Detection of eye movements during epileptic seizures by using image and signal processing techniques

Although epilepsy can be seen in all age groups, it is a common health problem that especially concerns the childhood age group. Detection of epileptic seizures and early warning systems are increasingly in demand and research. In our study, it was aimed to learn the effects of seizures on eye movements during EEG monitoring and to collect data by using machine vision and artificial intelligence techniques first; then to automatically detect the moment of seizure and even before and after the seizures if possible. Patients who applied to Akdeniz University Faculty of Medicine Pediatric Neurology Clinic between April 2018 and June 2020 were evaluated prospectively. In order to detect the eye findings of the patients, a special device supported by artificial intelligence was designed in cooperation with the Faculty of Engineering; eye movements were recorded by the designed device simultaneously with the Video-EEG in all patients. The detailed patient history, physical examination, laboratory tests, cranial MRI, interictal EEG, VEM and eye movements of 60 patients included in the study were examined. The results of 10 patients were evaluated who had sufficient data in accordance with the study design. Six of the patients were female (%60) and 4 were male (%40). The mean age was 105 ± 56 (50-216) months, the mean age during EEG was 93 ± 56 (38-204) months, and the mean age at first seizure onset was 40 ± 60 (1-180) months. All were mature births by gestational age. Five (%50) of them had additional diseases. Two patients had a history of status epilepticus (%20), 3 patients had a history of febrile convulsion (%30). One patient had a family history of febrile convulsion (%10), 4 patients had a family history of epilepsy (%40), and 2 patients had a parental consanguinity (%20). All patients were using antiepileptic drugs (AEDs). Seven patients (%70) received polytherapy and 3 patients (30%) received monotherapy; 4 patients (%40) were using more than 3 AEDs. Five patients (%50) had neuromotor and mental retardation, and 3 patients (%30) had abnormal findings on neurological examination. Pathological findings were detected in cranial MRI in 5 patients (%50). It was observed that all patients had seizures while awake, no patient had seizures during sleep. Four patients (%40) had absence, 2 patients (%20) had atonic, 2 patients (%20) had tonic, 1 patient (%10) had clonic, and 1 patient (%10) had tonic- clonic seizures. Eye opening was monitored during the seizure in all patients, and pathological eye movements were detected in all of the patients records. In 1 patient (10%), it was observed that eye movements started before the seizure activity on the EEG. Eye deviation was observed in 4 patients (%40); 2 (%20) had right deviation, 1 (%10) had left deviation, 2 (%20) had upward deviation and 1 (%10) had downward deviation. In 3 patients (%30), significant change in pupil diameter was detected during the ictal period; 2 had pupillary hippus and 1 had bilateral mydriasis. When the patients were grouped according to their demographic and clinical characteristics and the seizure durations recorded during VEM and pathological eye movement periods were examined; no statistically significant difference was found between the groups in terms of the duration of seizures and pathological eye movement in terms of gender, presence of additional disease, status history, FC history, family history of FC, epilepsy history in the family, and parental consanguinity (p>0.05). In patients with neuromotor retardation and mental retardation, the seizure durations detected on EEG and the duration of pathological eye movements were found to be statistically significantly longer than patients without retardation (p<0.05). Similarly, with the EEG and clinically detected seizure durations of patients with abnormal findings on neurological examination was found statistically significantly longer than patients with normal neurological examination (p<0.05). When the patients were grouped according to the number of antiepileptic drugs (AEDs) they used, with the EEG and clinically detected seizure durations of the patients using 3 or more AEDs was found statistically significantly longer than the patients using less than 3 AEDs (p<0.05). In our study, pathological eye movements in the preictal, ictal and postictal periods were detected and recorded with the device we developed in all patients. In 1 patient, it was possible to detect pathological eye movements before seizure activity on EEG started. With the development of the device we designed, it is predicted that eye movements can be evaluated and recorded much more objectively thanks to the analysis supported by artificial intelligence. Perhaps, with early detection of eye movements, it will be possible to warn the patient before the seizure activity begins. In the light of the data we have obtained, it has been seen that our study is promising. However, due to the limited number of patients, randomized controlled studies with more patients are needed to evaluate them with stronger data and analyzes. Keywords: Epilepsy, Epileptic eye movements, Seizure detection device