Leveraging AI Models for Enhanced Urine Analysis: Evaluating YOLOv8 and MobileNet-v2 in Dehydration Detection Post-COVID-19

The COVID-19 pandemic has underscored the critical importance of effective health management, particularly in monitoring hydration levels due to the virus’s extensive multisystem impacts. This study delves into the potential of advanced artificial intelligence (AI) models, specifically YOLOv8 and MobileNet-v2, in accurately analyzing urine color and levels as indicators of dehydration. By utilizing a dataset from Roboflow Universe, comprehensive data augmentation techniques were employed to enhance model training, ensuring robustness across various lighting conditions and diverse realworld scenarios. Both YOLOv8 and MobileNet-v2 models achieved perfect scores of $\mathbf{1. 0}$ for accuracy, recall, precision, and F1-score in the urine sample classification task. However, YOLOv8 demonstrated superior performance in real-time detection, positioning it as the more suitable candidate for clinical applications. The confusion matrix analysis revealed that YOLOv8 accurately classified all urine samples, while MobileNet had some misclassifications in certain categories. This study underscores AI’s transformative potential in revolutionizing health diagnostics, providing timely and accurate solutions for dehydration detection. The findings advocate for the future integration of AI models into portable health devices and mobile applications, democratizing access to timely diagnostic tools and empowering individuals to proactively monitor their hydration status. Such advancements are crucial in the post-COVID-19 landscape, where hydration management has gained heightened importance.