Clinical evaluation of JPEG2000 compression for digital mammography

Abstract

Medical images, such as computed radiography (CR), and digital mammographic images will require large storage facilities and long transmission times for picture archiving and communications system (PACS) implementation. American College of Radiology and National Equipment Manufacturers Association (ACR/NEMA) group is planning to adopt a JPEG2000 compression algorithm in digital imaging and communications in medicine (DICOM) standard to better utilize medical images. The purpose of this study was to evaluate the compression ratios of JPEG2000 for digital mammographic images using peak signal-to-noise ratio (PSNR), receiver operating characteristic (ROC) analysis, and the -test. The traditional statistical quality measures such as PSNR, which is a commonly used measure for the evaluation of reconstructed images, measures how the reconstructed image differs from the original by making pixel-by-pixel comparisons. The ability to accurately discriminate diseased cases from normal cases is evaluated using ROC curve analysis. ROC curves can be used to compare the diagnostic performance of two or more reconstructed images. The test can be also used to evaluate the subjective image quality of reconstructed images. The results of the test suggested that the possible compression ratios using JPEG2000 for digital mammographic images may be as much as 15 : 1 without visual loss or with preserving significant medical information at a confidence level of 99%, although both PSNR and ROC analyses suggest as much as 80 : 1 compression ratio can be achieved without affecting clinical diagnostic performance.

Perceived Sufficiency of Full-Field Digital Mammograms With and Without Irreversible Image Data Compression for Comparison with Next-Year Mammograms

“…Given the advantages of image compression for the transmission and storage of digital mammography files, there are very few studies that evaluate the compression ratios of mammograms 3–5. In a reader study with five experienced radiologists, Penedo and colleagues irreversibly compressed mammograms at 40:1 and 80:1 ratios using both the JPEG 2000 and object-based SPHIT methods 4.…”

“…Integration of FFDM systems into preexisting picture archiving and communications systems (PACS)—for the purposes of achieving the workplace efficiencies that can derive from digitized transmission, storage, and manipulation of images—has been difficult. File sizes associated with digital mammographic images after lossless compression are extremely large 3–6. Depending on the type of FFDM system used, individual lossless-compressed digital mammograms range in size from 2.5 to 16 MB.…”

Interpolation-based super-resolution reconstruction: effects of slice thickness

“…The PSNR in decibels (dB) between the original image and the registered image is defined by 22 PSNR ¼ 20 × log 10 MAX RMSE ;…”

A survey on evaluation methods for image interpolation

“…The PSNR metric can be also used which is closely related to RMSE. The PSNR in decibels (dB) between the original image I (x, y) and the interpolated imageÎ (x, y) is given by [27] PSNR = 20 × log 10 MAX I RMSE ,…”