Analysis of Classification Algorithms for brain tumor detection

EasyChair Preprint 1810

Abstract

Image classification of brain tumor MRI data is very critical and decisive work. In this paper, we implemented an algorithm to classify images of brain tumors from the data of those who do not have tumors using two efficient well-known classification techniques namely KNN and SVM.  The pre-processing techniques include Gaussian Filtering and Adaptive median filtering were implemented for both classifiers, by extracting the HoG features. When an impulse noise is dominant in such an image, a normal adaptive median filter is well-known way to remove the impulse noise. For both classifiers the accuracy has been determined, the results vividly show that SVM performs far better than K-NN with an accuracy of 81.1%, whereas the accuracy obtained for K-NN was 57.64%. The entire analysis is done using MATLAB 2019a software.

Keyphrases: Gaussian, HOG, MRI, Median Filter, Preprocessing Filter, SVM, feature extraction, k-NN

@booklet{EasyChair:1810,
  author    = {Sai Sasank Varthakavi and Devisetty Rohith Prasanna Babu and M B Amsitha and Sasi Jyothirmai Bonu},
  title     = {Analysis of Classification Algorithms for brain tumor detection},
  howpublished = {EasyChair Preprint 1810},
  year      = {EasyChair, 2019}}