Neural
Network with Bee Colony Optimization for MRI Brain Cancer Image Classification
Sathya
Subramaniam and Manavalan
Radhakrishnan
Department of Computer Science and applications, Periyar
University, India
Abstract:
Brain tumor is one of the foremost causes for the increase in mortality
among children and adults. Computer visions are being used by doctors to
analysis and diagnose the medical problems. Magnetic Resonance Imaging (MRI) is a medical
imaging technique, which is used to visualize internal structures of MRI
brain images for analyzing normal and abnormal prototypes of brain while
diagnosing. It is a non-invasive method to take picture of brain and the
surrounding images. Image processing techniques are used to extract meaningful
information from medical images for the purpose of diagnosis and prognosis. Raw
MRI brain images are not suitable for processing and analysis since noise and
low contrast affect the quality of the MRI images. The classification of MRI
brain images is emphasized in this paper for cancer diagnosis. It can consist
of four steps: Pre-processing, identification of region of interest, feature
extraction and classification. For improving quality of the image, partial
differential equations method is proposed and its result is compared with other
methods such as block analysis method, opening by reconstruction method and histogram
equalization method using statistical parameters such as carrier signal to
ratio, peak signal-to-ratio, structural similarity index measure, figure of
merit, mean square error. The enhanced image is converted into bi-level image,
which is utilized for sharpening the regions and filling the gaps in the
binarized image using morphological operators. Region of Interest (ROI) is
identified by applying region growing method for extorting the five features.
The classification is performed based on the extracted image feature to
determine whether the brain image is normal or abnormal and it is also,
introduced hybridization of Neural Network (NN) with bee colony optimization
for the classification and estimation of cancer affect on given MRI image. The
performance of the proposed classifier is compared with traditional NN classifier
using statistical measures such as sensitivity, specificity and accuracy. The
experiment is conducted over 100 MRI brain images.
keywords: MRI images, NN,
bee colony, PDE, biological analysis, feature extraction.
Received February 17, 2013; accepted
October 24, 2014