Image Data Reduction in Coronary Angiography -
Problems and Solutions
Ruediger Brennecke PhD FESC (Mainz, Germany)
Richard Kerensky MD FACC (Gainesville, Florida)
1. The need for image compression
2. What is lossless image compression and where is it used?
3. What is lossy image compression?
4. Lossless and lossy modes of image data compression in the DICOM3 standard
5. What do lossy images look like?
6. What are the goals of the International Compression Study?
6.1 Project I: The impact of lossy compression on identification of angiographic features
6.2 Project II: Quantitative coronary arteriography (QCA) in decompressed images
6.3 Project III: Side-by-side comparison of original and decompressed cine runs
7. Status of the study
1. The need for image compression
|The change from the cine film to digital methods of image exchange
and archival is primarily motivated by the ease and flexibility of handling
digital image information instead of the film media. While preparing this
step and developing standards for digital image communication, one has
to make absolutely sure that also the image quality of coronary angiograms
and ventriculograms is maintained or improved. Similar requirements exist
also in echocardiography.
Regarding image quality, the most critical step in going from the analog world (cine film or high definition live video in the catheterization laboratory) to the digital world is the digitization of the signals. For this step, the basic requirement of maintaining image quality is easily translated into two basic quantitative parameters:
|Spatial resolution||4 linepairs/mm||1024*1024 pixels|
|Data capacity per image||1 Megabyte (MByte)|
|Data rate||30 images per seconds||30 MByte per second|
|Data capacity per patient exam||2,400 images||2,400 MByte|
|Media||one film||four CD-R|
|Data for 10 years||30,000 films||120,000 CD-R|
Table 1 Scenario for replacement of cine film by digital imaging
with high resolution
|From Table 1 we see, that in this scenario the enormous data rate of
30 Megabyte per second has to be supported. This is much faster than even
advanced ATM networks (offering less than 20 Mbyte/s or 160 Mbit/s). Looking
for existing off-line media, real-time display from CD-R would require
a CD-R player with a data rate of 200X, while the fastest players available
presently deliver 50X (1X stands for a data rate of 150 KByte per second).
The total amount of data or the ‘data capacity’ required in this scenario
is even more frightening (see Table 1).
Computer technology, however, provides flexible principles for processing large amounts of information. Among the algorithms available is image data reduction or ‘image compression’. The principal approach in data compression is the reduction of the amount of image data (bits) while preserving information (image details). This technology is a key enabling factor in many imaging and multimedia concepts outside of medicine. So one has to ask if cardiology really will have to cope with these enormous and totally uncommon requirements concerning digital data rates and digital data capacity (Table 1), or if image compression can also be applied without problems in cardiac imaging.
At a closer look one observes that ad hoc approaches to image data compression have been applied in most digital imaging systems for the catheterization laboratory all the time. An example is recording the x-ray images with a smaller matrix of just 512 by 512 pixels (instead of the 1024 by 1024 pixel matrix often applied for real-time displays). In order to objectively assess these and other techniques of image data compression, some systematic knowledge of the tradeoffs implied in different modes of image data reduction is mandatory.
2. What is lossless image compression and where
is it used?
|When hearing that image data are reduced, one could expect that automatically
also the image quality will be reduced. A loss of information is, however,
totally avoided in lossless compression, where image data
are reduced while image information is totally preserved.
A simple example demonstrates one of the strategies applied. Let us assume that in one horizontal line of an image the following sequence of gray levels is encountered when starting from the leftmost pixel of that line and going to the right:
212 214 220 222 216 212 212 214 ...
These gray levels are usually stored as 8-bit-numbers (1Byte). Obviously much smaller numbers or ‘codes’ are involved if one transfers only the first value directly, followed by the differences to the preceding gray levels:
+212 +2 +6 +2 -6 -4 0 +2 ....
This strategy of data reduction is called ‘predictive encoding’, since we use the gray level of each pixel to predict the gray value of its right neighbor. Only the small deviation from this prediction is stored. This is a first step of lossless data reduction. Its effect is to change the statistics of the image signal drastically: typically 80% of the pixels in the resulting ‘difference image’ will now require just 8 graylevels (3 bits plus sign). Of course, we can still reproduce the original gray level values from these reduced data without any error if we only know the rule that was applied when generating the sequence.
Statistical encoding is another important approach to lossless data reduction. This term sounds very complex, but a similar trick in information coding had already been used by the famous American inventor Samuel Morse more than 150 years ago for his electromagnetic telegraph. A frequently occurring letter such as ‘e’ is transmitted as a single dot ‘ . ’, while an infrequent ‘x’ requires four Morse symbols ‘ - . . - ’. In this way the mean data rate required to transmit an English text is decreased as compared to a solution where each letter of the alphabet is coded with the same number of basic symbols. Accordingly in image transmission, short code words or bit sequences (one to four bits) will be used for frequently occurring small gray level differences (0, +1, -1, +2, -2 etc.), while long code words are used for the large differences (for instance the 212 in our example) with their very infrequent occurrence.
Statistical encoding can be especially successful if the gray level statistics of the images has already been changed by predictive coding. The overall result is redundancy reduction, that is reduction of the reiteration of the same bit patterns in the data. Of course, when reading the reduced image data, these processes can be performed in reverse order without any error and thus the original image is recovered. Lossless compression is therefore also called reversible compression. Data compression factors (number of bits required for uncompressed image data divided by number of bits for compressed image data) of 2 to nearly 4 can be attained by reversible compression. A poster (P1672) at this congress will present detailed data on the compression factors attainable.
In our example, for instance the data rate required for real-time cine display from a CD-R would be reduced by lossless compression from 200X to maybe 80X. This would still be extremely high. Therefore one needs to have also a closer look at methods of data compression that provide ‘lossy’ compression while promising higher compression factors.
3. What is lossy image compression?
|Lossy data compression has of course a strong negative connotation
and sometimes it is doubted quite emotionally that it is at all applicable
in medical imaging. However, please note that many imaging systems for
angiography primarily acquire images as a 1024*1024 pixel matrix while
they transfer only 512*512 pixels per image to local storage and to exchange
media. In this step, part of the image information is irreversibly lost.
So this is an example for a method of data compression that is lossy but
that provides a digital image recording format that is presently widely
accepted in cardiology. Therefore, instead of banning lossy compression
in general, we should discuss objective criteria for the acceptability
of specific methods of lossy data compression in coronary angiography.
We all know different strategies for tolerable lossy data reduction also from daily life. For instance, nobody will read all the information offered in a newspaper, so the overall process of information distribution by newspapers is an example of lossy information handling. The first step in this traditional type of ‘information processing’ occurs when the editors divide the incoming events into groups such as world politics, economy, local affairs, and sports. Each of these groups is presented on one (or several) specific pages of the newspaper. Moreover, on each of these pages, large headings draw the attention to those topics that are most important. Note that this first step of the overall process is essentially lossless.
It is this grouping of semantically similar information on certain pages (e.g. sports) which greatly simplifies selective ‘lossy’ data reduction as performed by the reader in a second step. While scanning the contents of the newspaper, he or she can concentrate on those few pages that are especially relevant to him or her (for instance politics). Time is saved by browsing very quickly through the other pages - those that are less relevant to the reader - and reading on those pages mainly the articles with the largest headings. Since the reader will experience a loss of information as compared to the total information offered, this step must be considered as ‘lossy data compression’. But since the information lost is not especially relevant to the reader, he or she will tolerate this and will be glad to use the limited time available for reading the most relevant information. Since the information lost in this process is systematically selected by the reader to be from the less relevant pages of the newspaper, this everyday type of lossy data compression can also be called ‘irrelevancy reduction’.
A trivial example for lossy compression of image data is selecting only 4 to 10 of the most relevant images for exposing them to a multiformat film that is to be sent to a referring physician. This is an interactive strategy of irrelevancy reduction. Usually, however, one applies an automatic algorithm using essentially the same two-step strategy of ‘irrelevancy reduction’ as described in the newspaper example above. For instance, in transform encoding one performs for each image of the cine run a mathematical transformation that is similar to the Fourier transform thus separating image information on gradual spatial variation of brightness (regions of essentially constant brightness) from information with faster variation of brightness at edges of the image (compare: the grouping by the editor of news according to the classes of contents). In the next step, the information on slower changes is transmitted essentially lossless (compare: careful reading of highly relevant pages in the newspaper), but information on faster local changes is communicated with lower accuracy (compare: looking only at the large headings on the less relevant pages). In image data reduction, this second step is called quantization. Since this quantization step cannot be reversed when decompressing the data, the overall compression is ‘lossy’ or ‘irreversible’.
Usually the main overall effect of transform encoding plus quantization is that small structures or ‘edges’ having low contrast are supressed. The applicability of this strategy is based on the experience that low-contrast edges usually do not contribute information that is important for image interpretation by the human visual system. Larger edge signals are essentially preserved in the quantization process.
Table 2 summarizes some of the characteristics of lossless and lossy image compression techniques as described in the previous paragraphs.
|Data compression strategy||lossless / reversible||lossy / irreversible|
- based on:
|statistics of data||meaning of information|
- results in:
|redundancy reduction||irrelevancy reduction|
|Key methods||predictive encoding plus
|transform encoding plus
|Examples for implementation||lossless JPEG mode||lossy JPEG mode|
|Compression factor (typical)||2 to 3||6 to 12|
Table 2 Overview on principal strategies and methods in lossless
and lossy (right) image compression.
4. Lossless and lossy modes of image data compression
in the DICOM3 standard
|The DICOM standard contains several options for lossless and lossy
image compression. For the existing basic X-ray angiographic application
profile, only a lossless method of data compression has been explicitly
selected. This lossless process corresponds to the two-step lossless process
(predictive coding followed by statistical encoding) described above.
As mentioned before, a lossy method of image data compression by a factor of 4 is implicit in the DICOM application profile mentioned above since it defines the image matrix as 512*512 pixels with 8 bit gray level resolution, while X- ray-video systems in the catheterization laboratory are often able to provide a resolution of more than 1000*1000 pixels. A corresponding loss of small detail resolution seems to be compensated to some part by upscanning the image data to the 1024*1024 pixel format for review and by digital edge-enhancement in order to amplify the contrast of vessel structures.
A more explicit use of lossy methods of data reduction has not yet been
included into the application profile. Before applying lossy compression
one has of course to carefully optimize the process of irrelevancy reduction
so that it retains all medically relevant information. More precisely,
one has to perform the following steps:
B) Definition of the relevant information content of the images (this can differ for different clinical tasks such as primary therapeutic decision making as compared to reviewing a case on the ward)
C) after selecting this strategy and defining the relevant information
content, one has to determine the ‘clinically acceptable’ amount of data
compression allowable (that is the data compression factor).
As the principal strategy for lossy data compression (step A. above), one of the lossy modes of the JPEG standard (Joint Photographic Expert Group) has been selected for testing by the ACC/ACR/NEMA Ad Hoc Group. The JPEG standard is widely applied outside of medicine. The lossy modes of the JPEG standard use transform encoding (DCT) as a first step followed by quantization. Note that the lossy JPEG compression is primarily performed on image subregions of typically 8 by 8 pixels, the so called blocks. At high compression factors, this can give rise to artifacts at the borders of the blocks, the so called ‘blockiness’ artifacts typical for JPEG images. Due to the necessity to apply edge-enhancement filters on the compressed images (see above), in cardiac X-ray imaging these artifacts become more visible than in the normal photographic applications of JPEG compression methods. The JPEG standard additionally defines the lossless mode of data reduction described before.
In the absence of agreed standards for lossy image compression in cardiac imaging, some vendors are following a strategy of recording on the CD-R media a lossy data track in addition to the lossless data. The lossless data are displayed when still images are viewed, while the lossy data are shown in dynamic review. This is done even though the ‘clinically acceptable’ compression factor is not yet known. This is called Dual-Mode encoding.
5. What do lossy images look like?
|While of course images coded by lossless techniques do not differ in any detail from the original images, lossy images may differ because of lost details or because of artifacts added in the compression process (for instance JPEG ‘blockiness’ artifacts). In the examples shown, the images are shown after edge enhancement, contrast stretching (‘windowing’) and upscanning (1024*1024) for improved visibility of small structures (including the ‘blockiness’ artifacts). The left image is uncompressed, and relatively high data compression factors have been selected for the two other images (CR= 12 and CR= 24).|
Fig.1 Comparison of original coronary angiogram (left) with two compression
Middle: JPEG data compression by factor of 12, right: factor of 24. See text for details.
6. What are the goals of the International Compression
|The objective of the study is to determine the clinical and diagnostic impact of utilizing lossy JPEG compression in digital coronary angiography, that is to assess whether the use of image compression degrades a physician’s ability to perform a variety of detection and diagnosis tasks. The impetus of the study is based on the Ad Hoc Group’s development towards finalising the x-ray angiographic DICOM standard for medical image exchange (not a standard for archival). The study is designed to determine the minimal compression factor which significantly affects feature detection, diagnostic information content, and the aesthetic appearance of angiographic images. The final design was developed by experts in the field of coronary angiography working in conjunction with technical experts in the field of digital imaging. Three projects that are described below are in progress to assess more specific areas of interest (Projects I, II, III). The study has been commissioned by the ACC/ACR/NEMA Ad Hoc Group and is sponsored by the American College of Cardiology (ACC), the European Society of Cardiology (ESC), and the National Electric Manufacturers’ Association (NEMA).|
6.1 Project I: The impact of lossy compression
on identification of angiographic features
|The purpose of this portion of the study is to determine if there is a level of compression where angiographers make errors in the detection of diagnostic features, that have been identified on uncompressed images. For example, if an uncompressed angiogram clearly displays a dissection, or filling defect, will lossy compression have an effect on the image such that an experienced angiographer will be less likely to detect this feature. Similarly, does lossy compression create artifacts that cause angiographers to detect features that are not seen on original images e.g. the appearance of calcification within a segment not present on original images. At first glance this question might seem to be a relatively straightforward one, however, the details of a study designed to answer these questions are effected by a number of variables including the well-known observer variability in coronary angiography. The final design of the studies took into consideration these confounding variables attempting to minimize or account for each and, therefore hopefully, isolate the effect of compression on diagnostic accuracy. This study will be done using images submitted from 12 high volume catheteriziation laboratories in the United States and Europe. Approximately 420 coronary cine runs (50,000 frames) were submitted for the study. From these 420 angiograms, 100 cine runs showing a variety of features were chosen by an expert panel for the final study. These 100 cine runs will be independently reviewed at different compression ratios by approximately 50 expert angiographers at different centers. Each reviewer will attempt to identify the features listed in Table 3. Complex stenosis is defined as multiple irregularities, overhanging edges, "haziness", or poorly seen edges.|
Table 3 Angiographic features to be detected in the compression study
|For each of the reviewers 1/4 of the images will be uncompressed, 1/4 will be compressed at a ratio of 1:6, 1/4 will be compressed at a ratio of 1:10, and 1/4 will be compressed at a ratio of 1:16. No observer will see a single cine run twice or any single cine run both compressed and uncompressed. Random effects logistic regression will be used to model the probability of a correct interpretation as a function of the level of compression.|
6.2 Project II: Quantitative coronary arteriography
(QCA) in decompressed images
|In Phase II of the study it will be investigated whether image compression and decompression has an effect on the quantitative assessment of vessel sizes by state of the art QCA analytical methods. The procedure will be as follows. From the available image runs, a frame selection will be done according to QCA standards. These selected frames will be compressed with the different compression factors and decompressed, and all the images including the original noncompressed images will be submitted in a blinded manner to the QCA technicians at the Core Laboratory at Leiden University Hospital in the Netherlands. QCA will be carried out with the Cardiovascular Measurement System (CMS). Statistical parameters to be investigated are the systematic differences between compressed/decompressed and the original images to assess whether the compression results in systematic over- or underestimation in the measured vessel dimensions, and the random differences to assess whether the compression leads to increased variability in the measurements.|
6.3 Project III: Side-by-side comparison of original
and decompressed cine runs
|This project presents the same cine runs as described for Project I, but in a side-by-side format, so that the original and the compressed/ decompressed images can be directly compared. The compression factors are again 6, 10 and 16 (in random order). The observer has to decide if there is a difference between both sides. A name for a test of this kind is 'just noticeable difference test' (JND-test). If the observer reports a difference, he or she is asked to define the side with better image quality. Of course, the observers are blinded to the fact on which side the compressed/ decompressed cine run is shown. If a quality difference is reported the observer is asked if a change in therapeutic decision making will result from this change in image presentation.|
7. Status of the study
|The described three study arms have been finalized by September 1998 and presently the resulting three papers are under review for publication.|
|The enormous requirements concerning data rate and storage capacity
of coronary angiograms can in principle be reduced by lossless and lossy
modes of image compression. A lossless mode is already included in the
X-ray angiographic application profile of DICOM3. Lossy modes provide higher
compression factors, but before their implementation within the standard
a study is required to assure that medically relevant information is not
lost in this process.
In the international study described, compressed/decompressed cine runs
will be presented both for the clinical decision making test (Project I)
and for the test of just noticeable differences (Project III). Both tests
have their advantages and limitations so that it seems to be necessary
to combine them. Project II will be added since QCA is a proven method
and can provide immediate quantitative results on the applicability of
lossy data compression. The effect of varying JPEG data compression levels
on the quantitative assessment of the degree of stenosis in coronary angiography.
For some of the results, see the contribution by Joan C. Tuinenburget al.
"The effect of varying JPEG data compression levels on the quantitative
assessment of the degree of stenosis in coronary angiography".
This text is based on the ACC/ ESC DICOM Primer, which is available at the DICOM booth of the ESC.
Address for correspondence:
Ruediger Brennecke PhD FESC
2. Medical Clinic
Mainz University Hospital
D 55101 Mainz Germany
Fax: +49 6131 17 6669