5. Tool environments and development methods (TOOLMET)

Abstract

The aim of the NESUMED project was to analyse neural networks and non-linear optimisation methods in certain medical applications. The applications were selected to meet the needs of the medical device industry.

In the medical signal analysis, a PC-based program for automated analysis of NMR spectrum series was developed. ³¹P NMR spectroscopy enables us to study intracellular energy metabolism noninvasively. The spectra contain NMR lines corresponding to certain metabolites in the sample. The relative concentrations of these metabolites change over time during exercise. By measuring an NMR spectrum every five seconds we obtained a time series of NMR spectra that could be analysed. Earlier analysis was carried out by a human expert who identified the peaks manually, which was too laborious a method for clinical use. Spectrum analysis leads to a non-linear optimisation problem which can be solved with iterative search methods. The Levenberg-Marquardt method was used, which is a compromise between the steepest descent method and the Taylor series method.

In another application we used multilayer perceptrons in the analysis of biochemical data measured to screen for Down’s syndrome in early pregnancy. In laboratory medicine there have been some promising publications on the usefulness of neural networks in interpreting laboratory test results. However, the number of example cases in most diseases is so small that the use of neural networks cannot be recommended. Laboratory tests used for screening are exceptions. Since the screening tests are taken on a large group of patients the number of examples cases will be sufficient. Furthermore, often thousands of patients are screened for a certain disease. This makes it critically important to have accurate models to conclude whether the patient needs further examination or not. A small inaccuracy in the calculations may lead to a large number of unnecessary tests. Differences in patient populations in different countries and even within one country have shown that there is a need to adapt laboratory methods and reference values to each population separately. In some multivariate calculations this cannot be done easily. Theoretically, there were two arguments supporting the use of neural networks in screening applications: 1) with a large set of examples there was a possibility to increase the accuracy of calculations based on screening data, and 2) neural networks have adaptive nature, which enables us to adapt the model easily to a new population.

Results

The NMR spectroscopy during exercise is a novel method which seems to be promising in pharmaceutical research and in diagnosing different diseases. However, the clinical usefulness of the method is as yet unclear and studies with different patient materials will be carried out in the near future to find out the best application areas. In this clinical research the program is in daily use.

In the screening of Down’s syndrome a large data set was collected from several central hospitals in Finland. When Down’s syndrome is screened using biochemical markers the number of screened cases is large but the number of affected pregnancies is very small. Therefore, neural networks are not feasible in detecting affected pregnancies. However, many patient-specific factors influence the levels of the biochemical markers (alphafetoprotein AFP and human chorion gonadotropin hCG) and these effects can be learned from data obtained from normal pregnancies. Furthermore, the effects are most likely non-linear and thus hard to model using conventional statistical techniques. In our preliminary results we have been able to develop a neural network model that can produce patient-specific reference values for both AFP and hCG from certain input data. As input variables we used factors that are known to influence the marker levels such as maternal weight, gestational age and maternal age. In a pilot study where we compared the reference value obtained from a neural network and the reference value based on conventional statistics the difference was not significant.

In another experiment we studied the transferability of a neural network model to another population. In this study it was shown that a neural network model that is developed in another population can be used as a basic model for another population. However, the basic model has to be fine-tuned to the new population by teaching it with a set of sample of the new population. When the learning factors are kept small the basic model is not destroyed but it is only adapted to perform better in the new population.

Based on the result of the Down’s syndrome screening example it is concluded that neural networks do not lead to significant improvements on the mathematical model itself. However, neural networks are more easily adapted to another population, which is an important feature in laboratory medicine.

Project information

Participants

• Turku University Central Hospital
• Department of Computer Science, University of Turku
• Wallac Oy
• Leiras Oy
• Mylab Oy
• Diter-Elektroniikka Oy

Project dates

Project volume

Responsible leader

Project manager

More information

5.2 Knowledge integration in the development of fuzzy systems (USTI)

Abstract

Results

Business achievements

The benefits of the systems developed include improved quality, less human work, and higher production. Utilisation remains on the responsibility of the industrial partners. The prototype of the lime kiln controller has lead to implementation of a one-line control system in a separate project.

Engineering achievements

Scientific achievements

Linguistic equation systems can be generalised and tuned to changing operating conditions. Compact representation is beneficial in large scale complex systems. The lime kiln application resulted a new type of intelligent controller and several simulation models that can be used in control design. Linguistic equations were introduced into soft sensors in the modelling of continuous cooking. The edge detection algorithm includes an online adaptation system for the membership functions. Nonlinear dynamics has been introduced into feature selection in forecasting component consumption in electronics manufacturing.

Project information

Participants

• UPM Kymmene, Wisaforest
• ABB Industry, Pulp and Paper
• Valmet Oyj, Paper Machines
• Nokia Access Systems Oy

• University of Oulu, Control Engineering Laboratory

Project dates

Project volume

Project manager

Publications

Tool and methodology development

Juuso E. K.: Linguistic Equations in Developing Fuzzy Systems. In: Closed Working Group Meeting of FALCON (Fuzzy Algorithms for Control), Mallorca, April 26-29, 1995, 10 pp.

Juuso E. K.: Adaptive Interfaces based on Soft Computing. In: Improving the Modelling and Simulation Process - Second workshop of the SiE Working Group, Brussels, June 29-30, 1995 (G. Vansteenkiste & H. Vangheluwe, ed.), 10 pp.

Juuso E. K.: Adaptive Interfaces in Synthetic Environments. In: Improving the Modelling and Simulation Process - Second Workshop of the SiE Working Group, Brussels, June 29-30, 1995 (G. Vansteenkiste & H. Vangheluwe, ed.), 8 pp.

Juuso E. K.: Symbolic Computation in Industrial Applications, In: Eurosim '95 - Poster Book, Argesim report No. 3 (F. Breitenecker & I. Husinsky, ed.).

Juuso E. K.: Linguistic Equations in System Development for Computational Intelligence. Proceedings of the Fourth European Congress on Intelligent Techniques and Soft Computing -EUFIT'96, Aachen, September 2 - 5, 1996 (H.-J. Zimmermannn, ed.), Aachen, 1996. Verlag Mainz. 1996, Volume 2, pp. 1127-1131.

Juuso E. K.: Computational Intelligence in Distributed Interactive Synthetic Environments. Simulation in Industry, Proceedings of the 8th European Simulation Symposium, Genoa, Italy, October 24-26, 1996 (Agostino G. Bruzzone & Eugene J. H. Kerckhoffs, eds.). Volume 1, 1996, pp. 157-162.

Juuso E. K.: Intelligent Methods in Diagnostical Process Analysis. Proceedings of XIV IMEKO World Congress, New Measurements - Challenges and Visions, Tampere 1-6 June 1997 (Jouko Halttunen, ed.), Volume VII, pp. 1-6.

Juuso E. K., Leiviskä K.: A Development Environment for Industrial Applications of Fuzzy Control. In  Proceedings of the Third European Congress on Intelligent Technoiques and Soft Computing -EUFIT'95, Aachen, August 28 - 31, 1995 (H.-J. Zimmermannn, ed.), volume 2, pp. 796-803, Aachen, 1995. Verlag und Druck Mainz.

Juuso E. K., Leiviskä K.: Adaptive Interfaces and Soft Computing. Proceedings of the Symposium on Modelling, Analysis and Simulation -CESA’96, Lille, France, July 9-12, 1996, IMACS, 1996, Volume 1, pp. 414-419.

Juuso E. K., Leiviskä K.: Lingvistisiin yhtälöihin perustuva sumeiden järjestelmien kehitys- ja viritysympäristö. Automaatio 97, Helsinki 23.-25.9.1997, Suomen Automaatioseura, CD-ROM, Esitelmä n:o 83.

uuso E., Yliniemi L., Leiviskä K.: Oppivat ja älykkäät järjestelmät - tutkimusta. Automaatioväylä (1996):6, pp. 34-37.

Lotvonen S., Kivikunnas S., Juuso E.: Tuning of a Fuzzy System with Genetic Algorithms and Linguistic equations. Proceedings of the 5th European Congress on Intelligent Techniques and Soft Computing -EUFIT'97, Aachen, September 8 - 11, 1997 (H.-J. Zimmermannn, ed.), Aachen, 1997. Verlag Mainz. 1997, Volume 2, pp. 1289-1293.

Myllyneva J., Juuso E. K.: FuzzyCon and FuzzyTune: control and tuning in PC. In: Closed Working Group Meeting of FALCON (Fuzzy Algorithms for Control), Mallorca, April 26-29, 1995, 6 pp.

Applications

Lime kiln control

Juuso E. K., Adaptiivinen sumea säätö, Annual Seminar of the National Technology Programme Adaptive and Intelligent Systems Applications, Helsinki, April 22, 1998, Tekes, Helsinki, 1998, 16 pp.

Juuso E. K., Intelligent Dynamic Simulation of a Lime Kiln with Linguistic Equations, Proceedings of the European Simulation Multiconference -ESM'98, Manchester, UK, June 16-19, 1998 (R. Zobel & D. Moeller, eds.), Volume 2, SCS, Delft, The Netherlands, 1998, pp. 23-28.

Juuso E. K., Ahola T., Leiviskä K.: Fuzzy Logic in Lime Kiln Control. Proceedings of TOOLMET’96 - Tool Environments and Development Methods for Intelligent Systems, Oulu, April 1-2, 1996 (Leena Yliniemi & Esko Juuso, eds.). Report A No 4, May 1996, pp. 111-119.

Juuso E. K., Ahola T., Leiviskä K.: Fuzzy Modelling of a Rotary Lime Kiln Using Linguistic Equations and Neuro-Fuzzy Methods. Proceedings of the 3^rd IFAC Symposium on Intelligent Components and Instruments for Control Applications - SICICA'97, Annecy, France, June 9-11, 1997 (L. Foulloy, ed.). pp.. 579-584.

Juuso E., Ahola, T., Oksanen P., Leiviskä K.: Application of Fuzzy Logic in Lime Kiln Control. Preprints of Control Systems ’98, Information Tools to Match the Evolving Operator Role, Sept. 1-3, 1998, Porvoo, Finland (R. Ritala, ed.), pp. 54-62.

Juuso E. K., Leiviskä K.: Modelling of Industrial Processes Using Linguistic Equations: Lime Kiln as an Example. Proceedings of the Fourth European Congress on Intelligent Techniques and Soft Computing -EUFIT'96, Aachen, September 2 - 5, 1996 (H.-J. Zimmermannn, ed.), Aachen, 1996. Verlag Mainz. 1996, Volume 3, pp. 1919-1923.

Juuso E. K., Leiviskä K.: Adaptive Interfaces and Soft Computing. Studies in Informatics and Control, Volume 6, No. 2, June 1997, pp. 117-124.

Digital image processing for recovery boiler control

Murtovaara S., Juuso E., Sutinen R.: Fuzzy Logic Detection Algorithm. Proceedings of IWISP’96 - Third International Workshop on Image and Signal Processing on the Theme of Advances in Computational Intelligence, 4-7 November 1996, Manchester, UK (B. G. Mertzios & P. Liatsis, eds.), 1996, pp. 423-426.

Web break indicator

Functional testing

Komulainen K., Juuso E., Leiviskä K.: Vikatietojen hyödyntäminen toiminnallisessa testauksessa. Automaatio 97, Helsinki 23.-25.9.1997, Suomen Automaatioseura, CD-ROM, Esitelmä n:o 82. 6 s.

Komulainen K., Juuso E. K., Leiviskä K., Kantola J.: Vikatietojen hyödyntäminen toiminnallisessa testauksessa. Automaatioväylä (1997):6, s. 21-23.

Forecasting

Mutka P., Juuso E. K., Leiviskä K., Frantti T.: Uncertainty Modelling in Chaotic Environment. Proceedings of European Symposium on Intelligent Techniques, Bari, Italy 20-21.3.1997, pp. 40-44.

TOOLMET Symposium

Yliniemi L., Juuso E.: Proceedings of TOOLMET’97 - Tool Environments and Development Methods for Intelligent Systems, Oulu, April 17-18, 1997. 201 pp.

5.3 Neural fuzzy systems application design workbench (NFSADW)

Publications

Virtanen, J. Zhou, Computational Intelligence, Lecture Notes, Umeå University, Jan. 1998.

[2] J. Bohlin, P. Eklund, V. Kairisto, K. Pulkki, L.-M. Voipio-Pulkki, A probabilistic network for the diagnosis of acute myocardial infarction, IFSA '97, Prague, June 1997, vol. 4, pp. 44-49.

[3] P. Eklund, A production line for generating clinical decision support systems, EPIA '95, Workshop on Fuzzy Logic and Neural Networks in Engineering, pp. 69-80.

[4] P. Eklund, What is the Role of Logic in Biomedical Engineering?, Multiple-Valued Logic, DAGSTUHL Seminar 9744, COST-MVL Conference, 27-31 October, 1997.

[5] P. Eklund, J. Forsström, Computational intelligence for laboratory information systems, Scand J Clin Lab Invest, {\bf 55} Suppl. 222 (1995), pp. 21-30.

[6] P. Eklund, K. Halin, A. Rajala, Computational intelligence for developing rating systems, EURO-BRUGES, Bruges, March 24-27, 1997.

[7] P. Eklund, L. Kallin, G. Selén, Computational intelligence for medical information analysis and refinement, Applications of Fuzzy & Neuro-Fuzzy Systems in Medicine and Bio-Medical Engineering (Eds. H .N. Teodorescu, A. Kandel, L.C. Jain), CRC Press, to appear.

[8] P. Eklund, T. Riissanen, Break Indicator Software for Paper Machines: Feature Extraction and Data Analysis, Report, Åbo Akademi University, August 8, 1997.

[9] P. Eklund, J. Zhou, Comparison of learning strategies for parameter identification in rule based systems, J. Fuzzy Sets and Systems, to appear.

[10] Q. Y. Hong, Simulation on the control strategy of car driving using fuzzy logic controller, MSc thesis, Åbo Akademi University, 1995.

[11] M. Karv, Regleringen av hissar med fuzzy kontroll, MSc thesis, Åbo Akademi University, 1996. (In Swedish.)

[12] S. Olli, En verktygslåda för utveckling av intelligent reglering, MSc thesis, Åbo Akademi University, 1997. (In Swedish.)

[13] T. Riissanen, Från manuell styrning till automation, MSc thesis, Åbo Akademi University, 1996. (In Swedish.)

[14] T. Riissanen, P. Eklund, Working with a fuzzy control application development workbench: Case study for a water treatment plant, Proc. EUFIT '96, Aachen, Second European Congress on Intelligent Techniques and Soft Computing, September 2-5, pp. 1142-1145.

[15] T. Riissanen, P. Eklund, Soft computing design, development and integration workbench: Joint software development, Proc. EUFIT '97, Aachen, Second European Congress on Intelligent Techniques and Soft Computing, September 8-11, pp. 2491-2494.

[16] G. Selén, DiagaiD: Ett generiskt verktyg för utveckling av beslutsstöd inom medicinsk informatik, MSc thesis, Åbo Akademi University, 1996. (In Swedish.)

[17] J. Zhou, Fuzzy controller tuning, PhLic thesis, Umeå University, 1997.

[18] J. Zhou, P. Eklund, Some remarks on learning strategies for parameter identification in rule based systems, Proc. EUFIT '95, Aachen, Second European Congress on Intelligent Techniques and Soft Computing, August 28-31, 1995, pp. 1911-1916.

[19] J. Zhou, P. Eklund, Parameter sensitivity in tuning fuzzy controllers, IEEE Int. Conf. Systems, Man and Cybernetics, Beijing, October, 1996.

5.4 Self configuration and optimization in fuzzy systems (SOHVI)

Publications

Rauma, T. An Adaptive Control Using Diagnosis Information. In proceedings of TOOLMET'97 - Tool Environments and Development Methods for Intelligent Systems, Oulu, Finland, April 17 - 18, 1997. Pp. 116 - 121. University of Oulu, Control Engineering Laboratory. ISBN 951-42-4648-9

Rauma, T. Diagnosis Information in Meta-Rule Adaptive Fuzzy Systems. In proceedings of the 23rd EUROMICRO Conference: New Frontiers of Information Technology, EUROMICRO'97. Budapest, HU, 1 - 4 Sept. 1997. Pp. 564 - 569. ISBN 0-8186-8129-2

Kurki, M., Rauma, T. & Alahuhta, P. Diagnostic Models in Mechnatronic Systems. Proceedings of the Fifth European Congress on Intelligent Techniques and Soft Computing (EUFIT'97), Aachen, Germany, September 8 - 11,1997. Aachen: Verlag Mainz. Vol III, pp. 2121 - 2125. ISBN 3-89653-200-6

Rauma, T. & Kurki, M. Fuzzy Logic Applications in Diagnosing Mechatronic Systems. To be published in: Reznik, L., Dimitrov, V., and Kapzyk. J. (eds.): Fuzzy Logic Applications in Education and Engineering. Physica Verlag: Lectures on Soft Computing. Heidelberg. (Will be published during the first months of 1998, approximately 300 pages).

Haapea, M. Elektroniikan tuotantolinjan arviointi sumealla logiikalla. 1996, Master of Sience thesis, University of Oulu, Department of Electrical Faculty. 57 pp. (In Finnish.)

Alahuhta P. Sumean mallinnuksen hyväksikäyttö mekatronisen laitteen vikadiagnostiikassa. 1996, Master of Science thesis, University of Oulu, Department of Electrical Faculty. 57 pp. (In Finnish.)

Figure 1. The result window of the CBR-based support tool for maintenance staff.

jukka.iivarinen@hut.fi
http://www.cis.hut.fi/neuronet/Tekes/5.shtml
Tuesday, 28-Nov-2000 15:51:03 EET