Pre-classification process symptom questionnaire based on fuzzy logic for pulmonary function test cost reduction

Abstract

In the past few years, developing of computeraided systems for disease classification has been investigated more extensively. Medical professionals use these systems as assistance in diagnosis since they perform the diagnosis based on larger, more complex set of new and previously stored information. Those computer-aided systems are equipped with graphical user interface that makes application in everyday situations more convenient. Disease classification in most computer-aided systems is based on expert systems. Beside the Artificial Neural Networks (ANNs), fuzzy logic (FL) or some other tools are often used for this purpose. This study presents the results of disease pre-classification process and determining the need for conducting respiratory function tests such as spirometry (SPIR), Impulse Oscillometry (IOS), or Body plethysmography and running the Fuzzy Logic – Artificial Neural Network (FL-ANN) Expert System for classification of respiratory diseases. This pre-classification algorithm optimizes time resources as well as reduces the costs of medical device use needed for testing of patient and costs of medical professional attending the measurement. Questions and symptoms used in pre-classification are based on Global Initiative for Asthma (GINA) and Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines. The pre-classification algorithm is validated on 5000 reports acquired from subjects prospectively enrolled in the Hospital in Sarajevo during the period of two year and CareFusion Database for the last 10 years. Sensitivity of 97.26% and specificity of 90.74% is achieved. It is shown that saving around 97% on pulmonary functions tests can be achieved by introducing these automated systems in everyday practices.