Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print February 19, 2003, 10.1148/radiol.2271020067
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2271020067v1
227/1/124    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ohgiya, Y.
Right arrow Articles by Munechika, H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Ohgiya, Y.
Right arrow Articles by Munechika, H.
(Radiology 2003;227:124-127.)
© RSNA, 2003

Computer Applications

Acute Cerebral Infarction: Effect of JPEG Compression on Detection at CT1

Yoshimitsu Ohgiya, MD, Takehiko Gokan, MD, Hiroshi Nobusawa, MD, Masanori Hirose, MD, Noritaka Seino, MD, Hidefumi Fujisawa, MD, Maiko Baba, MD, Kyoko Nagai, MD, Keisuke Tanno, MD, Nobuyuki Takeyama, MD and Hirotsugu Munechika, MD

1 From the Department of Radiology, Showa University School of Medicine, Tokyo, Japan (Y.O., T.G., H.N., M.H., N.S., M.B., K.N., K.T., N.T., H.M.); and Department of Radiology, Showa University Northern Yokohama Hospital, Japan (H.F.). From the 2001 RSNA scientific assembly. Received February 8, 2002; revision requested April 18; final revision received July 16; accepted August 29. Address correspondence to Y.O., Department of Radiology, Showa University School of Medicine, 1-5-8 Hatanodai Shinagawa-ku, Tokyo 142-0064, Japan (e-mail: yogiya@gd5.so-net.ne.jp).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
PURPOSE: To evaluate the effect of Joint Photographic Experts Group (JPEG) compression ratios of 10:1 and 20:1 on detection of acute cerebral infarction at computed tomography (CT).

MATERIALS AND METHODS: CT images obtained in 25 patients with acute cerebral infarction and 25 patients with no lesions were compressed by means of a JPEG algorithm at ratios of 10:1 and 20:1. Normal and abnormal sections (on original and compressed images) were reviewed by using a color soft-copy computed monochrome cathode ray tube monitor. Five observers rated the presence or absence of a lesion with a 50-point scale (0, definitely absent; 25, equivocal; and 50, definitely present). Diagnostic accuracy was evaluated with receiver operating characteristic (ROC) curve analysis. Significant difference was defined as a P value less than .05 for the area tested with a two-tailed paired Student t test.

RESULTS: At ROC analysis, no statistically significant difference was detected for all cases considered together (Az [area under the ROC curve] = 0.887 ± 0.038 [mean ± SD] on noncompressed images, Az = 0.897 ± 0.038 on 10:1 compressed images, and Az = 0.842 ± 0.073 on 20:1 compressed images; P > .05).

CONCLUSION: JPEG compression at ratios of 10:1 and 20:1 was tolerated in the detection of acute cerebral infarction at CT.

© RSNA, 2003

Index terms: Brain, CT, 10.12111 • Computed tomography (CT), image processing • Data compression • Images, interpretation


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Image data compression is important because the time and cost necessary for image transmission are decreased. Data compression also can decrease storage requirements and speed image retrieval. Image compression can be divided into lossless (reversible) and lossy (irreversible) compression methods. Lossless algorithms allow perfect reconstruction of the original image after compression, but these algorithms achieve only 2:1 to 3:1 reduction for medical images (1). Lossy algorithms provide higher ratios but do not perfectly reproduce the original image. Tolerance of image compression has been evaluated for various modalities and organs (2–11). We selected acute cerebral infarction to evaluate compressed CT images for several reasons. (a) Acute cerebral infarctions can be difficult to detect if they are small and subtle. (b) In a large multicenter trial on thrombolysis (12), the local investigators, composed mainly of nonradiologist physicians, were less sensitive and less specific in detecting hypoattenuating brain tissue than were the neuroradiologists (13). (c) CT is highly specific in the detection of irreversible ischemic brain damage within the first hours of stroke onset (13). We chose Joint Photographic Experts Group (JPEG) compression because it is widely used, and JPEG is a standards organization. The purpose of this study was to determine the effect of JPEG compression ratios of 10:1 and 20:1 on the detection of acute cerebral infarction at CT.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Image Selection
All CT scans were obtained in 25 patients with acute cerebral infarction and 25 patients with no lesions. The mean age of patients (14 men, 11 women) with acute cerebral infarction was 68.3 years ± 17.5 (SD) (range, 23–92 years). The mean age of the control subjects (10 men, 15 women) was 70.6 years ± 8.2 (range, 58–90 years). Twenty-five of the selected images from the 25 patients with no lesions, chosen randomly from all levels of the brain, were normal and served as controls. These controls included eight images of the posterior fossa and 17 images of the supratentorial compartment. Two neuroradiologists (N.S., Y.O) selected together the 25 normal images as controls. One image per control subject was selected for control images. The other 25 images were obtained in 25 patients with developing symptoms of acute cerebral infarction within 48 hours. Diagnoses were based on either CT findings obtained a few days after symptom onset or magnetic resonance (MR) imaging findings on diffusion-weighted images. One image per patient was selected. The 25 images selected by the two neuroradiologists together showed abnormalities with difficulty of detection owing to small size and/or subtle contrast. The institutional review board approved the study protocol, and informed consent was obtained from all participants.

Image Acquisition and Compression
All CT images were acquired by using a commercially available scanner (HiSpeed; GE Medical Systems, Milwaukee, Wis) between September 2000 and February 2001 at our institution. Section thickness was 10 mm at a 10-mm table interval in the supratentorial compartment and 5 mm at a 5-mm table interval in the posterior fossa, with conventional scanning parameters (140 kVp, 170 mA). The cases selected in the present study were retrieved from an optical disk archive by using software (Image VINS version 1.41; Yokokawa Denki, Tokyo, Japan). The image data were transferred in direct digital form to a personal computer (PC/AC; Dell, Tokyo, Japan), and images were interpreted on a 17-inch monochrome cathode ray tube monitor (1,024 x 768) (CV173; Totoku, Tokyo, Japan). JPEG software (Independent JPEG Group, August 1995) was used for the compression algorithm. In each case, the selected section was compressed at ratios of 10:1 and 20:1.

Image Review
Five senior radiologists (H.F., H.N., M.B., M.H., T.G), who were board certified and had served as nonneuroradiologists for 6–20 years (years of experience of each observer were 10 years for observer 1, 16 for observer 2, 6 for observer 3, 14 for observer 4, and 20 for observer 5), reviewed each set of 50 images compressed at the same ratios or noncompressed. To reduce reading-order effects, the readers interpreted each set of images noncompressed and compressed at ratios of 10:1 and 20:1 in random order, and review sessions were separated by at least 2 weeks. To maintain a constant objective observation condition, each observer reviewed the images of the brain independently and was allowed a limited viewing time (30 seconds per image) with a fixed window setting (window width, 100 HU; window level, 30 HU). Image review was performed in a quiet darkened environment. Observers scored each image for the presence or absence of an acute cerebral infarction with a 50-point scale: 0, definitely absent; 25, equivocal; and 50, definitely present.

Statistical Analysis
For each noncompressed or compressed image evaluated, a binomial receiver operating characteristic (ROC) curve was fitted to each observer’s confidence rating by means of a maximum likelihood estimation. The computer program ROCKIT 0.9B (C. E. Metz, Chicago, Ill) was used (14,15). The diagnostic accuracy with each noncompressed or compressed image was determined by calculating the area under each ROC curve (Az). Composite ROC curves were used to represent the performance of the five observers as a group and were calculated by averaging the binomial parameter values of the individual curves. To determine whether there was any significant difference in Az values for the three types of images, the two-tailed paired Student t test was performed. This test was used to compare differences between the patients with acute cerebral infarction and the control subjects with similar ages. A P value of less than .05 was considered to indicate a statistically significant difference. In addition, statistical performance, including sensitivity, specificity, accuracy, and positive and negative predictive values in detecting acute cerebral infarction, was evaluated for each compression ratio at a selected criterion. Thus, images with a rating of definitely present and probably present that the observers scored as 25 or higher were treated as having positive findings, whereas images with a rating of definitely absent and probably absent that the observers scored as less than 25 were treated as having negative findings. These results were then compared with those of standard diagnosis.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 REFERENCES
 
In a comparison of the ages of the patients with acute cerebral infarction and those of the control subjects, the difference was not significant (P > .05). Representative examples of brain CT images noncompressed and compressed at a ratio of 10:1 or 20:1 are shown in Figures 1 and 2. The Az values of noncompressed, 10:1 compressed, and 20:1 compressed images for each observer are shown in Table 1. With JPEG compression, no significant difference was noted in the Az values among noncompressed, 10:1 compressed, and 20:1 compressed images. The comparisons of the average ROC curves of noncompressed and compressed images are shown in Figure 3. The diagnostic value of each compression ratio is shown in Table 2.



View larger version (104K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1a. Transverse CT images show acute cerebral infarction in the left insular cortex in a 72-year-old man. A small area of hypoattenuation (arrows) can be seen on (a) the noncompressed original image and remains visible on (b, c) JPEG compressed images obtained at a ratio of 10:1 (b) and 20:1 (c).

 


View larger version (104K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1b. Transverse CT images show acute cerebral infarction in the left insular cortex in a 72-year-old man. A small area of hypoattenuation (arrows) can be seen on (a) the noncompressed original image and remains visible on (b, c) JPEG compressed images obtained at a ratio of 10:1 (b) and 20:1 (c).

 


View larger version (104K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1c. Transverse CT images show acute cerebral infarction in the left insular cortex in a 72-year-old man. A small area of hypoattenuation (arrows) can be seen on (a) the noncompressed original image and remains visible on (b, c) JPEG compressed images obtained at a ratio of 10:1 (b) and 20:1 (c).

 


View larger version (114K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2a. Transverse CT images show acute cerebral infarction in the left middle cerebral artery area in a 72-year-old man. Obscuration of sulci (arrows) in the left temporal lobe can be seen on (a) the original noncompressed image and remains visible on (b, c) JPEG compressed images obtained at a ratio of 10:1 (b) and 20:1 (c).

 


View larger version (114K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2b. Transverse CT images show acute cerebral infarction in the left middle cerebral artery area in a 72-year-old man. Obscuration of sulci (arrows) in the left temporal lobe can be seen on (a) the original noncompressed image and remains visible on (b, c) JPEG compressed images obtained at a ratio of 10:1 (b) and 20:1 (c).

 


View larger version (114K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2c. Transverse CT images show acute cerebral infarction in the left middle cerebral artery area in a 72-year-old man. Obscuration of sulci (arrows) in the left temporal lobe can be seen on (a) the original noncompressed image and remains visible on (b, c) JPEG compressed images obtained at a ratio of 10:1 (b) and 20:1 (c).

 

View this table:
[in this window]
[in a new window]

  		TABLE 1. Az Values for Each Type of Image

 


View larger version (30K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. ROC curves for noncompressed ({blacksquare}) and 10:1 ({diamond}) and 20:1 ({blacktriangleup}) JPEG compressed CT images used for the diagnosis of acute cerebral infarction. The diagnostic accuracy with images compressed at a ratio of 20:1 was slightly worse than that with noncompressed images, but the difference was not significant (P > .05).

 

View this table:
[in this window]
[in a new window]

  		TABLE 2. Diagnostic Performance for Prediction of Acute Cerebral Infarction by Using Each Type of Image

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
REFERENCES
 
Different types of images have different degrees of compression tolerance, where compression tolerance is defined as the maximum compression in which the decompressed image is acceptable for interpretation and esthetics. Investigators in multiple studies (2–11) suggested that compression ratios lower than 20:1 to 40:1 were considered acceptable for interpreting radiographs. Findings in other studies (2,3) support compression ratios at least as high as 20:1 for chest radiographs. Digitized skeletal radiographs tend to be more sensitive to compression than are digitized chest radiographs (3). In nuclear medicine studies, the acceptable compression ratio can be as high as 100:1 (4), whereas in studies of digital coronary angiograms, the ratio can be less than 15:1 (5,6). CT, MR, and ultrasonographic images exhibit fairly low tolerance to compression (<20:1) (7–11). CT images compress more readily than do MR images, and a single acceptable compression ratio cannot be assigned to a modality if the given organ systems are different. With regard to CT images, the acceptable compression ratio is 9:1 for mediastinal adenopathy and pulmonary nodules (9), 10:1 for detection of focal hepatic lesions (10), and 20:1 for the diagnosis of coronary artery calcification (11). In contrast, the findings of the present study suggested that the acceptable compression ratio for brain CT images was 20:1 for diagnosis of acute cerebral infarction.

The effect of compression on image quality depends on the image content and compression ratio. Most compression techniques consist of three steps: transformation, quantization, and encoding. Only the second step is a lossy step in the process. The algorithm of the JPEG method breaks the image into 8 x 8-pixel blocks and performs a discrete cosine transform on each block. The result is an 8 x 8 block of spectral coefficients with most of the information concentrated in relatively few coefficients. Quantization is performed, which approximates the larger coefficients; smaller coefficients become zero. These quantized coefficients are then reordered in a zigzag manner to group the largest values first, with long strings of zeros at the end that can be represented efficiently. It was commonly considered that most wavelet compression was superior to that of JPEG in radiologic images at a given compression ratio (7). In a previous study, investigators reported a small difference between the two algorithms at compression ratios of less than 20:1, which was an acceptable compression ratio for clinical diagnosis (11).

Teleradiology has become an important aspect of medicine, since it facilitates the delivery of a better medical service in rural areas by communicating with consultant radiologists at a central medical center. In teleradiology, compression techniques could reduce the cost and time necessary for image transmission. Compression could greatly facilitate image storage and speedy image retrieval if images could be compressed at a workstation. The compression technique could be used for image storage and transmission without compromising diagnostic value.

In conclusion, the findings of the present study suggest that JPEG compression at ratios of up to and including 20:1 were tolerated in the detection of acute cerebral infarctions at CT.


   FOOTNOTES
 
Abbreviations: Az = area under the ROC curve, JPEG = Joint Photographic Experts Group, ROC = receiver operating characteristic

Author contributions: Guarantors of integrity of entire study, Y.O., T.G., H.M.; study concepts, Y.O., T.G., H.M.; study design, Y.O., H.N., M.H.; literature research, Y.O., M.B., K.N., K.T.; clinical studies, Y.O., T.G., H.N., M.H., N.S., H.F., M.B., K.N., K.T.; data acquisition, H.N., M.H., N.S., H.F.; data analysis/interpretation, Y.O., M.B., K.N., K.T., N.T.; statistical analysis, K.N., K.T., N.T.; manuscript preparation, Y.O., N.S., H.F., M.B.; manuscript definition of intellectual content, Y.O., T.G., H.N., M.H.; manuscript editing, H.F., M.B., K.N., K.T., N.T.; manuscript revision/review, T.G., H.N., M.H., N.S.; manuscript final version approval, Y.O., T.G., H.M.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 

  1. Karson TH, Chandra S, Morehead AJ, Stewart WJ, Nissen SE, Thomas JD. JPEG compression of digital echocardiographic images: impact on image quality. J Am Soc Echocardiogr 1995; 8:306-318.[CrossRef][Medline]
  2. MacMahon H, Doi K, Snada S, et al. Data compression: effect on diagnostic accuracy in digital chest radiography. Radiology 1991; 178:175-179.[Abstract/Free Full Text]
  3. Gorldberg MA, Pivovarov M, Mayo-Smith WW, et al. Application of wavelet compression to digitized radiographs. AJR Am J Roentgenol 1994; 163:463-468.[Abstract/Free Full Text]
  4. Chameroy V, Di Paola R. High compression of nuclear medicine dynamic studies. Int J Card Imaging 1990; 5:261-269.[CrossRef][Medline]
  5. Rigolin VH, Robiolio P, Spero LA, et al. Compression of digital coronary angiogram does not affect visual or quantitative assessment of coronary artery stenosis severity. Am J Cardiol 1996; 78:131-135.[Medline]
  6. Silber S, Doee R, Zindler G, et al. Impact of various compression rates on interpretation of digital coronary angiogram. Int J Cardiol 1997; 60:195-200.[CrossRef][Medline]
  7. Erickson BJ, Manduca A, Palisson P, et al. Wavelet compression of medical images. Radiology 1998; 206:599-607.[Free Full Text]
  8. Maldijian JA, Liu WC, Hirschorn D, Murthy R, Semanczuk W. Wavelet transform-based image compression for transmission of MR data. AJR Am J Roentgenol 1997; 169:23-26.[Abstract/Free Full Text]
  9. Cosman PC, Davidson HC, Bergin CJ, et al. Thoracic CT images: effect of lossy image compression on diagnostic accuracy. Radiology 1994; 190:517-524.[Abstract/Free Full Text]
  10. Goldgberg MA, Gazelle GS, Boland GW, et al. Focal hepatic lesions: effect of three-dimensional wavelet compression on detection at CT. Radiology 1997; 202:159-165.[Abstract/Free Full Text]
  11. Zheng LM, Sone S, Itani Y, et al. Effect of CT digital image compression on detection of coronary artery calcification. Acta Radiol 2000; 41:116-121.[CrossRef][Medline]
  12. Hacke W, Kaste M, Fieschi C, et al. Randomised double-blind placebo-controlled trial of thrombolytic therapy with intravenous alteplase in acute ischaemic stroke (ECASS). Lancet 1998; 352:1245-1251.[CrossRef][Medline]
  13. Kummer RV, Bourquain H, Bastianello S, et al. Early prediction of irreversible brain damage after ischemic stroke at CT. Radiology 2001; 219:95-100.[Abstract/Free Full Text]
  14. Metz CE. ROC methodology in radiologic imaging. Invest Radiol 1986; 21:720-733.[Medline]
  15. Metz CE. Some practical issues of experimental design and data analysis in radiological ROC studies. Invest Radiol 1989; 24:234-245.[Medline]



This article has been cited by other articles:


Home page
 Am. J. Roentgenol.Home page
B. Kim, K. H. Lee, K. J. Kim, R. Mantiuk, S. Hahn, T. J. Kim, and Y. H. Kim
Prediction of Perceptible Artifacts in JPEG 2000 Compressed Chest CT Images Using Mathematical and Perceptual Quality Metrics
Am. J. Roentgenol., February 1, 2008; 190(2): 328 - 334.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Roentgenol.Home page
H. S. Woo, K. J. Kim, T. J. Kim, S. Hahn, B. Kim, Y. H. Kim, and K. H. Lee
JPEG 2000 Compression of Abdominal CT: Difference in Tolerance Between Thin- and Thick-Section Images
Am. J. Roentgenol., September 1, 2007; 189(3): 535 - 541.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2271020067v1
227/1/124    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ohgiya, Y.
Right arrow Articles by Munechika, H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Ohgiya, Y.
Right arrow Articles by Munechika, H.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE