1. Valentina Matovic, Mašinski fakultet Univerziteta u Beogradu, Serbia
2. Jasna Trbojević-Stanković, Medicinski fakultet Beograd, Serbia
3. Lidija Matija, Masinski fakultet, Univerzitet u Beogradu, Serbia
4. Dušan Šarac, Masinski fakultet, Univerzitet u Beogradu, Serbia
5. Aleksandra Vasić-Milovanović, Masinski fakultet, Univerzitet u Beogradu, Serbia
6. Andrija Petrović, Masinski fakultet, Univerzitet u Beogradu, Serbia
7. Nikola Stojiljković, Mašinski fakultet Univerziteta u Beogradu, Serbia
Chronic inflammation contributes to the pathogenesis of several complications in hemodialysis (HD) patients. The high concentration of C-reactive protein (CRP) is an indicator of an inflammatory condition. The increase in serum CRP level is an independent determinant of cardiovascular events in long-term HD patients.
The aim of this study was to predict CRP level in HD patients from the matrix of the spent dialysate fluid using near infrared spectroscopy (NIRS). The serum CRP values were presented in the form of a binomial variable, where zero indicated a normal CRP level (below 6 mg/l) and one represented CRP level that is beyond the normal limit (above 6 mg/l). We used several Machine Learning (ML) algorithms: Random Forest (RF), Logistic Regression, KNN (K-nearest neighbor), Support Vector Machine (SVM), Decision Tree Classifier, and Gaussian Naive Bayes (NB) to classify CRP level within reference range and CRP level beyond the normal limit. These classifier methods were used on the same dataset, and Area Under the Curve (AUC) evaluation was performed.
RF and KNN have shown the best classification accuracy for the prediction of CRP blood level, while LR, SVM Decision Tree and NB, have shown average accuracy. AUC score of RF algorithm was 95% with accuracy of 91%, and AUC score of KNN algorithm was 91% with accuracy of 75%.
The NIRS method with ML algorithms can be used to predict CRP blood level in HD patients.
Hemodialysis; Machine learning; spent dialysate; VIS-NIR; patient-specific; Hemodialysis; Machine learning; spent dialysate; VIS-NIR; patient-specific; Hemodialysis; Machine learning; spent dialysate; VIS-NIR; patient-specific
SIMPOZIJUM B - Biomaterijali i nanomedicina