Efficient ROI-based compression of mammography images

Abstract

Medical imaging and telemedicine are progressively employed on a large scale these days. The utilization of lossless compression techniques for medical image storage and transmission may not provide significant benefits in terms of data reduction. On the other hand, the utilization of lossy compression techniques may lead to loss of essential data, which affects the diagnosis process. The critical data in the region of interest (ROI) of a medical image should be compressed with high-quality compression with no data lost or destroyed, regardless of the remaining parts of the medical image (non-ROI). Generally, mammography is used for investigating breast malignancies and localization of small tumors. Therefore, to increase the compression efficiency of mammogram images without losing any essential data, we present a hybrid technique based on lossless compression for the critical data in the ROI and lossy compression for the remaining parts of the medical image (non-ROI). In this work, edge-directed prediction lossless compression is adopted for the ROI, while fractal lossy compression is proposed for the non-ROI of mammogram images. Mammogram images from the mammographic image analysis society (MIAS) database are used to test the proposed hybrid technique. Encoding time, decoded image quality in terms of peak signal-to-noise ratio (PSNR), and compression ratio are the metrics used to assess the quality of compression. The obtained results prove that the proposed technique achieves a high compression ratio, while maintaining an acceptable quality of the reconstructed images compared with other recent mammogram image compression techniques.