Combined Fuzzy-logic and Neural Network Classifiers for Uterine Fibroid in vulnerable women

Abstract

The diagnosis of uterine myoma often referred to as fibroid is poor mostly as women with the ailment present fewer symptoms at the early stage and get less clinical attention. Most times, the tumour is undiagnosed till it gets to an advanced stage, and may also be as a result of the relatively expensive diagnostic equipment used. Over the years the increase in uteri myoma has called for a model that could rapidly recognise the neoplasm at an early stage. In this study, we design a relatively specific model for the detection of uterine myoma. Matrix Laboratory (MATLAB version 7.5.0, R2007b) was used to implement the combined fuzzy logic classifier called adaptive neuro-fuzzy inference system (ANFIS) model. The dataset that was used in training the ANFIS model consist of 90 cases, which involve 78% (70 cases) of the entire dataset that was used in the training process of the system. The remaining 22% (20 cases) of the data were also used in the testing process of the system, which the system will get adapted to for subsequent recognition of similar cases. The ANFIS program was designed using a bell membership function that utilises a hybrid optimisation method with an error tolerance of 0.05. The training dataset was passed through the ANFIS for 30 epochs. At the end of the simulation, the system had a training error of 0.016169 with the training dataset and an average testing error of 0.010315. The system was able to classify approximately 99% of the test data set accurately. Therefore, we have successfully classified the symptoms of uterine fibroid using combined fuzzy-logic and neural network classifiers for vulnerable women.

Keywords: Artificial intelligence, ANFIS classifier, Fuzzy-logic, Fibroid & Uterine myoma.