Medical Image Archiving System Implementation with Lossless Region of Interest and Optical Character Recognition

Abstract

Digital medical images have been widely used in all stages of healthcare. It has a vital role to transfer and to store digital medical images for medical experts and patients. Since the large file sizes and storage space requirements, image compression has become a necessity. Instead of compressing the entire image, it is an option to compress the region of interest (ROI). Applying lossless methods to the whole image does not provide a sufficient advantage, however, when lossy techniques are used, the vital information of the medical image may be lost. In this study, a novel medical image archiving system implementation based on ROI and Optical Character Recognition (OCR) is proposed. Besides, a new dynamic file structure was used that was specially designed to produce better compression ratio and performance. The medical image is separated into the ROI and the non-ROI parts. JPEG-LS, a lossless compression algorithm, is applied to the ROI segment of the medical image. The OCR and Huffman coding algorithm is used for the non-ROI part of the image. The proposed method was evaluated using medical images of the actual patient and the produced compression ratio for the non-ROI image is between 92.12% and 97.84%. The average difference between the proposed method and the state-of-art in the literature is 83.80% for the non-ROI part. In conclusion, the proposed method provides an integrated solution to the medical image archiving problem.