classification of Chromosomes: A Comparative Study of Neural Network and Statistical Approaches.- 18.1 Introduction.- 18.1.1 Chromosome Analysis and its Applications.- 18.1.2 Chromosome Classification.- 18.1.3 Experimental Data.- 18.2 The Neural Network Classifier.- 18.2.1 Representation of Chromosome Features.- 18.2.2 Network Topology and Training.- 18.2.3 Incorporating Non-Banding Features.- 18.3 Classification Performance.- 18.3.1 Classification Experiments.- 18.3.2 Comparison with Statistical Classifiers.- 18.3.3 The Influence of Training-Set Size.- 18.4 The Use of Context in Classification.- 18.4.1 The Karyotyping Constraint.- 18.4.2 Applying the Constraint by a Network.- 18.4.3 Results of Applying the Context Network.- 18.5 Conclusion and Discussion.- 18.5.1 Comparison with Statistical Classifiers.- 18.5.2 Training Set Size and Application of Context.- 18.5.3 Biological Context.- 19 The Importance of Features and Primitives for Multi-dimensional/Multi-channel Image Processing.- 19.1 Introduction.- 19.2 The Image Data Level.- 19.3 From Image Data to Symbolic Primitives.- 19.4 Region Segmentation Quality and Training Phase.- 19.5 Validation of Image Segmentation.- 19.6 Segmentation Complexity and Quantitative Error Evaluation.- 19.7 Feature Description.- 19.8 Feature Selection.- 19.9 A Preliminary Overview of Application Results.- 19.10 Conclusion.