Detection of Respiratory Disorder and Performance Analysis from PPG Signals

Diseases related to respiration are the most common disorders in India and around the world. There are several techniques for the initial diagnosis of respiratory disorders. One of the methods to find respiratory disease is calculating the PCO2 level in the blood. Capnography is the measurement of the partial pressure of CO2 (carbon-dioxide) in the respiratory gases during inhalation and exhalation. It is a non-invasive measurement of PCO2. Through capnography, information about airway management on alveolar ventilation and pulmonary perfusion can be found. In this paper, various kinds of respiratory disorders are found using classifiers and the performance of classifiers is analyzed. Dimensionality reduction techniques based on Expectation Maximization (EM) and Fast Fourier Transform (FFT) are used. Classifiers such as Linear Regression (LR), Non Linear Regression (NLR), Bayesian Linear Discriminant Classification (BLDC), Logistic regression (LSR), PSO (Particle swarm optimization), KNN (K-Nearest Neighbor) and a hybrid classifier KNN-NLR This paper discusses about seven classifiers, and it provides an effective method for diagnosis. The PSO Classifier gave the highest accuracy of 90.2439% with a low error rate of 9.7609% for both EM and FFT.