We are overwhelmed by a deluge of data and, although its interpretation is challenging, fortunately, information technology comes to the rescue. One of the tools is artificial intelligence, allowing the identification of relationships between variables and their arbitrary classification. We focused on the assessment of both the remission of proteinuria and the deterioration of kidney function in patients with IgA nephropathy, comparing several methods of machine learning. It is of utmost importance to respond to subtle changes in kidney function, which will lead to a deceleration of the disease. This goal has been achieved by analyzing regression techniques, predicting the difference in serum creatinine concentration. We obtained the performance of the tested models which classified patients with high accuracy (Random Forest Classifier showed an accuracy of 0.8–1.0, Multi-Layer Perceptron an Area Under Curve of 0.8842–0.9035 and an accuracy of 0.7527–1.0) and regressors with a low estimation error (Decision Tree Regressor showed MAE 0.2059, RMSE 0.2645). We have demonstrated the impact of both model selection and input features on performance. Application of machine learning methods requires careful selection of models and assessed parameters. The computing power of modern computers allows searching for the models most effective in terms of accuracy.
Delayed-graft function (DGF) might be responsible for shorter graft survival. Therefore, a clinical tool predicting its occurrence is vital for the risk assessment of transplant outcomes. In a single-center study, we conducted data mining and machine learning experiments, resulting in DGF predictive models based on random forest classifiers (RF) and an artificial neural network called multi-layer perceptron (MLP). All designed models had four common input parameters, determining the best accuracy and discriminant ability: donor’s eGFR, recipient’s BMI, donor’s BMI, and recipient–donor weight difference. RF and MLP designs, using these parameters, achieved an accuracy of 84.38% and an area under curve (AUC) 0.84. The model additionally implementing a donor’s age, gender, and Kidney Donor Profile Index (KDPI) accomplished an accuracy of 93.75% and an AUC of 0.91. The other configuration with the estimated post-transplant survival (EPTS) and the kidney donor risk profile (KDRI) achieved an accuracy of 93.75% and an AUC of 0.92. Using machine learning, we were able to assess the risk of DGF in recipients after kidney transplant from a deceased donor. Our solution is scalable and can be improved during subsequent transplants. Based on the new data, the models can achieve better outcomes.
Background Lupus nephropathy (LN) occurs in approximately 50% of patients with systemic lupus erythematosus (SLE), and 20% of them will eventually progress into end-stage renal disease (ESRD). A clinical tool predicting remission of proteinuria might be of utmost importance. In our work, we focused on predicting the chance of complete remission achievement in LN patients, using artificial intelligence models, especially an artificial neural network, called the multi-layer perceptron. Methods It was a single centre retrospective study, including 58 individuals, with diagnosed systemic lupus erythematous and biopsy proven lupus nephritis. Patients were assigned into the study cohort, between 1st January 2010 and 31st December 2020, and eventually randomly allocated either to the training set (N = 46) or testing set (N = 12). The end point was remission achievement. We have selected an array of variables, subsequently reduced to the optimal minimum set, providing the best performance. Results We have obtained satisfactory results creating predictive models allowing to assess, with accuracy of 91.67%, a chance of achieving a complete remission, with a high discriminant ability (AUROC 0.9375). Conclusion Our solution allows an accurate assessment of complete remission achievement and monitoring of patients from the group with a lower probability of complete remission. The obtained models are scalable and can be improved by introducing new patient records.
The majority of recently published studies indicate a greater incidence and mortality due to Clostridioides difficile infection (CDI) in patients with chronic kidney disease (CKD). Hospitalization, older age, the use of antibiotics, immunosuppression, proton pump inhibitors (PPI), and chronic diseases such as CKD are responsible for the increased prevalence of infections. The aim of the study is to identify clinical indicators allowing, in combination with artificial intelligence (AI) techniques, the most accurate assessment of the patients being at elevated risk of CDI.
Background and Aims Children undergoing allogeneic hematopoietic stem cell transplantation (alloHSCT) are particularly vulnerable to acute kidney injury (AKI), especially in the early post-transplantation period. The major risk factors of AKI development are aggressive immunosuppression and infectious complications. In the meantime, malnutrition and hypermetabolic state of the patient, together with the routine intensive hydration during first 3 weeks after HSCT and subsequent forced diuresis, alter the serum creatinine concentration, modifying the estimated glomerular filtration rate (eGFR) value too. Therefore, the risk of underrating serum creatinine and overrating eGFR values is high, making the assessment of the degree of kidney damage during the first month after HSCT a challenge. Therefore, markers of tubular dysfunction and damage, like kidney injury molecule (KIM)-1, neutrophil gelatinase-associated lipocalin (NGAL) or interleukin (IL)-18, may be of added value while assessing renal function and analyzing the risk of AKI in this population. The aim of study was to assess the serum concentrations of damage biomarkers (KIM-1, NGAL, IL-18) in children undergoing alloHSCT, in relation to another surrogate marker of renal dysfunction, hyperfiltration. Another aim was to analyze the potential value of KIM-1, NGAL, and IL-18 as predictors of kidney damage in children after alloHSCT, with the use of artificial intelligence tools. Method The study group contained 22 children undergoing alloHSCT, followed up for 4 weeks after transplantation. Serum concentrations of KIM-1, NGAL, and IL-18 were assessed by ELISA in fixed time points (before HSCT, 1 day after HSCT, 1, 2 3, 4 weeks after transplantation). eGFR values (counted based on Schwartz formula) and the rate of hyperfiltration (eGFR > 140ml/min/1.73sq.m.) were evaluated at the beginning (before HSCT) and at the end (4 weeks after HSCT) of observation, when neither hydration nor diuretics were used. Statistical analysis was performed with the use of package Statistica, the comparisons between paired data were evaluated by using nonparametric tests (Friedman, Wilcoxon). Additionally, the patients within the database were randomly divided into two groups. The training group allowed to build a Random Forest Classifier (RFC) with the highest possible predictive power, while the testing group allowed to assess the effectiveness of prediction on new data and the clinical utility. Moreover, the contribution of individual variables was evaluated by GINI importance. Results KIM-1, NGAL, and IL-18 serum concentrations increased systematically until the 3rd week after HSCT, with statistically significant differences between subsequent observation points, then remained elevated until the 4th week after HSCT. Median eGFR values before transplantation and 4 weeks after HSCT were comparable, although the rate of patients with hyperfiltration increased. The RFC model built on the basis of 3 input variables, KIM-1, NGAL, and IL-18 concentrations in serum of children before HSCT, was able to effectively assess the rate of patients with hyperfiltration 4 weeks after the procedure. RF Classifier achieved AUROC of 0.8333, accuracy of 80.00%, positive predictive value of 0.8667, and sensitivity of 0.8000. The contribution of KIM-1, IL-18 and NGAL to the prediction in this model was comparable (33.73%, 32.77%, and 33.5%, respectively). Conclusion KIM-1, NGAL, and IL-18 are useful in assessing acute tubular damage in children after HSCT. Their values before HSCT may also serve as markers of incipient renal dysfunction 4 weeks after alloHSCT. The developed model seems a clinically useful tool to target patients who are at risk of kidney injury after HSCT. The Random Forest Classifier seems a promising tool for such analysis, that should be tested on a larger group of patients.
Background and Aims Progression of chronic kidney disease (CKD) is a compound process, where activation of immunocompetent cells and subclinical inflammation play pivotal role. Enhanced atrophy of the tubular cells, and finally, gradual fibrosis of tubulointerstitial tissue, are responsible for irreversible character of the disease. Multiple molecules influence above-mentioned processes. Growth differentiation factor (GDF)15, a member of TGF-β cytokine superfamily, is a marker of inflammation and an integrative signal in both acute and chronic stress conditions. Elevated serum concentrations of GDF15 were associated with increased risk of development and progression of CKD in adults, as well as with mortality in this group of patients. Our previous investigation revealed increased serum GDF15 concentrations in children on chronic dialysis. Epidermal growth factor (EGF), a tubule-specific protein, promotes proliferation, differentiation and migration of epithelial cells, and therefore, modulates regeneration of injured renal tubules. Decreased concentrations of EGF in urine were observed in variety of kidney diseases, including diabetic nephropathy, lupus nephritis or CKD. Our previous analysis of EGF serum concentrations in CKD children confirmed their decreased values on chronic dialysis. Neopterin is a product of activated monocytes and macrophages and serves as a marker of cell-mediated immunity. Elevated serum concentrations of neopterin were observed in CKD adult patients, our investigation revealed its increased values in children on chronic dialysis. None of the above mentioned markers was analyzed in the population of CKD children treated conservatively. Therefore, the aim of study was to assess the serum concentrations of EGF, GDF-15 and neopterin in children with CKD on conservative treatment and verify the usefulness of these markers in predicting CKD progression by means of artificial intelligence tools. Method The study group consisted of 153 children with pre-dialysis CKD stages 1-5 (stage 1 – 27 patients, stage 2 – 26 patients, stage 3 – 51 patients, stage 4 - 28 patients, stage 5 – 21 patients). EGF, GDF-15 and neopterin serum concentrations were assessed by ELISA. The patient database was implemented into the artificial neural network. In detail, the recursively selected subsets of input variables constituted the input layer of an artificial neural network built of perceptrons (multi-layer perceptron). Anthropometric data, biochemical parameters, EGF, GDF15 and neopterin were included into the model, serum creatinine and eGFR, as direct classifiers of CKD stage, were excluded. Various models were tested, regarding their accuracy, AUROC and Matthews correlation coefficient (MCC) values. Results EGF serum concentrations decreased gradually, whereas GDF15 and neopterin values rose systematically with CKD progression, keeping statistically significant inter-stage differences. Moreover, the most precise ANN model, among the tested artificial neural networks, contained EGF, GDF15 and neopterin as input parameters and classified patients into either CKD 1-3 or CKD 4-5 groups. This model has put new patients into appropriate classes with excellent Accuracy of 96.77%, AUROC 0.9169 and Matthews correlation coefficient (MCC) of 0.9157. Conclusion The presented model of an artificial neural network, with serum concentrations of EGF, GDF15 and neopterin as input parameters, may serve as a useful predictor of CKD progression in the pediatric population. It suggests the essential role of inflammatory processes, defined by newly discovered markers, in the renal function decline towards advanced stages of CKD in children.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.