Decentralized Tiny Ensemble Learning for Privacy-Preserving Compliance Prediction in Operating Room Environments

Abstract

Healthcare environments, particularly operating rooms, face significant challenges in maintaining optimal conditions to prevent postoperative infections. These challenges include the absence of real-time monitoring systems, the limited availability of labeled medical data, and the difficulty of deploying intelligent solutions in resource-constrained settings. Additionally, there is a growing need to explore emerging technologies in IoT and TinyML to modernize healthcare monitoring and address existing gaps in system intelligence, scalability, and privacy. This paper presents a privacy-preserving intelligent system designed to predict the compliance of operating room environments with ISO standards by continuously monitoring key parameters such as temperature, humidity, air quality, and pressure. The system employs a decentralized Tiny ensemble learning architecture, utilizing ESP32 microcontrollers as edge learners and a Raspberry Pi 4 as the aggregator node. This approach improves prediction accuracy to 98.74 % while preserving data privacy through localized processing. To address data scarcity, recent synthetic data generation techniques from the literature are employed to create comprehensive datasets for training and validation.

Ähnliche Arbeiten