Intelligent Medical Diagnosis and Decision Support Model Based on Neural Networks and Rule Based System رسالة ماجستير

Abstract

The human body is a vital source of data, which is important for human health. Medical imag ing is one of the processes that produce different kind s of human body data. While some data in the form of images, the others are signals. This data daily used to d iagnose different kinds of diseases. There are many different bio signals that can be collected; some important signals are Electrocardiogram (ECG), Electroencephalogram (EEG) and Electromyogram (EMG). These signals are collected from different human organ s and utilized in diagnos ed different diseases. The designing and implementation of intelligent computer programs that try to emulate with human intelligence are a sign of the integration of various sciences and areas of knowledge. One important field is the improvement that allows appropriate assistance to physicians in decision making. The development of technologies associated with Artificial Intelligence (AI) techniques that are applied on medicine, represents a novel perspective, which can reduce co sts, time, and medical errors. The integration between artificial intelligence and the medical system is vital, and a lot of efforts have been made in this area. While this field still needs more and more investigation , intelligent medical and diagnostic decision support systems could consume these amounts of data, and utilize it to improve healthcare. The using of Artificial Intelligence methods in medical diagnosis can Benchmark from several of its main techniques such as expert systems (diagnosis based on rules, probabilities), fuzzy logic (diagnosis based on classification), neural networks (diagnosis based on training and recognition), applied data mining (diagnosis through the pattern recognition). VI In this thesis, In this thesis, a new metha new method od are are producedproduced to support the take of medical decisions by combining to support the take of medical decisions by combining the intelligent computational systems with medicalthe intelligent computational systems with medical. In general, there is an approximate shared . In general, there is an approximate shared procedure follow to manipulate with these problems;procedure follow to manipulate with these problems; starts with destarts with de--noising the signals, noising the signals, andand thenthen applies feature extraction methods (reduction and selection). applies feature extraction methods (reduction and selection). While the last task is to classify or While the last task is to classify or recognize a different pattern used in medical diagnoses to make a decision in determined medical recognize a different pattern used in medical diagnoses to make a decision in determined medical cases.cases. Therefore, this study proposed a technique thaTherefore, this study proposed a technique that using useful physiological variables for t using useful physiological variables for diagnosis heart disease, diagnosis heart disease, a hybrid system that combined expert systems and neural networks for a hybrid system that combined expert systems and neural networks for the implementation of Intelligent Medical Diagnosis System in Decision the implementation of Intelligent Medical Diagnosis System in Decision Support in medical Support in medical application is application is used. used. AnotherAnother goal goal that that isis achieved was the use of optimization of neural network achieved was the use of optimization of neural network parameters by optimization algorithms with the objective of parameters by optimization algorithms with the objective of enhancienhancing ng thethe system. Asystem. A hybrid hybrid system that combines Genetic Algorithm (GAssystem that combines Genetic Algorithm (GAs),), BiogeographyBiogeography--Based Optimization (BBO) with Based Optimization (BBO) with neural networks (NNs) [GAsBBOneural networks (NNs) [GAsBBO--MLPNNs]], BBO and particle swarm optimization (PSO) , BBO and particle swarm optimization (PSO) methods combined with the neural network was used to improve the performance of the systems. methods combined with the neural network was used to improve the performance of the systems. The idea of this thesis concentrates oThe idea of this thesis concentrates on the medical diagnosis system for heart disease using n the medical diagnosis system for heart disease using aartirtificial intelligence techniques. ficial intelligence techniques. The proposed method produces better performance than The proposed method produces better performance than previous works, where previous works, where the GAsBBOthe GAsBBO--MLPNNs method performance parameters result MLPNNs method performance parameters result represented as 9represented as 94.5% 95.6%, 89.94.5% 95.6%, 89.94% % accuracyaccuracy, G, G--mean, and Fmean, and F--measure respectively.measure respectively. TheThe Intelligent Medical Diagnosis Intelligent Medical Diagnosis System has achievedSystem has achieved a a prediction prediction accuracy of 95.097% using accuracy of 95.097% using NNeuroeuro--FFuzzy modeluzzy model with triangular membership function.with triangular membership function.