Computational Intelligence Processing in Medical Diagnosis

by Schmitt, Manfred; Teodorescu, Horia-Nicolai; Jain, Ashlesha; Jain, Ajita; Jain, Sandyha

• ISBN13: 9783790814637
• ISBN10: 3790814636
• Format: Hardcover
• Copyright: 2002-05-01
• Publisher: Physica Verlag

Summary

Computational intelligence techniques are gaining momentum in the medical prognosis and diagnosis. This volume presents advanced applications of machine intelligence in medicine and bio-medical engineering. Applied methods include knowledge bases, expert systems, neural networks, neuro-fuzzy systems, evolvable systems, wavelet transforms, and specific internet applications. The volume is written in view of explaining to the practitioner the fundamental issues related to computational intelligence paradigms and to offer a fast and friendly-managed introduction to the most recent methods based on computer intelligence in medicine.

Table of Contents

An introduction to computational intelligence in medical diagnosis
H.-N. Teodorescu
L.C. Jain
What is computational intelligence?	1	(2)
Why CI in medicine and especially in medical diagnosis?	3	(1)
CI in medical diagnosis	4	(2)
Data mining and knowledge discovery	6	(1)
Qualitative reasoning methods	6	(1)
Issues related to CI management in medicine	6	(1)
The prospects of CI in medicine	7	(6)
Acknowledgments	9	(1)
References	10	(3)
Computational intelligence techniques in medical decision making: the data mining perspective
V. Maojo
J. Sanandres
H. Billhardt
J. Crespo
Background - artificial intelligence in medicine	13	(3)
Data mining	16	(15)
Knowledge discovery in databases	16	(2)
Methods	18	(3)
Statistics and pattern recognition	21	(2)
Machine learning	23	(4)
Artificial neural networks (ANNs)	27	(3)
Data mining tools	30	(1)
Applications in medicine	31	(3)
Limitations of data mining in medicine	34	(2)
Conclusions	36	(9)
References	37	(8)
Internet-based decision support for evidence-based medicine
J. Simpson
J.K.C. Kingston
N. Molony
Introduction	45	(2)
The protocol assistant - feasibility assessment	47	(7)
Feasibility: organizational issues	48	(2)
Feasibility: technical issues	50	(2)
Feasibility: project & personnel issues	52	(2)
Representing clinical protocols	54	(6)
Knowledge Acquisition and Modeling Using PROforma	54	(2)
``Running'' a clinical protocol using JESS	56	(2)
Representing and reasoning with clinical uncertainty	58	(2)
Design and implementation of the protocol assistant	60	(5)
System design	60	(3)
User interface design	63	(1)
Implementation	64	(1)
Evaluation and future work	65	(6)
Acknowledgments	68	(1)
References	68	(3)
Integrating kernel methods into a knowledge-based approach to evidence-based medicine
K. Morik
T. Joachims
M. Imhoff
P. Brockhausen
S. Ruping
Introduction	71	(5)
Data acquisition and data set	76	(4)
Data acquisition	76	(2)
Data set	78	(1)
Statistical preprocessing	79	(1)
Data-driven acquisition of state-action rules	80	(7)
Support vector machine	80	(1)
Learning the directions of interventions	81	(1)
Learning when to intervene	82	(2)
SVM rules in evidence based medicine	84	(1)
More learning tasks	85	(2)
Medical knowledge base	87	(6)
Knowledge acquisition and representation	87	(4)
Validating action-effect rules	91	(1)
Integrating learned decision functions with the knowledge base	92	(1)
Using the knowledge base of effects to validate interventions	93	(2)
Validating learned decision rules	93	(1)
Validating proposed interventions	94	(1)
Comparison with related work	95	(1)
Conclusions	96	(5)
Acknowledgements	97	(1)
References	97	(4)
Case-based reasoning prognosis for temporal courses
R. Schmidt
L. Gierl
Introduction	101	(1)
Methods	102	(4)
Case-based reasoning	102	(2)
Prognostic model	104	(1)
State abstraction	105	(1)
Temporal abstraction	105	(1)
CBR retrieval	106	(1)
Applications	106	(17)
Kidney function courses	107	(1)
Objectives	107	(1)
Methods	108	(7)
Learning a tree of prototypes	115	(3)
Evaluation	118	(2)
Prognosis of the spread of diseases	120	(2)
Searching for similar courses	122	(1)
Adaptation	122	(1)
Generalization of our prognostic method	123	(2)
Summary	125	(4)
References	125	(4)
Pattern recognition in intensive care online monitoring
R. Fried
U. Gather
M. Imhoff
Introduction	129	(2)
Curve fitting	131	(6)
Median filtering	137	(2)
Statistical time series analysis	139	(4)
Intervention analysis	143	(6)
Statistical process control	149	(7)
Online pattern recognition based on statistical time series analysis	156	(7)
Dynamic linear models	156	(2)
ARMA modeling	158	(2)
Trend detection	160	(3)
Conclusion	163	(10)
Acknowledgements	165	(1)
References	165	(8)
Artificial neural network models for timely assessment of trauma complication risk
R. P. Marble
J.C. Healy
Artificial neural network models	173	(7)
Background	173	(3)
Neural networks and statistical analysis	176	(2)
Neural networks in medicine	178	(2)
A neural network model for predicting the incidence of coagulopathy in victims of blunt injury trauma	180	(6)
Model description	181	(1)
Results	182	(2)
Remarks	184	(2)
Prospects for refining and utilizing neural models in trauma care settings	186	(11)
Sensitivity analysis, pruning, and rule extraction	186	(2)
Trauma systems development	188	(1)
References	189	(8)
Artificial neural networks in medical diagnosis
Y. Fukuoka
Introduction	197	(1)
Foundations of artificial neural networks	198	(7)
Artificial neuron	198	(1)
Network architectures	198	(1)
Learning algorithms	199	(1)
Back-propagation	200	(3)
Self-organizing map	203	(2)
Applications to biomedicine	205	(16)
Pattern classification with BP	206	(1)
Clinical data	207	(1)
Bioelectric signals	208	(2)
Image analysis	210	(1)
Pattern classification with SOM	211	(2)
Data compression with BP	213	(1)
System modeling with BP	214	(2)
More detailed reviews	216	(1)
Chronic stress evaluation using ANNs	216	(3)
Gene expression data analysis with SOM	219	(2)
Conclusion	221	(8)
References	222	(7)
The application of neural networks in the classification of the electrocardiogram
C.D. Nugent
J.A. Lopez
N.D. Black
J.A.C. Webb
Introduction to the classification of the electrocardiogram	229	(4)
Diagnostic utilities of the ECG	230	(1)
Introduction to computerized classification	231	(2)
Fundamentals of the 12-lead ECG	233	(4)
The 12-lead ECG and associated nomenclature	234	(3)
Computerized classification of the 12-lead ECG	237	(4)
Classification	239	(2)
Neural networks in 12-lead ECG classification	241	(10)
The artificial neuron	241	(3)
The MLP and ECG classification	244	(7)
Summary	251	(11)
References	254	(8)
Neural network predictions of significant coronary artery stenosis in women
B.A. Mobley
W.E. Moore
E. Schechter
J.E. Eichner
P.A. McKee
Introduction	262	(3)
Systems enabling the avoidance of unnecessary angiography	262	(1)
Women and angiography	263	(1)
Other clinical predictions by neural network	264	(1)
Methods	265	(11)
Development of the data set from the SCA&I database	265	(5)
Artificial neural network	270	(2)
Patient files	272	(2)
Logistic regression	274	(1)
ROC analysis	275	(1)
Results	276	(9)
Neural network training and cross validation	276	(1)
Network application to the cutoff determination file	276	(2)
Network application to the test file	278	(3)
Relative weights of the neural network	281	(1)
Logistic regression	282	(3)
Discussion	285	(3)
Patients and data	285	(1)
Patient files	286	(1)
Cutoff determination file	287	(1)
Predictive systems	287	(1)
Network weights	287	(1)
Conclusions	288	(4)
Acknowledgments	288	(1)
References	289	(3)
A modalar neural network system for the analysis of nuclei in histopathological sections
C.S. Pattichis
F. Schnorrenberg
C.N. Schizas
M.S. Pattichis
K. Kyriacou
Introduction	292	(2)
The need of quantitative analysis in diagnostic histopathology	292	(1)
A brief overview of the use of artificial neural network - (ANN) systems in diagnostic histopathology	293	(1)
Quantitative analysis in immunocytochemistry	293	(1)
Material	294	(1)
Modular neural network system	295	(11)
Detection of nuclei: the receptive field-squashing function (RFS) module	295	(2)
Step 1: convert color image to optical density image	297	(1)
Step 2: compute the receptive field filter	297	(1)
Step 3: apply iteratively the receptive field and the squashing function	297	(1)
Step 4: threshold bimodal histogram	298	(1)
Step 5: revise the list of detected nuclei	298	(1)
Detection of nuclei: the feedforward neural network (FNN) module	299	(1)
Step 1: color image to optical density image conversion	300	(1)
Step 2: histogram stretching and thresholding	300	(1)
Step 3: SV expansion and feedforward neural network identification of image blocks	301	(1)
Step 4: calculation of the exact nuclei locations	301	(1)
Combination of detection modules	301	(1)
Nuclei classification and diagnostic index calculation	302	(1)
Step 1: extract features for each nucleus	303	(1)
Step 2: classify each nucleus	304	(1)
Step 3: compute diagnostic index	304	(1)
System validation	304	(2)
Results	306	(6)
Detection example	306	(2)
ROC analysis	308	(2)
Classification and diagnostic index computation module	310	(2)
Discussion	312	(2)
Future work	314	(9)
References	315	(6)
Appendix A: Semi-quantitative diagnostic index	321	(2)
Septic shock diagnosis by neural networks and rule based systems
R. Brause
F. Hamker
J. Paetz
Introduction	323	(2)
The data	325	(6)
The data context	326	(1)
Data problems and preprocessing	326	(2)
Selecting feature variables	328	(1)
Basic statistical analysis	329	(2)
The neural network approach to diagnosis	331	(8)
The network	331	(1)
The network architecture	332	(1)
Treatment of missing values	333	(1)
Training and diagnosis	334	(1)
The training and test performance	334	(1)
The problem of medical data partition	335	(1)
Selection and validation of a neural network	336	(2)
Results for septic shock diagnosis	338	(1)
The neuro-fuzzy approach to rule generation	339	(6)
The rule extraction network	340	(3)
Application to septic shock patient data	343	(2)
Conclusions and discussion	345	(12)
Acknowledgments	348	(1)
References	348	(4)
Appendix A: The network adaptation and growing	352	(1)
Adaptation of the layers	352	(1)
Growing of the representation layer	353	(1)
Appendix B: The main rule building algorithm	354	(1)
Appendix C: The rule shrinking procedure	355	(2)
Monitoring depth of anesthesia
J. W. Huang
X.-S. Zhang
R.J. Roy
Introduction	357	(2)
Computational intelligence (CI) for DOA	359	(9)
Fuzzy logic assessment	360	(1)
Fuzzy inference process	361	(2)
Why not fuzzy?	363	(1)
Artificial neural networks	363	(3)
Neuro-fuzzy modeling	366	(2)
ANN-based CI model for MLAEP	368	(7)
MLAEP-derived parameter extraction	369	(1)
Wavelet transformation	370	(2)
System design based on ANN for MLAEP	372	(1)
ANN system: experiment results	373	(2)
Neuro-fuzzy based CI model for EEG	375	(7)
EEG-derived parameter extraction	376	(1)
Complexity analysis	376	(1)
Regularity analysis	376	(1)
Spectral entropy analysis	377	(1)
ANFIS - ``derived'' fuzzy knowledge model	377	(2)
System design based on ANFIS for EEG	379	(1)
ANFIS system: experiment results	380	(2)
Discussions	382	(9)
ANN versus ANFIS	382	(1)
EEG versus MLAEP	383	(1)
Performance issues	384	(1)
Acknowledgments	385	(1)
References	386	(5)
Combining evolutionary and fuzzy techniques in medical diagnosis
C.A. Pena-Reyes
M. Sipper
Introduction	391	(1)
Background	392	(11)
Fuzzy modeling	392	(4)
Evolutionary computation	396	(4)
Evolutionary fuzzy modeling	400	(3)
Fuzzy systems for breast cancer diagnosis	403	(4)
The WBCD problem	403	(2)
Fuzzy-system setup	405	(2)
A fuzzy-genetic approach	407	(4)
The evolutionary setup	407	(1)
Results	408	(3)
A fuzzy coevolutionary approach: fuzzy CoCo	411	(10)
Cooperative coevolution	411	(2)
The coevolutionary algorithm	413	(3)
The evolutionary setup	416	(3)
Results	419	(2)
Concluding remarks	421	(6)
References	422	(5)
Genetic algorithms for feature selection in computer-aided diagnosis
B. Sahiner
H.P. Chan
N. Petrick
Introduction	427	(2)
Genetic algorithms	429	(3)
Encoding	430	(1)
Initial population	431	(1)
Fitness function	431	(1)
Genetic operators	431	(1)
Working parameters	431	(1)
Feature selection and Gas	432	(2)
Applications in CAD	434	(41)
Classification of malignant and benign microcalcifications	436	(1)
Feature extraction	437	(1)
Data set	437	(1)
Morphological feature space	438	(2)
Texture feature space	440	(2)
GA implementation	442	(2)
Classification	444	(1)
Results	445	(7)
Discussion	452	(1)
Classification of mass and normal breast tissue	453	(1)
Data set	453	(2)
Morphological features	455	(1)
Texture features	455	(1)
Classification	455	(1)
GA implementation	456	(1)
Results	457	(4)
Discussion	461	(2)
Classification of malignant and benign masses	463	(1)
Data set	464	(1)
Image transformation	464	(2)
Texture features	466	(1)
Classification	467	(1)
GA implementation	467	(2)
Results	469	(5)
Discussion	474	(1)
Conclusion	475	(10)
References	476	(9)
Index	485	(4)
List of contributors	489