Automated localization of Epileptic Focus Using Convolutional Neural Network

Cuixia Feng, Hulin Zhao, Jun Zhang, Zhibiao Cheng, Junhai Wen*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

3 Citations (Scopus)

Abstract

Focal cortical dysplasia (FCD) is one of the most common causes of intractable epilepsy. The automatic localization of magnetic resonance (MR) images of epileptic lesions caused by FCD can be performed by using the convolutional neural network (CNN) technology in the field of artificial intelligence, which is helpful for doctors to better diagnose. This study trained a four-layer CNN, including the 8630 learning parameters and three convolutional layers followed by the pooling layer and one full connection layer, and finally, output an image as a disease probability value. The network can classify the MR images with the disease or not, and the accuracy of the final classification of the new data is 92.45%. On the basis of correct classification, the accurate rate of FCD lesions localization can reach 92.86% by using this network.

Original languageEnglish
Title of host publicationBDET 2020 - 2020 2nd International Conference on Big Data Engineering and Technology
PublisherAssociation for Computing Machinery
Pages72-75
Number of pages4
ISBN (Electronic)9781450376839
DOIs
Publication statusPublished - 3 Jan 2020
Event2nd International Conference on Big Data Engineering and Technology, BDET 2020 - Singapore, Singapore
Duration: 3 Jan 20205 Jan 2020

Publication series

NameACM International Conference Proceeding Series

Conference

Conference2nd International Conference on Big Data Engineering and Technology, BDET 2020
Country/TerritorySingapore
CitySingapore
Period3/01/205/01/20

Keywords

  • Convolutional neural network
  • Epilepsy
  • Focal cortical dysplasia
  • Localization
  • MR images

Fingerprint

Dive into the research topics of 'Automated localization of Epileptic Focus Using Convolutional Neural Network'. Together they form a unique fingerprint.

Cite this