Automatic tumor lesion detection and segmentation using modified winnow algorithm

Abstract

Automated recognition of brain tumors in magnetic resonance images (MRI) is a difficult procedure due to the variability and complexity of the location, size, shape, and texture of these lesions. Due to intensity similarities between brain lesions and normal tissues, most approaches make use of multi-spectral MRI images. However, the time, cost, and data process restrictions for collecting multi-spectral MRI necessitate developing a lesion detection and segmentation approach that can detect lesions using a single anatomical MRI image. In this paper, we present a fully automatic system, which is able to detect the MRI images that include tumor and to segment the tumor area. Fully anisotropic complex Morlet transform, and dual tree complex wavelet transform are introduced for tumor textural characterization. Perhaps most importantly, we propose a novel feature selection technique that is based on regularized Winnow algorithm. An active contour model implemented with selective binary and Gaussian filtering regularized level set (SBGFRLS) is used for final segmentation step. The experimental results on both simulated and real brain MRI data prove the efficacy of our technique in successfully segmenting brain tumor tissues with high accuracy and low computational complexity.