Study of the Sine-SSA-BP Model in Classification of Cognitive Impairment by Eye Movement Features for Patients with Epilepsy The Sine-SSA-BP Model in Classification for epilepsy

Most studies reported that patients with epilepsy could suffer from attention dysfunction and other social cognitive impairment but there were few studies on automatic detection for patients with epilepsy of cognitive impairment based on eye-tracking technology. The current study aimed to explore objective and nontraumatic method of assisting in the detection of patients with epilepsy of cognitive impairment. Thirty-seven patients with epilepsy of cognitive impairment and twenty-nine healthy controls performed the Attention Network Test (ANT) based on eye-tracking technology. The random forest algorithm combined with the principal component analysis was applied to extract the main eye-tracking features, and then the back propagation neural network model was carried out to identify patients with epilepsy of cognitive impairment. To improve the accuracy of the classification model, the sparrow search algorithm (SSA) with Sine chaotic mapping was used to optimize the initial weight and threshold of the back propagation (BP) network. The results showed that compared with the BP network accuracy of 60.00% and the BP network optimized by the SSA accuracy of 90.45%, the BP network optimized by SSA with Sine chaotic mapping had the higher classification accuracy of 96.67%. It proved that Sine-SSA-BP based on eye-tracking features can facilitate early detection for patients with epilepsy of cognitive impairment.