Prediction of Chronic Kidney Disease Using Ammonia Gas Sensor with Supports of Other Non-Invasive Attributes
作者
Mae M. Garcillanosa,Jherome Arroyo,Hamina Villanueva,Patricia Anne Gonzales
出处
期刊:日期:2024-06-14卷期号:: 63-67被引量:2
标识
DOI:10.1145/3678935.3678946
摘要
Chronic kidney disease (CKD) is a condition characterized by gradual loss of kidney function over time and it has become a major public concern with its rising prevalence. Diagnosing CKD is generally invasive, costly, time-consuming, and often risky. To address the disadvantages of CKD detection, we proposed a prediction model using non-invasive predictive parameters (blood pressure, appetite, diabetes, coronary, hyper-tension, pedal edema, and anemia) and the measurement of ammonia concentration in breath to classify patients with chronic kidney disease. The mean breath ammonia levels of CKD patients were significantly higher with a value of 3.868 ppm versus those of healthy participants with a mean ppm value of 0.678 ppm. Decision tree classifier is applied to the extracted features for classification. By using bootstrap aggregating (bagging) machine learning applied to decision trees in MATLAB, the accuracy of the model in the training datasets showed 99.02% of success rate in classifying the patients, while an accuracy of 100% was acquired in predicting the five samples of test data. The overall aim of this study is to construct a model designed to detect chronic kidney disease. The system is focused mainly on the breath samples for both classes of participants, with and without chronic kidney disease, with the aid of non-invasive attributes in improving machine learning accuracy. Using machine learning classification, particularly decision tree method, the system allows to build a model based on the breath samples and non-invasive attributes collected from the participants.