Secure RealTime Detection of RiskProne Patient Behavior in Hospitals via Multimodal Video Analytics

In modern healthcare environments, patient safety is a top priority, especially for elderly or mobility-impaired individuals. Falls and risky movements can lead to serious injuries, prolonged hospital stays, and increased healthcare costs. This project presents a secure, real-time system for detecting risk-prone patient behaviors using multimodal video analytics. Leveraging the YOLOv8 deep learning algorithm and a custom-trained dataset from Roboflow, the system identifies critical actions such as falling, lying down, sitting, standing, moving, and interactions with medical aids like walkers and wheelchairs. A live webcam feed serves as the primary input source, enabling continuous monitoring without human intervention. The front-end is developed using Streamlit to provide an intuitive and responsive web interface, while the back-end is powered by Python for real-time processing and detection. This solution offers hospitals an effective way to enhance patient care, reduce risks, and enable quick response to emergencies. By integrating AI with healthcare monitoring, the system demonstrates a scalable and practical approach to real-world medical safety challenges.

Keywords:
fall detection, patient safety, real-time monitoring, YOLOv8, video analytics, hospital surveillance, deep learning, behavior detection, Streamlit, healthcare AI