Anzeige
Erschienen in:
2019 | OriginalPaper | Buchkapitel
The Application of Machine Learning Algorithms to Diagnose CKD Stages and Identify Critical Metabolites Features
verfasst von : Bing Feng, Ying-Yong Zhao, Jiexi Wang, Hui Yu, Shiva Potu, Jiandong Wang, Jijun Tang, Yan Guo
Erschienen in: Bioinformatics and Biomedical Engineering
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Abstract
Background: Chronic kidney disease (CKD) is a progressive and heterogeneous disorder that affects kidney structures and functions. Now it becomes one of the major challenges of public health. Early-stage detection, specialized stage treatments can significantly defer or prevent the progress of CKDs. Currently, clinical CKD stage diagnoses are mainly based on the level of glomerular filtration rate (GFR). However, there are many different equations and approaches to estimate GFR, which can cause inaccurate and contradictory results.
Methods: In this study, we provided a novel method and used machine learning techniques to construct high-performance CKD stage diagnosis models to diagnose CKDs stages without estimating GFR.
Results: We analyzed a dataset of positive metabolite levels in blood samples, which were measured by mass spectrometry. We also developed a feature selection algorithm to identify the most critical and correlated metabolite features related to CKD developments. Then, we used selected metabolite features to construct improved and simplified CKD stage diagnosis models, which significantly reduced the diagnosis cost and time when compared with previous prediction models. Our improved model could achieve over 98% accuracy in CKD prediction. Furthermore, we applied unsupervised learning algorithms to further validate our models and results. Finally, we studied the correlations between the selected metabolite features and CKD developments. The selected metabolite features provided insights into CKD early stage diagnosis, pathophysiological mechanisms, CKD treatments, and drug development.