There is much confusion regarding compression in digital radiography. Much of the confusion comes from otherwise well meaning salesmen disparaging other vendor’s equipment and horrible JPEG artifacts that are common in digital photography. This tutorial is aimed at clearing up misconceptions with regard to compression as it applies to medical images.

The Need for Compression:

The imaging dataset for digital radiographic images is large. Many studies can reach sizes of 20-50MB. Anyone who has ever attempted to e-mail a 20MB attachment and had it rejected by the receiving server knows 20MB is an unwieldy file size for teleradiology applications. (As an aside this is one reason why E-mail is entirely unacceptable for teleradiology.) Even with broadband internet connections, these large files can impede teleradiology and remote image archival.

If used appropriately, image compression can be effective method of managing these large data sets. Image compression greatly facilitates teleradiology and dramatically reduces the requirements (and therefore cost) of archiving medical images.

What is compression?

Data compression on digital images reduces redundancy of the image data in order to be able to store or transmit data in an efficient form. Reducing redundancy decreases the size of image files. The degree of compression applied to an image is called the “compression ratio.”

A compression ratio is the ratio between the uncompressed size to the compressed size. Compression ratios are generally written as 5:1 or 40:1. A compression method that compresses a 10MB file to 2MB has a compression ratio of 10/2 = 5, often notated as an explicit ratio, 5:1 (read "five to 1”)¹ .

What compression methods are available for medical images?

There are many different compression methods available to be applied to medical images. The two broad categories of image compression are lossy (also known as irreversible) and lossless compression methods.

Lossless compression methods use statistical redundancy within in an image to allow for perfect image reconstructions. However, the decrease in file size with lossless compression is only 2:1-4:1 for medical images². This small decrease in file size does not overcome the issue of large file sizes and teleradiology applications. RLE and JPEG-LS are examples of lossless compression methods.

Lossy compression methods can decrease file sizes by very large ratios but do not perfectly reproduce the original image. Lossy compression is used if some loss of fidelity is acceptable. Lossy data compression algorithms introduce relatively minor differences and represent the picture, video, or audio using fewer bits³. JPEG and JPEG2000 are the two most common methods of compression that are applied to medical images.

JPEG is good for general purposes, however, at high compression ratios it produces dramatic and characteristic “blockiness” artifacts⁴. This limits the amount of compression that can be applied to medical images before artifacts degrade the images.

JPEG2000 compression is a form of compression called wavelet compression or DWT (discrete wavelet transform) compression. Wavelet compression creates a compressed image by treating the images as a signal or wave. A technical discussion of how wavelet compression works can be found here . The advantage of wavelet based compression methods such as JPEG2000 is that higher compression ratios can be applied while introducing less artifact to the image⁴.

How does irreversible compression affect images?

At appropriate compression ratios, the change to an image is imperceptible (sometimes called visually lossless). With increased compression, statistical noise is removed from the images. These “denoised” images are generally preferred by observers and may actually improve diagnostic accuracy⁴. As compression levels are increased, blurring of the image becomes apparent. As the compression is increased further characteristic compression artifacts are encountered².

What does the research say about image compression? Will I miss lesions if I read on compressed images?

The worry with applying irreversible compression to medical images is that a lesion will be missed because subtle findings will not be represented in the compressed image. Research, however, has shown that if appropriate levels of compression are applied to an image, there is no loss of diagnostically significant information. The most clinically relevant studies about image compression focus on objective measurement of diagnostic accuracy using blinded evaluation methods⁴.

Recently, a broad review of the literature and legal implications of image compression was performed by Canada Health Infoway which is a nonprofit organization that promotes the adoption of electronic health information systems in Canada⁵.

The conclusion of a literature review of over 120 papers determined that “based on scientific studies, irreversible compression is a clinically acceptable option for the compression of medical images.⁵”

Are there any legal implications relative to making a diagnosis off if images with appropriate levels of compression?

The legal review from the CHI study concluded that would not be unreasonable, and consequently could not trigger liability under the cause of action in negligence, if such reliance has the support of at least a respectable minority of radiologists.

In short, within Canada, if the professional body adopts irreversible compression as a standard of practice, and if institutions deem the use of irreversible compression provides economic and practical operational benefits as well as contributes to better quality care, the exposure to legal risk is no greater than with current practice. The key presumption, however, is that the use of irreversible compression does not impact the visual quality of an image⁵.”

“The regulatory review, also conducted by Bull, Housser and Tupper LLP, concluded that none of the regulatory bodies in Canada, UK, European Union, Australia and the United tates make any statements preventing or endorsing the use of irreversible compression⁵.”

The American College of Radiology does not currently propose a standard or suggest recommended compression but leaves the topic of compression to the physician. As stated in the ACR Technical Standard for Teleradiology⁶:

“Data compression may be performed to facilitate transmission and storage. Several methods, including both reversible and irreversible techniques, may be used under the direction of a qualified physician, with no reduction in clinical diagnostic image quality. The types and ratios of compression used for different imaging studies transmitted and stored by the system should be periodically reviewed by the responsible physician to ensure appropriate clinical image quality.”

What level of compression is acceptable?

The amount of compression that can be applied to an image will vary with the compression method, modality, and organ system being examined. The CAH study recommended the following:

When purchasing a digital radiography system, the ability to send compressed images and view compressed images is a value added feature!

DICOM provides for compression in a vendor independent manner. Every DICOM device supports a number of services like storage and print management. Imaging devices support one or more coding schemes for these services. Theses coding schemes or transfer syntaxes have names and numbers associated with them. For example, Implicit VR Little Endian (default) = 1.2.840.10008.1.2 and Jpeg lossless = 1.2.840.10008.1.2.4.57 are two transfer syntaxes that are commonly encountered.

If you are going to include teleradiology in your practice or want to do an offsite backup, it would be worthwhile to ask your vendor if they support the nonproprietary DICOM compression transfer syntaxes⁷ such as JPEG2000 1.2.840.10008.1.2.4.91 or “0.91” or Jpeg Lossy compression 1.2.840.10008.1.2.4.51 or “0.51”

There are several proprietary compression methods that are also used in veterinary digital radiography systems. Image Channel, Real Time Image, and X-scan are three of these. These compression methods are certainly useful and diagnostic, but unlike the nonproprietary method of 0.91 JPEG2000 wavelet compression in DICOM, these proprietary systems mean that your viewer and server are “linked” and function as a unit.

If your digital radiography system does not support compression for image transmission, there are a number of third party applications that you can use to easily add this functionality to your system

Practically speaking – what does this all mean to me?

The bottom line on all of this is that:

1. When used appropriately, irreversible compression such as JPEG2000 wavelet compression is an acceptable method of decreasing teleradiology transfer times and storage requirements in veterinary medicine.
2. The improvements in transfer times in teleradiology can be dramatic. In our practice, noncompressed DICOM images may take 4 minutes to send using standard DICOM transfer protocols but with the DICOM 0.91 transfer syntax, images will take about 8-10 seconds to transfer and there is no perceptible loss of image quality.
3. The time to consider data compression is BEFORE you purchase your digital radiography system. Ask your vendor if they offer you non-proprietary ways to send compressed images for teleradiology. If not, there are inexpensive third party applications that can be helpful in this regard.

References:

1. http://en.wikipedia.org/wiki/Data_compression_ratio
2. Erickson, B. Irreversible Compression of Medical Images (SCAR White Paper). November 2000.
3. http://en.wikipedia.org/wiki/Image_compression
4. Koff, DA. An Overviwe of Digital Compression of Medical Images: Can We Use Lossy Image Compression in Radiology? Can Assoc Radiol J 2006; 57(4):211-217.
5. Will the Use of Irreversible Compression become a Standard of Practice? Winter 2006. SCAR News. Volume 18. Issue 1. http://www.siimweb.org/WorkArea/showcontent.aspx?id=1126
6. 6. ACR Technical Standard for Teleradiology. Revised ed. American College of Radiology; 2002.
7. http://medical.nema.org