Читать книгу Smart Healthcare System Design - Группа авторов - Страница 15

Karthikeyan M.P.1*, Krishnaveni K.2 and Muthumani N.3

¹Department of Computer Science, PPG College of Arts and Science, Coimbatore, India

²Department of Computer Science, Sri Ramasamy Naidu Memorial College, Sattur, India

³ PPG College of Arts and Science, Coimbatore, India

Abstract

The classification of medical data is the demanding challenge to be addressed among all research issues since it provides a larger business value in any analytics environment. Medical data classification is a mechanism that labels data enabling economical and effective performance in valuable analysis. Proposed research has indicated that the quality of the features may cause a backlash to the classification performance. Also squeezing the classification model with entire raw features can create a bottleneck to the classification performance. Thus, there is necessity for selecting appropriate features for training the classifier. In this proposed, a system is proposed that can use multiple channel real-time EEG signals to predict the onset of an epileptic seizure. The system is given a select number of EEG channels as input and reports back the corresponding epileptic seizure state at every second and the Hybrid Artificial Neural Network with Support Vector Machine (HANNSVM) based classifications are done as a simulation of real-time dynamic predictions and are dependent upon past predictions that were made. As a result, the sensitivity must be controlled such that seizures aren’t predicted more often than they actually occur. Statistical analysis of accuracy values and computational time portrays that the proposed schemes provide compromising results over existent methods.

Keywords: Computer aided diagnosis, K-nearest neighbor, artificial neural network, electroencephalography, Internet of Things, support vector machine, brain modeling feature exraction