AI – Enhanced Nighttime Seizure Surveillance

AI - Enhanced Nighttime Seizure Surveillance

Mr. JAGAN M                                       Mr. RAJAGUHAN A                                   Mr. PRANESH N S

Department of Artificial Intelligence            Department of Artificial Intelligence           Department of Artificial Intelligence

and Machine Learning                                 and Machine Learning                                  and Machine Learning

Sri Shakthi Institute of Engineering              Sri Shakthi Institute of Engineering             Sri Shakthi Institute of Engineering

and Technology, Coimbatore, India              and Technology, Coimbatore, India             and Technology, Coimbatore, India

Mr. PRATHAP E                                                   Ms. NIVEDHA S

Department of                                                     Assistant   professor

Artificial Intelligence                                     Department of Artificial Intelligence

and Machine learning                                              and Machine Learning

Sri Shakthi Institute of Engineering                        Sri Shakthi Institute of Engineering

and Technology, Coimbatore, India                        and Technology, Coimbatore, India

Abstract – Nighttime seizures pose serious health risks to individuals with epilepsy, often going undetected until critical complications occur. This system tackles these challenges by using advanced computer vision and AI to create a non-invasive, real-time seizure detection solution. It integrates technologies like MediaPipe, OpenCV, and edge computing to analyze video streams and detect seizure-specific patterns through body pose estimations and physiological inputs. High-resolution video feeds, enhanced by infrared capabilities, enable effective nighttime surveillance. MediaPipe's Pose Estimation and Face Mesh modules ensure accurate real-time tracking of body movements, facial expressions, and hand gestures. OpenCV aids in preprocessing by removing noise, segmenting motion areas, and capturing subtle body changes critical for detection. Designed with efficiency and patient comfort in mind, the system uses local edge computing to reduce latency and enhance data privacy by avoiding cloud transmissions. Training with annotated datasets helps distinguish between regular sleep movements and seizure-related activities, reducing false alarms. Integration with wearable and environmental sensors strengthens the system by capturing physiological changes and contextual room data. Outputs include real-time alerts, visual overlays on detected video frames, and detailed activity logs. This innovative solution combines machine learning, real-time analytics, and advanced vision technologies, paving the way for reliable, non-invasive seizure detection systems in epilepsy care and broader healthcare monitoring scenarios.

Key Words: OpenCV, MediaPipe, Vision Transformers (ViG), Artificial Intelligence (AI), Electronic Health Records (EHR), Machine Learning (ML), Convolutional Neural Networks (CNNs),