Decision support in acute abdominal pain
Abdominal pain is one of the commonest symtoms in acute surgical wards. A quick correct diagnosis and treatment is not seldom lifesaving, but the diagnosis is in many cases difficult, especially for unexperienced doctors. Several attempts have therefore been made before to build decision support systems for acute abdominal pain, but with only limited success. Based on a large material from a standardized clinical examination (cf. Laurell, 2006), we intend to develop a new kind of decision support using the theory of Bayesian belief nets (BBNs; cf. Husmeier et al 2005). In contrast to earlier so-called "naive Bayesian" approaches, BBNs take account of dependencies between variables when calculating the probabilities for the different diagnoses. We also plan to develop an alternative system from the same data but instead based on artificial neural networks (Dybowski & Gant 2001). Recently, we have successfully applied SVMs (support vector machines) to a sub-problem (Nalin et al 2008, Åberg et al 2009, Björnsdotter Åberg et al 2009).
A masters thesis from our group (Nalin, 2006) has adressed the actual work situation in an acute surgical ward in order to find out what need there is for such improved decision making procedures. Nalin is now continuing these investigations as a doctoral student in cognitive science. Our future plans include collecting new data from surgical wards in Western Sweden and constructing a teaching program for medical students using actual cases from the database.
The project leader is Lars-Erik Hansson, associate professor of surgery at the Sahlgrenska Academy. Major project partners are Richard Dybowski (InferSpace) and Helge Malmgren, Department of Philosophy, GU. The project has hitherto mainly been funded by Region Västra Götaland.
References:
L.-E. Hansson, Akut buk. Studentlitteratur, Lund 2002.
R. Dybowski & V. Gant (eds)., Clinical applications of artificial neural networks. London 2001
D. Husmeier, R. Dybowski & S. Roberts (Eds.), Probabilistic Modeling in Bioinformatics and Medical Informatics. Springer 2005.
H. Laurell, Acute Abdominal Pain. Thesis, Uppsala University 2006.
K. Nalin, Den ideala kliniska beslutsprocessen. En studie av arbetsprocessen på en kirurgisk akutmottagning (The ideal clinical decison process. A study of the work process in an acute surgical ward). Masters thesis in Cognitive Science, University of Gothenburg 2006.
Nalin, K., Malmgren, H., Gunnarsson, U. et al., Automatic computer-based diagnosis in acute abdominal pain. Poster presented at Medicinteknikdagarna, Gothenburg, Sweden, Oct. 14-15, 2008. Philosophical Communications, Web Series 50. University of Gothenburg.
Åberg, M., Nalin, K., Hansson, L.-E. & H. Malmgren, Towards an automatic diagnosis system for acute abdominal pain. To be published in the proceedings from Biosignals 2009 (Porto, Portugal).
Björnsdotter Åberg, M., Nalin, K., Hansson, L.-E. & Malmgren, H. (2009), Support vector machine diagnosis of acute abdominal pain. Communications in Computer and Information Science (Springer), in print.
Back to main (Swedish) ANN page
Back to
main
(English) ANN page
Last update: March 8, 2009 by Helge Malmgren