Enhanced Optimization of Fluid Volume Control in Hemodialysis Using Federated Learning

Abstract:
The medical field is undergoing significant advancements due to the rapid development of artificial intelligence and machine learning. This research aims to predict hydration in pediatric hemodialysis using hybrid machine-learning models with hyperparameters for prediction accuracy. Data from pediatric hemodialysis at the University Children's Hospital in Tiršova was collected, and parameters such as weight, blood pressure, lean tissue index, fat tissue index, body mass index, total body water, extracellular water, adipose tissue mass, body cell mass, and bioimpedance were adjusted for training. The model was configured for every pediatric patient and retrained after every treatment to make individualized predictions with the highest accuracy. The proposed model uses measurable parameters to estimate hydration and provide improved recommendations to the physician, yielding better results than typically used state-of-the-art competing methods. Federated learning with hyperparameters represents a novel, simplified, safe, and efficient approach to predicting hydration in children, making it easier for physicians to supervise the hemodialysis process in pediatric patients.

Bio: 

Dr. Vladimir M. Mladenovic is a Professor at the Faculty of Technical Sciences in Cacak, University of Kragujevac. Prof. Mladenovic’s research is in Artificial intelligence, Wireless communication, IoT. He is a dean and head of digital innovation lab at the Faculty of Technical Sciences and has over 20 years of research experience on wireless communications and signal processing and systems. He has authored more than 100 research articles, 10 books, and 40 patents.