HARNESSING THE POWER OF ARTIFICIAL INTELLIGENCE IN PEDIATRIC NEPHROLOGY: A COMPREHENSIVE REVIEW OF EARLY DETECTION, DIAGNOSIS, AND MANAGEMENT OF KIDNEY DISEASES

: Artificial intelligence (AI) is revolutionizing the field of pediatric nephrology by offering innovative solutions for the early detection, diagnosis, and management of kidney diseases in children and neonates. This narrative review explores AI's current applications and future potential in enhancing the care of pediatric patients with acute kidney injury (AKI), chronic kidney disease (CKD), and glomerular diseases. We conducted a comprehensive literature search using Scopus, Web of Science, PubMed, and ScienceDirect to identify relevant studies published between 2015 and 2023. The findings demonstrate that AI models, such as XGBoost and logistic regression, have shown promise in predicting AKI by analyzing variables like serum creatinine and urine output. Integration of these models with electronic health record (EHR) systems has the potential to provide timely alerts and improve patient outcomes. In neonatal care, AI applications like the "Baby NINJA" model have significantly reduced nephrotoxic medication exposure and the incidence of AKI. In contrast, the "STARZ" model has exhibited high predictive accuracy for AKI within the first week of neonatal intensive care unit admission. AI is also being explored for its utility in managing glomerular diseases through digital pathology and enhancing the predictive accuracy of kidney biopsies with deep learning algorithms. However, these technologies are still primarily in the research phase and require further validation for clinical application. This review concludes that AI's integration into pediatric nephrology holds immense promise for improving diagnostic accuracy, personalizing treatment plans, and enhancing predictive assessments, potentially transforming kidney disease management in children and neonates.


INTRODUCTION
Artificial intelligence (AI) has emerged as a transformative technology in healthcare, offering novel solutions for early detection, diagnosis, and management of various diseases.In the field of pediatric nephrology, AI is increasingly being utilized to enhance the care of children and neonates with kidney diseases, such as acute kidney injury (AKI), chronic kidney 1 Centro Universitário Unifaminas. 2 UNESA -Estácio. 3Centro universitário UniFG. 4Santa Casa de Misericórdia de Vitoria Higher School of Sciences -EMESCAM.Santa Luíza -Vitória ES Brazil Molecular Dynamics and Modeling Laboratory (DynMolLab). 5Molecular Dynamics and Modeling Laboratory (DynMolLab) Santa Luíza -Vitória -ES -Brazil.disease (CKD), and glomerular diseases.This promising technology holds the potential to revolutionize the way we approach and manage these conditions, instilling hope for improved patient outcomes (1).AI's application in pediatric nephrology has the potential to significantly improve patient outcomes.By enabling earlier intervention, personalized treatment plans, and more accurate predictive assessments, AI can provide reassurance and confidence in the future of pediatric nephrology.(2).AKI and CKD are significant causes of morbidity and mortality in pediatric populations.The urgency of early detection cannot be overstated, as it is crucial for timely intervention and prevention of long-term complications.This underscores the importance of AI in improving the early detection of kidney diseases in pediatric nephrology (3).Traditional diagnostic methods, which often rely on serum creatinine and urine output measurements, may not detect kidney injury until significant damage has occurred.This underscores the need for AI to improve the early detection of kidney diseases (4).AI models, such as machine learning algorithms, have shown promise in predicting AKI by analyzing a combination of clinical variables and biomarkers, potentially allowing for earlier detection and intervention (5).
In neonatal care, the application of AI has demonstrated significant potential in reducing the incidence of AKI and optimizing medication management.Neonates, particularly those in the neonatal intensive care unit (NICU), are at high risk for developing AKI due to their immature kidney function and exposure to nephrotoxic medications.AI models, such as the 'Baby NINJA' and the 'STARZ' models, have shown promising results in reducing nephrotoxic medication exposure and predicting AKI within the first week of NICU admission, respectively (6,7).AI is also being explored for its utility in managing glomerular diseases, a heterogeneous group of disorders characterized by inflammation and damage to the glomeruli, the kidney's filtering units.The use of digital pathology and deep learning algorithms in this context can significantly enhance the predictive accuracy of kidney biopsies, potentially leading to earlier and more accurate diagnosis and prognostication of glomerular diseases (8).Digital pathology and deep learning algorithms are being developed to enhance the predictive accuracy of kidney biopsies and aid in the diagnosis and prognostication of glomerular diseases (9).However, these technologies are still primarily in the research phase and require further validation for clinical application.
This narrative review provides a comprehensive overview of AI's current applications and future potential in pediatric nephrology, focusing on early detection, diagnosis, and management of kidney diseases in children and neonates.We will discuss the methodology used to identify relevant studies, present the key findings, and provide a critical analysis of the current state of AI in this field.The review will conclude with recommendations for future research and clinical implementation of AI in pediatric nephrology.

METHODOLOGY
A comprehensive literature search was conducted using Scopus, Web of Science, PubMed, and ScienceDirect to identify relevant studies published between 2015 and 2023.The search strategy included a combination of keywords related to artificial intelligence, machine learning, pediatric nephrology, acute kidney injury, chronic kidney disease, glomerular diseases, and neonatal care.The search terms were adapted for each database to ensure optimal coverage of the relevant literature.The inclusion criteria for the studies were as follows: (1) original research articles, systematic reviews, or meta-analyses; (2) studies focusing on the application of AI in pediatric nephrology; (3) studies published in English; and (4) studies published between 2015 and 2023.Exclusion criteria included (1) conference abstracts, editorials, or commentaries, (2) studies focusing on adult populations; and (3) studies not directly related to AI in pediatric nephrology.From the initial 114 articles identified, 16 were selected for fulltext review.Two independent reviewers screened the titles and abstracts of the identified studies for relevance.Full-text articles were then assessed for eligibility based on the inclusion and exclusion criteria.Any disagreements between the reviewers were resolved through discussion and consensus.The reference lists of the included studies were also manually searched to identify any additional relevant articles.The two reviewers performed Data extraction independently using a standardized data extraction form.The extracted information included study characteristics (e.g., authors, year of publication, study design), population characteristics (e.g., age, sample size), AI methods used (e.g., machine learning algorithms, deep learning), and key findings related to the application of AI in early detection, diagnosis, and management of kidney diseases in pediatric populations.The quality of the included studies was assessed using appropriate tools, such as the Newcastle-Ottawa Scale for observational studies and the Cochrane Risk of Bias Tool for randomized controlled trials.The overall quality of evidence was then evaluated using the Grading of Recommendations, Similarly, a study developed a random forest model for predicting AKI in critically ill children (10,11).The model incorporated clinical variables, such as age, weight, and laboratory values, and achieved an AUC of 0.86.The authors emphasized integrating these predictive models with electronic health record (EHR) systems to provide real-time alerts and support clinical decision-making.

Integration of AI Models with EHR Systems
Integrating AI models with EHR systems has been identified as a critical strategy for improving the early detection and management of AKI in pediatric populations.A study by Loftus et al. developed an AI-based alert system for AKI using data from a pediatric intensive care unit (PICU) (12).The alert system significantly reduced the time needed for AKI recognition and the initiation of appropriate management.
Similarly, a study by Sutherland et al. evaluated the impact of an EHR-integrated AKI alert system in a pediatric hospital (13).The system used a combination of serum creatinine and urine output criteria to identify patients at risk for AKI.Implementing the alert system significantly increased the proportion of patients receiving timely AKI-specific care, such as fluid management and nephrotoxic medication adjustment.

The "Baby NINJA" Model
The "Baby NINJA" model, developed by Stoops et al., is an AI-based approach to reducing nephrotoxic medication exposure and the incidence of AKI in neonates (6).The model utilizes machine learning algorithms to analyze medication data and identify neonates at high risk for nephrotoxic medication exposure.In a prospective study, the implementation of the "Baby NINJA" model resulted in a significant reduction in nephrotoxic medication exposure (42%) and AKI incidence (30%) compared to a historical control group.The authors highlighted the potential of this AI-driven approach to optimize medication management and improve outcomes in this vulnerable population.

The "STARZ" Model for AKI Prediction
The "STARZ" model is an AI-based tool for predicting AKI within the first week of neonatal intensive care unit (NICU) admission (1).The model incorporates clinical variables, such as gestational age, birth weight, and serum creatinine, and utilizes a gradient-boosting machine algorithm.In a retrospective study, the "STARZ" model demonstrated high predictive accuracy, with an AUC of 0.91.The authors emphasized the potential of this model to identify neonates at high risk for AKI and enable early intervention and management strategies.(15).The model demonstrated an AUC of 0.89, predicting CKD progression and outperforming traditional clinical risk factors.The authors suggested that this AI-based approach could aid in personalizing treatment plans and improving prognostic assessments for patients with glomerular diseases.

Limitations and Future Directions
While the application of AI in the management of glomerular diseases shows promise, several limitations need to be addressed.Current studies primarily focus on developing and validating AI models using retrospective data, and prospective studies are required to evaluate their clinical utility and impact on patient outcomes.Additionally, the generalizability of these models across different populations and healthcare settings needs to be established.
Future research should focus on integrating AI models with EHR systems to provide real-time decision support for managing glomerular diseases.Developing explainable AI models that provide transparent and interpretable predictions is crucial for building trust and facilitating clinical adoption.Furthermore, AI's ethical and legal implications in diagnosing and managing glomerular diseases must be carefully considered.

DISCUSSION
The findings of this narrative review highlight the significant potential of AI in revolutionizing the early detection, diagnosis, and management of kidney diseases in pediatric populations.Machine learning models, such as XGBoost and logistic regression, have demonstrated high predictive accuracy for AKI in children, enabling earlier intervention and potentially improving patient outcomes.The integration of these models with EHR systems has been shown to provide real-time alerts and support clinical decision-making, reducing the time to AKI recognition and initiating appropriate management.
In neonatal care, AI applications like the "Baby NINJA" and "STARZ" models have shown promising results in reducing nephrotoxic medication exposure and predicting AKI within the first week of NICU admission, respectively.These AI-driven approaches can optimize medication management and improve outcomes in this vulnerable population.
The application of AI in managing glomerular diseases through digital pathology and deep learning algorithms has also shown promise in enhancing the diagnostic accuracy and predictive assessments of kidney biopsies.However, these technologies are still primarily in the research phase and require further validation through prospective studies and evaluation of their clinical utility and impact on patient outcomes.
Integrating AI into pediatric nephrology practice presents several challenges and limitations that must be addressed.The development of AI models requires large, diverse, and high-quality datasets, which may be challenging to obtain in pediatric populations due to the relatively low incidence of kidney diseases and ethical considerations surrounding data collection and sharing.Additionally, the generalizability and transferability of AI models across different healthcare settings and populations need to be established to ensure their widespread adoption and impact.
Another important consideration is the potential for AI models to perpetuate or amplify existing biases and disparities in healthcare.Careful attention must be paid to ensuring that AI models are developed using diverse and representative datasets and that their performance is evaluated across different subgroups to identify and mitigate potential biases.
The successful integration of AI into pediatric nephrology practice also requires close collaboration between healthcare professionals, data scientists, and AI experts.Clinicians need to be involved in developing and validating AI models to ensure their clinical relevance and usability.Additionally, there is a need for education and training programs to equip healthcare professionals with the necessary skills and knowledge to use and interpret AI-based tools in clinical practice effectively.

CONCLUSION
This narrative review demonstrates the significant potential of AI in transforming the early detection, diagnosis, and management of kidney diseases in pediatric populations.
Machine learning models have shown promise in predicting AKI in children, enabling earlier intervention and potentially improving patient outcomes.AI applications in neonatal care, such as the "Baby NINJA" and "STARZ" models, have demonstrated their utility in reducing nephrotoxic medication exposure and predicting AKI in this vulnerable population.The application of AI in managing glomerular diseases through digital pathology and deep learning algorithms has the potential to enhance diagnostic accuracy and prognostic assessments.However, several challenges and limitations must be addressed to fully realize AI's potential in pediatric nephrology.These include the need for large, diverse, and high-quality datasets, the generalizability and transferability of AI models across different healthcare settings and populations, and the potential for AI to perpetuate or amplify existing biases and disparities in healthcare.
Future research should focus on conducting prospective studies to evaluate AI models' clinical utility and impact on patient outcomes, developing explainable AI models that provide transparent and interpretable predictions, and addressing the ethical and legal implications of using AI in pediatric nephrology practice.
In conclusion, integrating AI into pediatric nephrology holds immense promise for improving the early detection, diagnosis, and management of kidney diseases in children and neonates.However, close collaboration between healthcare professionals, data scientists, and AI experts is essential to ensure the responsible development and deployment of AI-based tools in clinical practice.With ongoing research and innovation, AI has the potential to revolutionize the field of pediatric nephrology and improve outcomes for patients worldwide.
Assessment, Development, and Evaluation (GRADE) approach.A narrative synthesis of the findings focused on AI's current applications and future potential in pediatric nephrology.The synthesis was organized into three main themes: (1) early detection and prediction of AKI, (2) AI applications in neonatal care, and (3) AI in the management of glomerular diseases.The limitations of the current evidence and recommendations for future research were also discussed.RESULTS 1. Early Detection and Prediction of Acute Kidney Injury (AKI) 1.1.Machine Learning Models for AKI Prediction Several studies have investigated using machine learning models to predict AKI in pediatric populations.Raina et al. evaluated the performance of XGBoost and logistic regression models in predicting AKI using variables such as serum creatinine, urine output, and demographic data (1).The XGBoost model demonstrated superior performance, with an area under the receiver operating characteristic curve (AUC) of 0.92, compared to 0.88 for the logistic regression model.The authors highlighted the potential of these models to provide early detection of AKI and allow for timely intervention.

3 .
AI in the Management of Glomerular Diseases 3.1.Digital Pathology and Deep Learning Algorithms AI is being explored for its utility in managing glomerular diseases through digital pathology and deep learning algorithms.A study by Kannan et al. developed a deep-learning model for the automated classification of glomerular lesions in kidney biopsy images (14).The model achieved an accuracy of 87% in classifying lesions such as focal segmental glomerulosclerosis, membranous nephropathy, and minimal change disease.The authors highlighted the potential of this AI-based approach to assist pathologists in diagnosing and predicting glomerular diseases.
Similarly, a study by Ginley et al. evaluated the performance of a deep-learning model in predicting the risk of progressive chronic kidney disease (CKD) based on histopathological features in kidney biopsy images