Classification of brain tumor image based on High Grade and Low Grade using CNN with LSTM

Abstract

Magnetic resonance imaging (MRI) is a normally available imaging method for the
examination of these cancers or tumors, but the vast amount of data provided by MRI prohibits
within a reasonable time restricting by manual segmentation based on the procedure of precise
quantifiable measurements in medical practice. Gliomas are the most prevalent and destructive
brain tumors, resulting at its highest grade in a very small life of expectancy. In general, gliomas
can be classified into High (HG) and Low (LG) Grades. The precise diagnosis of the glioma
allows to define the progression of the infection and to choose the therapeutic plan. Although
classification of medical images has attained remarkable success using a Convolution Neural
Networks (CNNs), but it is dispute for finding out the medical image exactly over 3D. This paper
has proposed to evaluate the CNN with Long Short Term Memory (LSTM) for analyzing glioma
of LG and HG 3D brain tumor image. In order to classify the volume of 3D brain tumors, the
pre-trained VGG-16 gets extracted and served into LSTM for learning higher level features
while feature extraction. Based on the results produced by this method, the best method is
suggested for brain tumor identification and further, for analysis of MRI images.