Scimago Lab
powered by Scopus
call: +1.631.470.9640
Mon-Fri 10 am - 2 pm EST


eISSN: 1643-3750

Computer educational program for diagnosis of liver diseases based on HEPAR II

H. Wasyluk, A. Oniśko

Med Sci Monit 2003; 9(2): 18-19

ID: 15297

Published: 2003-05-20

Background:Computer-based support of diagnosis of liver diseases has been a subject of numerous investigations (e.g,Adlassaning and Horak,1995,Bobrowski 1992,Lucas 1994).We have built a Bayesian network model called HEPAR II,that employs techniques of decision analysis. The model aims at supporting clinical diagnostic process of liver diseases (OniEko et al,2000, 2001).The HEPAR II system is a continuation of a work on a single-disorder diagnosis model,HEPAR I,created at the Institute of Biocybernetics and Biomedical Engineering PAN in cooperation with physicians of the Medical Center of Postgraduate Education in Warsaw (Bobrowski,1992,Wasyluk, 1995,2002). The computer system we present is based on the idea of gathering and processing updated clinical data of clinical patients with liver diseases. This database comes from the Department of Gastroenterology at the Institute of Food and Feeding in Warsaw.The database was founded in 1990;each hepatological case is described by circa 160 various medical variables and by clinical and histopathological diagnoses. Material/Methods:The structure of the HEPAR II model was built on the basis of both medical literature and knowledge of the experts in hepatology.The version of HEPAR II model we used in our investigations has been constructed on a model of Bayesian network and consisted of 73 nodes, representing 11 different liver diseases,66 variables reflecting results of clinical signs and symptoms as well as diagnostic laboratory tests .Numeric parameters of the model have been specified based on the HEPAR database. In the initial phase of developing the final program, a prototype of the educational program for diagnosis of liver disorders has been prepared. It consists of a few windows which allow for an interaction with the HEPAR II model. First three windows give an opportunity to introduce data of a diagnosed patient. An interactive diagnostic interface to HEPAR II consists of a number of windows,each connected with different group of clinical findings of a patient. On the left side of each window (vide Fig.1)is displayed a list of findings. Right-clicking on any of the features brings up a pop-up menu that lists all possible values of the selected variable. The right side of the window presents an ordered list of possible diagnoses with their associated probabilities, presented also graphically. The probabilities are updated immediately after entering each finding and a newly ordered list of possible disorders appears in a second. Thus, one can verify instantly whether introducing a new feature into the model will change probabilities of particular diseases.Results:The prototype of the educational program based on the HEPAR II model has been presented to the groups of physicians, participating in courses of medical informatics available for those specializing in family medicine. The presentation of the program aimed at acquainting them with possible diagnostic strategies of the model through diagnostic simulations .Applying this method physicians could assess and verify diagnostic value of signs symptoms and tests resulting from diagnostic examinations of a particular patient .Such a simulation enabled them choosing of optimal diagnostic strategy in determining type of medical tests to be performed in particular clinical situation. Despite of medical and ethical aspects of diagnostic procedures, such a computer assisted identification of a laboratory test of the highest diagnostic power for a particular patient evidently diminishes costs of diagnostic process and accelerate applications of specific therapy. In frames of the process of initial testing of the prototype of our educational program ,we have performed experimental investigation, in which the diagnostic accuracy of the HEPAR II model was compared with diagnostic accuracy of a group of family doctors .It proved, that the diagnostic accuracy of the model (70%)was higher than the accuracy of physicians (33%). Moreover, these physicians, using this computer assisted diagnostic system, Increased their diagnostic accuracy from 33%to 65%. Above results suggest that presented system would be useful as an educational tool for students and specializing physicians in supporting medical diagnosis. Presented system is still under development and new ways of communication of physicians with computer assisted diagnostic system are currently verified successively by physicians attending courses provided by the Department of Informatics of the Medical Center for Postgraduate Education in Warsaw