MRI Medical Image Fusion Using Gradient Based Discrete Wavelet Transform

Abstract

Medical diagnosis and treatment for brain pathology can be improved by fusing the medical images. This paper proposes a strategy for fusion of different types of MRI brain images like T1-weighted (T1w), T2-weighted (T2w) and FLAIR images. MRI is most widely and commonly used scanning technique especially for detecting all kinds of brain tumor and stroke. Here a gradient based discrete wavelet transform is proposed for fusing the different types of MRI images of same patients for better diagnosis of brain pathology. The first step includes the DWT decomposition for getting the fused image 1. The second step involves gradient measure to the fused image 1 for getting the final fused image with good clarity for the abnormality detection. Several image fusion techniques have been proposed so for, few among them is considered here for performance analysis. The methodology considered here include discrete wavelet transform (DWT), principal component analysis (PCA) and dual tree complex wavelet transform (DTCWT). The performance measure considered here include fusion symmetry, standard deviation, mutual information, average gradient and entropy. Experimental results suggest that T2w and FLAIR fusion result is better for tumor detection using proposed method compared to other fusion results like T1w and T2w fusion, T1w and FLAIR fusion. In spite of all possible combination of inputs, proposed method is giving 90% accuracy results compared to other techniques.