期刊名称：International Journal of Electrical and Computer Engineering
电子版ISSN：2088-8708
出版年度：2021
卷号：11
期号：5
页码：3964-3976
DOI：10.11591/ijece.v11i5.pp3964-3976
语种：English
出版社：Institute of Advanced Engineering and Science (IAES)
摘要：In the compression of medical images, region of interest (ROI) based techniques seem to be promising, as they can result in high compression ratios while maintaining the quality of region of diagnostic importance, the ROI, when image is reconstructed. In this article, we propose a set-up for compression of brain magnetic resonance imaging (MRI) images based on automatic extraction of tumor. Our approach is to first separate the tumor, the ROI in our case, from brain image, using support vector machine (SVM) classification and region extraction step. Then, tumor region (ROI) is compressed using Arithmetic coding, a lossless compression technique. The non-tumorous region, non-region of interest (NROI), is compressed using a lossy compression technique formed by a combination of discrete wavelet transform (DWT), set partitioning in hierarchical trees (SPIHT) and arithmetic coding (AC). The classification performance parameters, like, dice coefficient, sensitivity, positive predictive value and accuracy are tabulated. In the case of compression, we report, performance parameters like mean square error and peak signal to noise ratio for a given set of bits per pixel (bpp) values. We found that the compression scheme considered in our setup gives promising results as compared to other schemes.
关键词：discrete wavelet transform;magnetic resonance imaging;PSNR;region of interest;SPIHT;support vector machine
