Project 13: Gait malfunction identification and classification Organization: Vienna University of Technology, Project Abstract Name: Institute of Software Technology Primary Contact: Dieter Merkl Address: Institute of Software Technology Vienna University of Technology Resselgasse 3/188 A-1040 Vienna Austria Email: dieter@ifs.tuwien.ac.at URL: http://www.ifs.tuwien.ac.at/ifs/people/dieter Based on data recorded with ground reaction force measurement platforms we are working towards a tool for human gait malfunction identification. The data is recorded with patients during their stay at an Austrian rehabilitation centre after injuries to their locomotive system. Issues of interest are the representation of ground reaction force data for subsequent identification of gait malfunctions and tools for the identification and classification of gait malfunctions. So far we were primarily interested in evaluating different architectures of artificial neural networks to perform that task. Duration: 6 years Number of People: 5 Tools Developed: Supervised/Unsupervised Neural Networks Academic Disciplines: Computer Science, Medicine Project Related Publications: Keywords: Gait Analysis, Pattern Recognition, Classification, Learning, Neural Networks "Identification of Gait Patterns with Self-Organizing Maps Based on Ground Reaction Force", European Symposium on Artificial Neural Networks, Bruges, Belgium, April 24-26, 1996. "Clinical Gait Analysis: Issues and Experiences", 1EEE S~vmposium on Computer-BasedMedical Systems, Maribor, Slovenia, June 11-13, 1997. Project 14: Data mining Organization: Duke University ofprenatal data Abstract Comparing inductive machine learning with neural nets and traditional statistical techniques for 'best' predictors of women at risk for PRETERM BIRTH. Preterm babies having higher mortality and morbidity outcomes and often cost in excess of $1 million in the first year of life. We're mining 10 years of prenatal data for possible patterns of preterm birth risk. Primary Contact: Linda Goodwin, RN, and Ph.D. Address: Box 3322 Duke University Durham, NC 27710 Email: linda, goodwin@duke.edu URL: http://www.duke.edu/~goodw010/collaboratory/OB team.ht ml Duration: 6 years Number of People: 10 Tools: Developed: LERS; data extract/cleaning programs Purchased: PREDICT neural net software, SPSS, SAS Academic Disciplines: Nursing, BME, CS, Statistics, and Medicine Keywords: induction, neural networks, machine learning, classification, preterm birth Funding Sources: National Library of Medicine Project Related Publications: Woolery (now Goodwin), L, & Grzymala-Busse, J., "Machine Learning and Preterm Birth Risk Assessment", Journal of the American Medical Informatics Association, 1, 6, 439-446, 1994. Goodwin, L. Prather, J. Schlitz, K. Iannacchione, MA. Hage, M. Hammond, E. Grzymala-Busse, J., "Data Mining Issues for Improved Birth Outcomes", In Barret, S. and Wright, C. (Eds) Biomedical Sciences Instrumentation, 34, 291-296,1998.

/lp/association-for-computing-machinery/data-mining-of-prenatal-data-zHjiPbftAX