الفهرس 
Introduction
Data acquisitionOnly 14 pages are availabe for public view
 |  | from | 218 |  |  |
| | | from | 218 | | |
Abstract
Intelligent systems are increasingly being deployed in medicine and health care to practically aid the busy clinician and to improve the quality of patient care. In the present thesis, an intelligent system is designed to help clinical personnel to interpret esophageal motility assessment tests: The lower Esophageal Sphincter (LES) and the tubular esophagus body. The suggested intelligent system combines artificial intelligence methodologies with biomedical signal acquisition, processing, analysis, classification and evaluation providing intelligent processing of esophageal motility data. It can help in monitoring, in measurement supervision, and in making diagnoses of esophageal motility records (manometric traces). A group of 77 cases representing different abnormalities associated with dysphagia diseases, were used in training and testing the developed systems. Computer programs were developed for processing and feature extraction of parameters to be used in the classification procedure. Two intelligent techniques were developed, for automatic diagnosis. The first technique was based on Artificial Neural Networks and the second utilizes Expert System methodology. A neural network classification technique has been utilized using a three-layer feed-forward neural network. The network was trained using back propagation algorithm with adaptive learning rate and z-score weighting normalization. The classification accuracy has been estimated for the two kinds of studies; LES and Esophagus Body studies. Training was performed, using the leave-one-out technique for classes having cases less than ten and the hold-out technique for larger class sizes. The classification results obtained from the neural network classifier reach 97.4 to 100% correct classification, for the Esophageal Body studies and LES respectively. A combination of both studies of LES and Esophageal Body has been done to get an overall diagnosis for the whole trace of a specific patient. The overall percentage classification reaches 100% correct classification. The higher classification rate is due to the fact that cases, which have been misclassified in the Esophageal Body study, are correctly classified when using information extracted from both the LES and Esophageal Body studies. An Expert System has been developed which span a fair range of esophageal motility human expertise, specifically for detecting deseases related to the dysphagia affecting the esophagus. Written in Prolog, the expert is designed using backward chaining method for inferring the knowledge base. The knowledge base was built, using built-in database features in Prolog language, containing Objects as different classes of normality and abnormalities and Attributes as descriptive paramters associated with this Object/Class. An explanatory interface was designed for reasoning and explaining the procedure of inferring unknown object and the final diagnostic-decision. The Expert System correctly classified 94.8% (eight-category classification) of 77 cases. This demonstrates that expert systems can be a feasible approach in building more robust esophageal monitoring systems. The results show a substantial level of agreement between the developed system and the physician. Interpretation of the results indicates that gastroentrology field could be an area of interest for the application of Artificial Intelligence techniques.v.