Application of the Fuzzy Min-Max neural networks to medical diagnosis

In this paper, the Fuzzy Min-Max (FMM) neural network along with two modified FMM models are used for tackling medical diagnostic problems. The original FMM network establishes hyperboxes with fuzzy sets in its structure for classifying input patterns into different Output categories. While the first modified FMM model uses the membership function and the Euclidian distance to classify the input patterns, the second modified FMM model employs only the Euclidian distance for the same process. Unlike the original FMM network, the two modified FMM models undergo a pruning process, after network training, to remove hyperboxes with low confidence factors. To assess the effectiveness of the three FMM networks in medical diagnosis, a set of real medical records from suspected Acute Coronary Syndrome (ACS) patients is collected and used for experimentation. The bootstrap method is used to analyze the results statistically. Implications of the experimental outcomes are discussed, and the potential of using the FMM networks a decision support tool for medical prognostic and diagnostic problems is demonstrated.