International Journal of Cardiology 60 (1997) 195-200

IMPACT OF VARIOUS COMPRESSION RATES ON INTERPRETATION OF DIGITAL CORONARY ANGIOGRAMS.

Sigmund Silber MD, Rolf Dörr MD, Gunnar Zindler MD, Holger Mühling MD, Thomas Diebel RN

Cardiac Catheterization Laboratories, Dr. Müller Hospital, Am Isarkanal 36, Munich, Germany

Abstract:

According to the ACC/ACR/NEMA/ESC-guidelines, digital techniques should replace cinefilm for coronary angiography. The ad hoc group of experts recent-ly chose CD-R (CD recordable) as trans-port media and the JPEG standard for image compression. To avoid a possible loss of image quality, the guidelines allow a maximal data compression of only 2:1. This, how-ever, leads to a considerable limitation: coronary angiograms cannot be viewed in real-time directly from CD.
Since the possible influence of higher compression rates on image quality of co-ronary angiograms had not been investigated in a controlled study, we evaluated 8 va--rious compression rates (ranging from 5:1 to 43:1) according to a prospective, randomized and blin-ded protocol. Four independent observers assessed 1440 angiograms using a semiquantitative score. We found that an-gio-grams with a compression rate of 5:1 and 6:1 did not lead to a clinically relevant de-terioration of image quality, whereas 11:1 was still acceptable, but 22:1 becomes un-acceptable.
Since no clinically relevant loss of information at a compression rate of 6:1 was experienced in our study, a modification of the ACC/ACJ/NEMA/ESC-guidelines allowing higher com-pression rates should be con-sidered.

Key words: cardiac catheterization; coronary artery disease; computer application; CD-ROM; ACC standard

Introduction

The advantages of the filmless cathlab are obvious: No use of chemistry and subsequent environmental protection, reduction of radiation exposure, simpler hand-ling and finally, reduction of costs [1]. Therefore, replacing cinefilm by digital tech-niques is generally desired.
The ad hoc group of experts achieved a major task by establishing a general technical standard for digital archiving and interinstitutional exchange of coronary angiograms: the guidelines of the ACC (American College of Cardiology), ACR (American College of Radiology) and NEMA (National Electrical Manufacturers Association) [2]. The logical format is the in medicine well established DICOM 3.0-standard (Digital Imaging and Communications in Medicine). The phy-si-cal format is the CD-R (compact disc-recordable). Its write-once-technique warrants a high degree of data security, using the industry standard of "orange book" [3]. CD-R is faster and more robust than digital tapes like DAT, Exabyte or DLT. The capacity of a single CD-R with its 680 MB is sufficient to store 99% of cardiac catheterizations [2].
The technique chosen for data compression is JPEG (Joint Photographic Experts Group). However, the expert group recommended a maximal data compression of 2:1 (lossless compression), directly leading to a considerable limitation of its practicability: even with the up-to-date CD-technology, coronary an-giograms cannot be replayed in real-time directly from the CD-R. But exactly this option of immediately viewing coronary angiograms in real-time (like cinefilm) will ascertain the worldwide acceptance and distribution of CD-R for archiving and exchanging coronary angiograms. Since higher compression rates would allow real-time viewing of coronary angiograms directly from CD-R, but the possible deterioration of image quality using various higher compression rates had not yet been established in a controlled stu-dy with coronary angiograms, we performed the following pro-spective randomized and blinded investigation.

Methods

Diagnostic cardiac catheterization was performed in standard technique after puncture of the right femoral artery; coronary angiograms were obtained with 5F-Judkins-ca-the-ters (Baxter). The contrast agent was identical in all patients (Solutrast™, Byk-Gulden). Twenty four consecutive patients were enrolled (15 with coronary artery di-se-a-se, 2 with valvular disease, 2 with dilative cardiomyopathy and 5 without abnormal fin-dings). The mean age was 58.7 ± 9.3 years, the mean height 170.4 ± 8.9 cm and the mean weight was 80 ± 11.7 kg. LV-EF was 68 ± 11%. The loss-less compressed angiograms (General Electric Advantx-Hiline-Sy-stem, 512x512 pixels, 256 grey levels, Huffman encoding and differential pulse-code modu--lation, DPCM [4]) served as reference.
Angiograms were digitized at 25 frames per second using various JPEG-compression rates with the Me-dia 100™-system (Data Translation, USA) installed in a Power-PC (Macintosh™ (8100/110; 80 MB RAM) and stored on Seagate-Barracuda™ hard disks (SCSI-2 fast disk array, 9 GB) using Remus™ Software from Tri-llium-Research. The Media 100™-system performs a single frame by frame JPEG da-ta compression of 786x576 pixels. We chose 8 various compression rates: 5:1, 6:1, 7:1, 9:1, 11:1, 14:1, 22:1 and 43:1. In blinded and randomized sequences, four experienced observers (A, B, C and D, three invasive car-diologists, one expe-rienced technician) evaluated the angiograms. These real-time movies were con-tin-u-ously displayed and compared with the reference angiogram as bet-ter (1 point), equal (2 points), good but somewhat worse (3 points), accept-able (4 points) or un-accep-table (5 points). The overall assessment included general image quality, image contrast, delineation of stenoses, visability of small ves-sels and collaterals as well as disturbing signals, blurring effects and block artifacts possibly introduced by higher JPEG compression. A total of 1440 assessments had to be made by the observers (45 coronary angiograms, 23 left coronary ar-te-ry, 22 right coronary artery with 8 different compression rates). To exclude the possible influence of different views, all angiograms were evaluated in the 30° RAO projection. The statistical analysis was performed using Scheffé's test. p < 0.05 was regarded as significant.

Figure 1: Mean values for image quality according to the semiquantitative score overall results (for p-values and significances please see Table I).

Results

Figure 1 depicts the overall results: most angiograms were classified as equal or good up to a compression rate of 6:1, followed by a mean classification as good to acceptable. The rate of 43:1 was unacceptable. Analyzing the semi-quan-titative score, compression rates from 5:1 to 11:1 did not show statistically sig-ni-fi-cant- differences among each other (Table I).

Table I: Significance of differences between the various compression rates overall results (*= p<0.05). For mean values please see Figure 1.

The higher compression rates of 14:1, 22:1 and especially 43:1 were significantly wor-se. The first and closest pair of compression rates showing a statistically significant difference was 6:1 vs. 14:1 (p = 0.03, Table I). There was no difference between the assessment of the left or right coronary artery (3.3 ± 0.7 vs. 3.4 ± 0.6, p = 0.27).
Interobserver variability: the mean values of the four observers were 3.3 ± 0.9, 2.9 ± 1.1, 3.3 ± 0.7 and 3.4 ± 0.8. Although these mean values were close, the individual analysis revealed significant differences between individual observers (Figure 2, Table II):

Table II: Significance of differences between the four observers A, B, C, and D (*= p<0.05). For mean values please see Figure 2.

Observer B classified the quality of the angiograms constantly significantly better than A and D, whereas observer C scored significantly better than A, but worse than B. Therefore, viewing the data in analogy to Figure 1, the over-all trend shows a similar tendency with the individual curves, however, vertically shifted (Figure 2).

Figure 2: Inter-observer (A,B,C,D) variability: Mean values according to the semiquantitative score (for p-values and significances please see Table II).

Discussion

Coronary angiography remains the gold standard for the diagnosis of coronary artery di-sease. Most coronary angiograms are still stored on 35 mm cinefilm. With the increasing number of centers performing coronary interventions and the availability of faster and cheaper computer systems, digitizing co-ro-nary angiograms has become increasingly accepted since the mid-80s [5]. Today, approximately 75% of the cathlabs use digital techniques for the immediate assessment of coronary an-giograms for diagnostic and interventional procedures [2]. These digital angiograms are usually then erased from the hard disk.
For archiving digital angiograms, the prerequisites for replacing cinefilm are clearly defined: Firstly, image quality should not be inferior to cinefilm [1, 6]. Since the ACC/ACR/NEMA-standard was not yet established until re-cently, many cathlabs de-cided on a filmless-digital, but not standardized archiving. Others opted for a filmless-analog ar-chi-ve using video techniques: videotapes offer several advanta-ges: they are in wide-spread use (and therefore simplify inter-institutional exchange), offer easy handling, instant replay capability and are inexpensive [7, 8, 9]. The ACC/ACR/NEMA-expert group, which was recently joined by the ESC (Europe-an Society of Car-diology), however, is very much concerned about the increasing use of video techniques for archiving and for inter-institutional exchange. The ex-pert group says that S-VHS angiograms are - at their best - on-ly half as good as di-gi-tal angiograms [1, 2, 10]. Even the use of analog la-ser disks (e.g. LDA) may be only an interim solution [6]. In contrast, however, others showed that ana-log laser disks may be sufficient for clinical purposes [11, 12]. Even S-VHS, par-ti-cularly with edge-enhancement, offers an acceptable quality comparable to cine-film [13]. Disadvantages of videotapes are the loss of image quality after co-pying as well as their sensitivity to external influences.

Digital techniques offer several advantages over video techniques: no loss of quality of copies, the possibility of zooming and the "au-ra" of modern technology. Digital coronary angiography is similar to cinefilm, even for the quantification of lesions [14], with its tendency to overestimate ste-no-ses in small vessels [15].
The minimum requirement for digitizing coronary angiograms is a pixel resolution of 512x512 at 256 grey levels (8 bit), resulting in 256 KB per frame (512x512x8). Using a data compression of 2:1 according to the ACC/ACR/NEMA/ESC-guidelines, a da-ta flow of approximately 3.9 MB/s for NTSC (30 frames/s, USA) or 3.1 MB/s for PAL/SECAM (25 frames/s, Europe) can be calculated. These challenging data streams, however, cannot be accomplished by CD-players: even the fa-stest commercially available CD-players with their "12x-speed" (i.e. 12 times as fast as au-dio CD-players), achieve only up to 1.8 MB/s. Therefore, CD-players can-not pro-vide real-time viewing of digital angiograms directly from CD. This leaves only two choi-ces: Viewing coronary angiograms in slow motion (appr. 8 frames/s) or accepting longer wai-ting times until the angiograms are copied from CD-R to a hard disk, which may take up to 15 minutes per study.
To reduce the data streams, the following possibilities exist: The field of view may be cut, choo-sing the most important, representative part of the image (reduction of da-ta per frame). The acquisition speed may also be reduced to, for instance, 12.5 frames/s in Europe or to 15 frames/s in the USA. This, however, is not recommended by the expert group [2, 6]. Recently, the creation of new quan-ti-zation tables was proposed in order to reduce the amount of redundancy as well as some irrelevant information and noise [16]. One might on the other hand wait until newer and faster CD-players are developed. This is, however, highly unlikely, because a 24x-speed player would be necessary to replay approximately up to 3.5 MB/s. The new DVD-standard (Digital Versatile Disc) will not only in-crease the storage capacity per disc (up to 19 GB), but also allow a higher replay speed [17].
Therefore, the only realistic possibility available today is to increase the compression rates: in con-trast to lossless compression algorithms (including variable-length bit codes (Huffman co-des and variants), dictionary-based compression (Lempel-Ziv variants) and arith-me-tic coding [18], higher compression rates carry the inherent risk of deteriorating ima-ge quality due to compression artifacts with their uncertain clinical im--pact [1]. These techniques were traditionally classified as "lossy" (de-struc-tive, irreversible).
The ACC/ACR/NEMA/ESC-group has chosen the JPEG standard, because JPEG is already established in medicine, especially in many angiography systems. The JPEG compression algorithm with its inherent risk of blocky artifacts (8x8 pixel-DCT-blocks) is generally accepted to be lossless only for a compression rate of 2:1. This historical classification into lossless and lossy compression is based on phy-sical-mathematical considerations and not on physiologic-clinical criteria. Recently, the "los-sy" compression rates of 3:1 and 4:1 were classified as "not entirely loss free" [19].
As our study has shown, a JPEG data compression of 5:1 and 6:1 yielded the best results for "lossy compression" and did not lead to a clinically visible loss of image quality. In particular, blocking artifacts were not cli-ni-cal-ly relevant. Furthermore, there is enough safety margin from the limit of clinically acceptable image quality.
Other studies comparable to ours with coronary angiograms using various compression rates were not conducted or not pu-blished. Koning et al. mentioned that a compression rate of 4:1 does not lead to a significant deterioration of visual interpretation of 19 coronary an-gio-grams [19]. Several groups have worked on the problem of data compression un-der clinical circumstances with static pictures: JPEG compression with 10:1 of con-ven-tio-nal X-ray images, computer tomograms and ultrasound images showed no visible differences compared with the originals [20]. Using the FFBA-technique and a compression rate of 20:1 for X-ray images of the chest, acceptable results were obtained [21]. For the JPEG algorithm, however, 15:1 already delineated a loss of image quality [17, 22]. On the other side, JPEG compression rates of 40:1 in dermatology did not influence diagnostic performance, although resulting in a significantly lower rating [22]. Since image quality was significantly more deteriorated in images of skeletal muscle com-pared with images of the abdomen [23], one can assume that the results for one type of image cannot necessarily be extrapolated to images of other organs. Mo-vies are possibly more sensitive to compression rates than static images. In a recently published study, lossy JPEG (15:1) compression of coronary angiograms did not alter the diagnostic assessment of lesion severity [24|. However, no conclusions could be drawn re-gar-ding the accuracy of other compression rates [24].
The results of our study are only applicable for JPEG-compression, as required by the ACC/ACR/NEMA/ESC-standard. Other compression algorithms, like MPEG-1 (CD-I) with 352x288 pixels and "interframe interpolations", should not be used for digital co-ro-na-ry angiography as well as other restricted formats like AVI, low level Quick Time, Cinepack, In-deo or Xing. The influence of adaptive ("dynamic") compression algorithms, i.e. diffe-rent compression rates depending on image contents, is not known. Newer com-pression algorithms like MPEG-2 or MPEG-4 with lower compression rates as com-pared to MPEG-1 may be promising. However, they are not accepted as a cardiology standard.
Another advantage of reducing data streams offers easier real-time transmission of coronary angio-grams through networks (ATM, FDDI) to other locations, for example to heart surgeons or referring physicians.

Study limitations:

It was our purpose to investigate the influence of various compression rates on image quality of coronary angiograms in consecutive patients referred for invasive diagnostic procedures. Further studies are necessary to analyze the influence of the severity of stenoses, of the lesion types (plaque, thrombus, dissection) and the assessment of the results of interventions. In par-ticu-lar-, a similar study using various compression rates with quantitative lesion analysis [25] depending on image contrast and the use of different contrast agent osmolality is necessary. Digital systems may need higher requirements for scientific work than for clinical practice [13, 26].

Conclusions:

Following the new international standard using CD-R as archiving and exchange media for digital co-ronary angiography results in a dilemma between complying with official guidelines (2:1 compression) and prac-ti-cability (no real-time viewing from CD-R). Since no clinically relevant loss of information at a com-pression rate of 6:1 was experienced in our study, a modification of the ACC/ACJ/NEMA/ESC-guidelines allowing higher com-pression rates should be considered.

References

1, Nissen SE, Pepine CJ, Bashore TM, Block PC, Bonchek LI, Brinker JA, Carabello B, Douglas JS Jr, Elion JL, Hirshfeld JW Jr, Holmes DR, Johnson WL, Klinke WP, Levin DC, Mancini GBJ, Mullins CE, Thomas JD, Topol EJ, Vogel JHK, Wondrow MA: Cardiac Angiography Without Cine Film. Erecting a Tower of Babel" in the Cardiac Catheterization Laboraty. J Am Coll Cardiol 1994; 24:834-837
2, Nissen et al. for the ACC/ACR/NEMA ad hoc group. American College of Cardiology, American College of Radiology and Industry Develop Standard for Digital Transfer of Angiographic Images. J Am Coll Cardiol 1995; 25:800-802
3, Pelanek GA. DICOM Cardiology Exchange. In: Digital Cardiac Imaging in the 21st Century: A Primer, edited by Kennedy TE, Nissen SE, Simon R, Thomas JD, Tilkemeier PL, 1997, 24-30
4, Huffman DA. A method for the construction of minimum redundancy codes. Proc of the Inst of Radio Eng 1952; 40:1098-1101
5, Eigler N, Pfaff M, Whiting J, Thasana N, Forrester J. The role of digital angiography in the evaluation of coronary artery disease. Int J Cardiol 1986; 10:3-13
6, Simon R, Brennecke R, Heiss O, Meier B, Reiber H, Zeelenberg C. Report of the ESC Task Force on Digital Imaging in Cardiology. Eur Heart J 1994; 15:1332-1334
7, Gray JE, Wondrow MA, Smith HC, Holmes DR Jr. Technical considerations for cardiac laboratory high-definition video systems. Cathet Cardiovasc Diagn 1984; 10:73-86
8, Holmes Jr DR, Smith HC, Gray JE, Wondron MA. Clinical evaluation and application of cardiac laboratory high-definition video systems. Cathet Cardiovasc Diagn 1984; 10:63-71
9, Laboratory Performance Standards Committee. Interim standard: videotape recorders for transferring cardiac cinefluorographic studies. Cathet Cardiovasc Diagn 1990; 21:58-60
10, Zerbst J, Becker T, Simon R. Filmless Cardiac Catheterization: Are Analog Videodisks or Videotapes Adequate Substitutes for the Cinefilm ? Circulation 1995; 92: I-2
11, Chamuleau SA, Hanekamp WB, Appelman YE, Koch KT, Peters RJ, Kok WE, la Rievière GA, David GK, Pick JJ: Analogue Laser Optical Disc Versus Cinefilm Angiography. A Comparative Study. Circulation 1995; 92:I-21
12, Haase J, Preusler W, Göhring S, Semmler N, Kruse K, Schwarz F, Störger H, Reifart N. How Reliable are Computer-assisted Geometric Coronary Measurements Following Laser-Disc-Archiving (LDA): A Comparison with QCA on Cinefilm. Circulation 1995; 92:I-22
13, Ozaki Y, Keane D, Herrman JP, Foley D, Haase J, den Boer A, di Mario C, Serruys PW. Coronary arteriography for quantitative analysis: Experimental and clinical comparison of cinefilm and video recordings. Am Heart J 1995; 129:471-475
14, Haase J, Nugteren SK, Montauban van Swijndregt E, Slager CJ, Di Mario C, de Feyter PJ, Serruys PW. Digital geometric measurements in comparison to cinefilm analysis of coronary artery dimensions. Cathet Cardiovasc Diagn 1993; 28:283- 290
15, Gurley JC, Nissen SE, Booth DC, Harrison M, Grayburn P, Elion JL, DeMaria AN. Comparison of Simultaneously Performed Digital and Film-Based Angiography in Assessment of Coronary Artery Disease. Circulation 1988; 7 8:1411- 1420
16, Onnasch DG, Prause GP, Ploger A. Objective methods for optimizing JPEG compression of coronary angiographic images. Int J Card Imaging 1995; 11:151- 162
17, Elion JL, Whiting JS. Clinical use of lossy image compression in digital angiography. Am J Cardiol 1996; 78:219-220
18, Gillespy T 3rd, Rowberg AH. Displaying radiologic images on personal computers: image storage and compression. J Digit Imaging 1994; 7:1-12
19, Koning G, van Meurs BA, Haas H, Reiber JHC. Effect of Data Compression on Quantitative Coronary Measurements. Cathet Cardiovasc Diagn 1995; 34:175- 185
20, Yamamoto LG. Using JPEG image compression to facilitate telemedicine. Am J Emerg Med 1995; 13:55-57
21, Aberle DR, Gleeson F, Sayre JW, Brown K, Batra P, Young DA, Stewart BK, Ho BK, Huang HK. The Effect of Irreversible Image Compression on Diagnostic Accuracy in Thoracic Imaging. Invest Radiol 1993; 28:398-403
22, Sneiderman C, Schosser R, Pearson TG. A comparison of JPEG and FIF compression of color medical images for dermatology. Comput Med Imaging Graph 1994; 18:339-342
23, Goldberg MA, Pivovarov M, Mayo-Smith WW, Bhalla MP, Blickman JG, Bramson RT, Boland GW, Llewellyn HJ; Halpern E. Application of wavelet compression to digitized radiographs. Am J Roentgenol 1994; 163:463-468
24, Rigolin VH, Robiolio PA, Spero LA, Harrawood BP, Morris KG, Fortin DF, Baker WA, Bashore TM, Cusma JT. Compression of digital coronary angiograms does not affect visual or quantitative assessment of coronary artery stenosis severity. Am J Cardiol 1996; 78:131-135
25, Koning G, Baretta P, Zwart P, Reiber JHC. Effect of Lossy Image Compression on QCA Results. Circulation 1995; 92:I-22
26, Fritsch JP, Negwer F, Renneisen U, Brennecke R. Visual and quantitative analysis of coronary angiograms after irreversible data compression. Eur Heart J 1994; 15:46

Correspondence to:

Sigmund Silber, MD, FACC

Professor of Medicine

Dr. Müller Hospital

Am Isarkanal 36

D-81379 München

Tel: (+ 49 89) 74 21 51-0

Fax (+ 49 89) 74 21 51 31

e-mail: ssilber@med.de