Acute Myocardial Infarction (AMI) is one of the most critical diseases for its highincidence of disease and death rate. Because of its diversity of clinical symptoms anddifficulty of diagnosis, misdiagnosis is not uncommon, which leads to inappropriatetreatment and waste of money. As a result of that, it's very important to find aneconomical and effective laboratory examination method in order to diagnose timelyand correctly.To fulfill all the above criteria, POPFNN (a Pseudo Outer-product Based FuzzyNeural Network), a new kind of Fuzzy Neural Network, and the application of GeneticAlgorithms into automatically construct Member Function, are raised in this thesis. Onthe basis of the advantage of Enzymes and Biochemical Markers representativeMyocardial Injury, it can improve the analysis accuracy and increase the diagnosis rateof AMI while making judgment of period of AMI as well, which can reduce both themisdiagnosis rate and the economic burden of patients.In this project, the GA is capable of automatically constructing membershipfunctions from the input training data. It utilizes the powerful search capability ofgenetic algorithm to generate the best membership functions for the input training datafrom a pool of population. Guided by a fitness evaluation function, provided withsufficient running time, the system will converge to derive an optimal or near-optimalgroup of membership functions. The rule base of the fuzzy system is derived by usingthe POP Rule Leaming Algorithm to automatically construct fuzzy rules. During thisprocess, both fuzzy membership functions and fuzzy rules will become more optimaland it requires very few complicated manually tuned preset parameters. This makes thedesign of the system fast and cost-effective.The proposed Diagnosis System can be divided into three main modules: 1) aSolution Generation Module; 2) a Fitness Evaluation and Assignment Module; 3) aSolution Implementation Module. The first and second modules are represented by a GAs, while the last module is a FRBS. The Solution Generation Module consists of theGAs different level populations and the adopted genetic operations. The chromosomesin these populations keep evolving new possible solutions. Then these possiblesolutions are decoded from GA chromosomes and implemented into fuzzy rule basesystems in the Solution Implementation Module. Training data are presented to thefuzzy rule base systems and the results are used for fitness evaluation in the FitnessEvaluation and Assignment Module, which assigns proper fitness value to eachchromosome. The Solution Generation Module evolves the genetic populations basedon these fitness values. As the evolution stops, the learning process terminates.In this thesis, domestic and oversea similar investigations are introduced, which isfollowed by the relative knowledge of medicine and theory of POPFNN and GA. Thenthe theory of the Diagnosis System and design flow of POPFNN are illustrated. At last,GA automatically construct Member Function and Fuzzy Rule Base generated by POPRule Learning Algorithm are attained, through which diagnosis of AMI is achieved. |