Strategies for overcoming data scarcity, imbalance, and feature selection challenges in machine learning models for predictive maintenance.

Predictive maintenance harnesses statistical analysis to preemptively identify equipment and system faults, facilitating cost- effective preventive measures. Machine learning algorithms enable comprehensive analysis of historical data, revealing emerging patterns and accurate predictions of impending system failures. Common hurdles in applying ML algorithms to PdM include data scarcity, data imbalance due to few failure instances, and the temporal dependence nature of PdM data. This study proposes an ML-based approach that adapts to these hurdles through the generation of synthetic data, temporal feature extraction, and the creation of failure horizons. The approach employs Generative Adversarial Networks to generate synthetic data and LSTM layers to extract temporal features. ML algorithms trained on the generated data achieved high accuracies: ANN (88.98%), Random Forest (74.15%), Decision Tree (73.82%), KNN (74.02%), and XGBoost (73.93%).

Clinical Characteristics and Early Outcomes of Hospitalized COVID-19 Patients with End-Stage Kidney Disease in Saudi Arabia.

BACKGROUND: The prognosis of the novel coronavirus disease 2019 (COVID-19) may be poor in patients with end-stage kidney disease (ESKD). Limited information is available on the clinical characteristics and outcomes of such patients in Arab countries. The present study aimed to address this gap. METHODS: This retrospective cohort study included 101 patients with ESKD who were hospitalized for COVID-19 between May 1, 2020 and December 31, 2020 at King Fahad Central Hospital (single center), Jazan Province, Saudi Arabia. Chi-square test, independent samples t-test, Mann-Whitney U-test, and Cox regression were performed in the statistical analysis. RESULTS: Of the 101 hospitalized COVID-19 patients with ESKD, 20 patients died (19.8%). Patients aged ≥ 65 years had a significant mortality risk. Mortality was higher in male patients (70%) than in female patients (30%). The most common comorbidities were hypertension (88%), diabetes mellitus (47%), and heart disease (17%). The presence of diabetes and cardiovascular diseases along with ESKD increased the mortality risk [relative risk (RR) = 6.5 and RR = 3.8, respectively]. The most frequently reported clinical symptoms were fever (50%), shortness of breath (41%), and cough (35%). In total, 17% of the patients were admitted to the intensive care unit and required mechanical ventilation. Most patients had bilateral lung infiltrates (88%). Hemoglobin levels and platelet counts were significantly lower in the deceased patients than in the surviving ones. Moreover, compared with the surviving patients, higher degree of lymphocytopenia, neutrophilia, ferritin, D-dimer, blood urea nitrogen, and aspartate transaminase were detected in the deceased patients. Septic shock (20%) and respiratory failure (19%) were the most prevalent complications resulting in death. CONCLUSION: COVID-19 patients with ESKD have an increased risk of poor outcomes and mortality. The mortality risk increases with an increase in age and the presence of other comorbidities, such as diabetes and cardiovascular disease. Elevated levels of inflammatory markers correlate with disease severity and are associated with in-hospital mortality in this population.

Recessively inherited severe aortic aneurysm caused by mutated EFEMP2.

Familial aortic aneurysm (AA) is mostly inherited as an autosomal dominant disorder. However, recessively inherited AA has also been observed but in association with skin manifestations of cutis laxa, which is caused by a mutated EFEMP2 gene. In the present study, we recruited 9 patients, from 4 unrelated consanguineous families, with recessively inherited AA. The index cases, their parents, and siblings underwent clinical evaluation and cardiac imaging. In the affected subjects, the clinical presentation ranged from sweating and cyanosis at 3 months of age to incidental findings in an asymptomatic adult. The echocardiogram revealed a wide spectrum of severity of the AA, with a Z-score varying from 5 to 33. Intrafamilial variability was also evident; 2 unrelated subjects were detected at 17 and 20 years of age through family screening. The skin manifestations of cutis laxa were not found in any patient. In 1 family, genome-wide single-nucleotide polymorphism analysis detected a homozygous block, shared by 2 affected siblings, on chromosome 11 at q13. Sequence analysis of EFEMP2, located on chromosome 11 at q13, identified a novel homozygous mutation (p.E161K) in all 9 affected subjects. In this largest cohort of reported patients with a mutated EFEMP2 gene, we illustrate the phenotypic spectrum of inherited AA due to a novel EFEMP2 mutation. In conclusion, our work suggests that in families with apparently recessively inherited AA, molecular analysis of EFEMP2 gene might be warranted.

Prevalence of anemia in predialysis chronic kidney disease patients.

To evaluate the prevalence of anemia in a large cohort that comprises patients in different stages of chronic kidney disease (CKD) in the kingdom of Saudi Arabia (KSA), we conducted a multi-center cross-sectional study of a cohort of CKD patients who have not started dialysis. The study patients were recruited from the nephrology clinics in 11 different medical centers distributed all over the regions of the KSA. For the estimated glomerular filtration rate (GFR), we used the Chronic Kidney Disease-Epidemiology Collaboration (CKD-EPI) equation. There were 250 study patients who fulfilled the criteria for the study. The patients were stratified according to their GFR as follows: stage 1: 19 patients, stage 2: 35 patients, stage 3: 67 patients, stage 4: 68 patients, and stage 5: 61 patients. The composite of proteinuria and abnormal imaging in stages 1 and 2 was satisfied in 100% of the cases. The prevalence of anemia was elevated for the hemoglobin levels below 12 g/dL (the level at which the evaluation of anemia in CKD should be initiated) in the different stages of CKD, that is, 42%, 33%, 48%, 71%, and 82% in the stages from 1 to 5, respectively. The prevalence was also elevated for the hemoglobin levels below 11 g/dL (the minimum hemoglobin level at which therapy should be initiated with erythropoietin), that is, 21%, 17%, 31%, 49%, and 72%, respectively for stages from 1 to 5. In conclusion, we found a large prevalence of anemia among the CKD population in Saudi Arabia, and the burden of patients who require treatment with erythropoietin is considerably large. However, the response to therapy will not require large doses according to the availability of long-acting erythropoiesis stimulating agents, which will render the therapy more convenient and less expensive.

What is the most appropriate formula to use in estimating glomerular filtration rate in adult Arabs without kidney disease?

BACKGROUND: In clinical practice, the glomerular filtration rate (GFR) is often estimated by the Modification of Diet in Renal Disease (MDRD) or Cockcroft-Gault (CG) formulae. No data are available, however, on the performance of these formulae in Arab individuals. METHODS: Plasma creatinine samples were obtained from 90 consecutive normal Arab kidney donors for the estimation of GFR (eGFR) using the simplified MDRD and CG formulae. The GFR was measured in these donors with chromium labelled EDTA {[51Cr] EDTA). Bias was assessed by calculating the difference between the measured GFR and the calculated GFR using each of the two formulae; precision was calculated using the r value of the regression analysis. RESULTS: The group studied consisted of 90 donors, of whom 64 were males (71%). The mean age was 30.8 years (+/- 9.8) and mean BMI was 25.7 (+/- 5.7). The measured GFR (mean 112.4 +/- 17.5) correlated better with the calculated GFR by CG formula (mean 107.7 +/- 29.7) and showed poor correlation with the GFR estimated by the MDRD (mean 89.2 +/- 13.8); bias = 4.8 and 23.3, respectively (p = 0.1 and < 0.0001, respectively). The correlation with CG formula was better in males (bias = 2, p = 0.5) and those under 30 years of age (bias = 1.0, p = 0.9). Based on our data, we calculated a correction factor to the CG formula to improve the correlation with the measured GFR in Arab individuals. By multiplying the CG formula by 1.0446, the bias was reduced from 4.8 (p = 0.1) to 0.0 (p = 0.5) with an increase in precision from 0.2 (p = 0.05) to 0.43 (p = 0.0001). Using CG formula, the frequency for values within 30% of the mean of the measured value was 75%, which improved to 80% using the revised formula. CONCLUSIONS: CG formula was found to be the most appropriate for calculation of GFR in Arab individuals. It is possible to reduce the bias and improve precision in Arab individuals with normal renal function by multiplying the result obtained by CG formula by 1.0446.