Novel Discrete Component Wavelet Transform for detection of cerebrovascular diseases

Abstract

Detection and diagnosis of a disease with a single image can be tedious and difficult for doctors but with the adaptation of medical image fusion, a path for additional improvements can be paved. The objective of this research is to implement different fusion algorithms based on conventional and proposed hybrid techniques. Based on performance metrics it has been observed that the novel method, Discrete Component Wavelet Transform (DCWT) shows remarkable results in comparison to the traditional techniques. As per the enhancement methods, Binarization, Median Filter, and Contrast Stretching have been considered to compare the contrast performance with Contrast Limited Adaptive Histogram Equalization. Certain modifications to each enhancement method were made related to the selection of parameters. Thus, better qualitative and quantitative values were observed in Discrete Component Wavelet Transform. The different attributes were calculated from the fused images which were classified using various machine learning techniques. Maximum accuracy of 97.87% and 95.74% is obtained using Discrete Component Wavelet Transform for Support Vector Machine (SVM) and k Nearest Neighbor (kNN) (k = 4) respectively considering the combination of both features Grey Level Difference Statistics and shape.