Simple Search:

Bayesian network modeling of gastrointestinal bleeding


Citation

Aisha, Nazziwa (2013) Bayesian network modeling of gastrointestinal bleeding. Masters thesis, Universiti Putra Malaysia.

Abstract / Synopsis

Acute gastrointestinal bleeding (GIB) is a common medical emergency with 50-150 per 100,000 people admitted per year. Although 80 percent of GIB cases stop spontaneously, it is important to determine the source of bleeding and establish adiagnosis such that possible recurrences are prevented and that the most suitable management may be given in future episodes. In the emergency room, when a patient shows signs of hematemesis (vomiting of red blood), it is obvious that the patient has upper gastrointestinal bleeding. In the absence of hematemesis however,the source of bleeding remains unclear. While the diagnosis of GIB is best done by a gastroenterologist, it is not always feasible, due to scarcity of resources and time. A reliable classification model would be very helpful in diagnosing patients more efficiently and effectively targeting the scarce resources.Current review of the literature, did not reveal any model that predicts the source of GIB in the absence of hematemesis. This thesis uses a graphical modeling approach,specialcally Bayesian networks, to model the different outcomes of GIB. One key advantage of Bayesian network models in this context is their ability to predict the outcome with partial observations on variables or attributes. The four outcome variables predicted are: source of bleeding, need for urgent blood resuscitation,need for urgent endoscopy, and disposition. Performance of the models is assessed by classification or prediction accuracy, area under curves, sensitivity and specificity values. The Bayesian network models provide good accuracy for the prediction of the source of bleeding and need for urgent blood resuscitation but did not do well on predicting need for urgent endoscopy, and disposition. The models require further validation if they are to be used in clinical settings.


Download File

[img]
Preview
PDF
FS 2013 73 IR.pdf

Download (1MB) | Preview

Additional Metadata

Item Type: Thesis (Masters)
Subject: Mathematical statistics
Subject: Bayesian statistical decision theory
Call Number: FS 2013 73
Chairman Supervisor: Mohd Bakri Adam, PhD
Divisions: Faculty of Science
Depositing User: Mas Norain Hashim
Date Deposited: 14 Mar 2019 09:50
Last Modified: 14 Mar 2019 09:50
URI: http://psasir.upm.edu.my/id/eprint/67578
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item