Seizure intensity estimation using video recording

Epilepsy is one of the most common neurological diseases worldwide. It affects around 50 million people worldwide. Therefore, it is believed to be one of the most common neurological diseases globally. It is defined as a state of recurrent seizures, which induce, in many cases, involuntary and abnormal movements . These movements may involve a part of the body (partial seizures) or the entire body parts (generalized seizures). This motor manifestation during seizures represents a clinical marker of seizure identification and has a significant role in evaluating epilepy in addition to the treatment impact on the progression of the disease.The current method of epileptic seizures analysis are qualitatively evaluated by professionals with visual inspection. Threrefore it is a highly subjective method and susceptible to human errors and bias. In this thesis, an automatic tool (i.e. pose estimation) was used to obtain 3D kinematic analysis of epileptic seizures from stereo video-recordings. The data samples and the automatic tool used in this thesis are provided by Neuroeventlabs Oy. Seizure events belong to a patient with Lennox–Gastaut syndrome (LGS). They consist of four nocturnal seizure events recorded dur-ing sleep at home. 3D kinematic analysis of epileptic seizure events was conducted for intensi-ty assessment. To provide a qualitative assessment, visual assessment was done on the sei-zure events as well as a seizure-free event during REM sleep to assess their severity on a sim-ple rating scale having categorical values (i.e. weighted score) from (0-9). The main goal of the thesis is to conduct a comparative experiment between the human visual assessment (as the gold standard in the field) and the automatic quantitative approach in a control-free environment for to demonstrate whether it is possible to use the current computer vision and deep learning techniques for conducting motion analysis during seizures