Design of a Fuzzy Input Expert System Visual Information Interface for Classification of Apnea and Hypopnea

Abstract

In this study, a Fuzzy input Expert System (FES) is developed to detect the patients' PSG results to linguistic statements (apnea-hypopnea). All of the randomly selected 1318 PSG data taken from 15 patients (12 males (80%) and three females (20%)) from St. Vincent University Hospital / University College Dublin Sleep Apnea Database were studied and applied to the FES model. It is understood from the literature that three signals (airflow, SpO(2), and Rib movements) are the primary indicators of apnea and hypopnea. Thus, this study's three important parameters were chosen as input variables to classify the apnea-hypopnea in this study. The output variable DIS (disease) was defined as A (Apnea) and H (Hypopnea). A rule base (consisting of 75 rules) was created using membership functions in the light of AASM's 2012 scoring criteria and an expert's opinion. Since it is the most preferred method, this study uses the center-of-gravity/area (centroid) method for defuzzification. The limit values for each fuzzy expression were created. These parameters were symbolically classified. Membership functions and the degree of the membership function were defined. 231 apnea and 1029 hypopnea events have been successfully detected at 97.5% and 95.2%, respectively. A confusion matrix has been formed for calculating the performances of FES, and accuracy was found to be 97.5%. An interface program was developed using Matlab Graphical User Interface programming language, where some sample results were checked. Thus, results have been converted into understandable linguistic expressions. It can be said that the detection performance of the system developed is good by looking at the results of the correct detection. It is shown that detecting apnea and hypopnea using FES are reliable, consistent, and successful results and helps doctors make quick and reliable diagnoses without any risks.