Maximum-likelihood detection of neonatal clonic seizures by video image processing

In this paper we consider the use of a well-known statistical method, namely Maximum-Likelihood Detection (MLD), to early diagnose, through a wire-free low-cost video processing-based approach, the presence of neonatal clonic seizures. Since clonic seizures are characterized by periodic movements of parts of the body (e.g., hands, legs), by evaluating the periodicity of the extracted signals it is possible to detect the presence of a clonic seizure. The proposed approach allows to differentiate clonic seizure-related movements from random ones. While we first consider a single-camera scenario, we then extend our approach to encompass the use of multiple sensors, such as several cameras or the Microsoft Kinect RBG-Depth sensor. In these cases, data fusion principles are considered to aggregate signals from multiple sensors.