Brain MRI classification using discrete wavelet transform and bag-of-words

Automatic CAD system able to detect correctly the unhealthy brain in magnetic resonance imaging (MRI) scanning is represented in this paper. The new system exploited Discrete Wavelet Transform (DWT) and Bag-of-Words (BoW) to extract image features. Support vector machine (SVM) was used in classification step. We employed 256×256 images from three datasets (DS-66, DS-160, DS-255) provided by Harvard Medical School, to evaluate our method. 10*k-fold stratified Cross Validation (CV) technique was applied to validate the system performance. The Accuracy reached respectively 100%, 100%, and 99.61% for DS-66, DS-160, and DS-255 datasets. The overall computation time is about 0.027 s for each MR image. A comparative study with several works showed efficiency and robustness of our scheme.

I. Introduction

MR brain images diagnosis represents a critical task. Wrong diagnosis can provide severe results for patient healthy. Human brain is characterized by a structure complexity, which makes its analysis very hard. In addition, the analysis and viewing expert are very limited compared to the large amount of MR images. Analyzing these images manually has several disadvantages as time-consuming. Moreover, it is very exhausting to keep a high level of concentration during the classification that gives rise to increase the false hit rate. Therefore, an automatic system is required to analyze MR images, where CAD is a promising solution.