INTERNATIONAL JOURNAL OF SCIENTIFIC DEVELOPMENT AND RESEARCH International Peer Reviewed & Refereed Journals, Open Access Journal ISSN Approved Journal No: 2455-2631 | Impact factor: 8.15 | ESTD Year: 2016
open access , Peer-reviewed, and Refereed Journals, Impact factor 8.15
A CONVOLUTION NEURAL NETWORK ALGORITHM FOR BRAIN TUMOR IMAGE SEGMENTATION
Authors Name:
Priya K
, Dr.O.Saraniya
Unique Id:
IJSDR1703028
Published In:
Volume 2 Issue 3, March-2017
Abstract:
Gliomas are the most common occuring condition in brain tumor. It is aggressive in nature leading to a very short life span and predicting the presence of high grade glioma has many real time difficulties. Even under treatment patients do not survive an average of more than 14 years, this situation requires a technique that can predict the presence of tumor at the early stage. For these resons we propose an automatic segmentation method based on convolution neural network in MRI (Magnetic Resonance Imaging) images. Convolution Neural Network (CNN) is an artificial neural network in which several neurons are connected in the way of visual cortex. Convolution layers have fewer weights which make CNN easy to train. In this paper, a sequence of trained data is fed to the pre-processor which estimate the bias field correction. This image is normalized and standard deviation is calculated. Using fuzzy clustering the presence of tumor cell is determined by patch extraction from different patches.. In CNN, Back propagation is used to calculate the gradient of loss function with respect to all weights in the network. Pooling and data augmentation technique are used in CNN.
Keywords:
Magnetic Resonance Imaging, Glioma, Brain
Cite Article:
"A CONVOLUTION NEURAL NETWORK ALGORITHM FOR BRAIN TUMOR IMAGE SEGMENTATION", International Journal of Science & Engineering Development Research (www.ijsdr.org), ISSN:2455-2631, Vol.2, Issue 3, page no.183 - 188, March-2017, Available :http://www.ijsdr.org/papers/IJSDR1703028.pdf
Downloads:
000337067
Publication Details:
Published Paper ID: IJSDR1703028
Registration ID:170101
Published In: Volume 2 Issue 3, March-2017
DOI (Digital Object Identifier):
Page No: 183 - 188
Publisher: IJSDR | www.ijsdr.org
ISSN Number: 2455-2631
Facebook Twitter Instagram LinkedIn