MACHINE LEARNING-BASED DETECTION AND COMPARATIVE ANALYSIS OF THE SWIFT RESPIRATORY DISTRESS SYNDROME (SRDS) BASED ON VOCAL

Abstract

Patients with severe conditions such as sepsis, pneumonia, are at increased risk for developing Swift Respiratory Distress Syndrome (SRDS), fulminant inflammatory lung damage. Unfortunately, many people who acquire SRDS are not diagnosed with the condition and so may not get therapy that might improve their prognosis. Due to the clinical nature of SRDS, diagnostic confusion (label uncertainty) may arise while treating a patient. In addition, a chest x-ray is necessary for the diagnosis; however, this is test that isnt always readily accessible in a clinical context. For this reason, we develop machine learning-based model for assessing the risk of SRDS, using patient's respiration sounds as both data for training and testing, as well as random forest, gradient boosting, & LR, & comparing results with SVM