Diagnostic Quality Evaluation of Compressed Medical Images for Telemedicine Applications Seddeq E. Ghrare, M. Alauddin M. Ali, M. Ismail, K. Jumari Universiti Kebangsaan Malaysia (UKM) Faculty of Engineering { seddeq, mama, mahamod, kbj }@vlsi.eng.ukm.my

Abstract Many techniques for achieving data compression have been introduced. The fundamental goal of image data compression is to reduce the bit rate for transmission or storage while maintaining an acceptable reproduction quality, but it is natural to raise the question of how much an image can be compressed and still preserve sufficient information for a given telemedicine application. Evaluation of the quality of compressed medical image for telemedicine applications still remains an important issue. In this paper, the evaluation of diagnostic quality of compressed medical images using objective and subjective testing will be presented. Three different medical image modalities which are CT, MRI, and X-ray have been compressed and decompressed using DWT for different compression ratios. The quality of the reconstructed images has been measured objectively using objective measures such as MSE, MAE, SNR, and PSNR. Ten non specialist observers have been involved to carry out the subjective evaluation. Based on the quality of the reconstructed images, the PSNR obtained has been between 35.3dB to 58.0dB for CT scan images, 38.6dB to 55.0dB for MRI and 34.5dB to 51.0dB for x-ray images. For clinical applications such as telemedicine or teleradiology, the compression ratio of 30:1 is acceptable for CT images, and a compression ratio of 40:1 is acceptable for MRI, and compression ratio of 20:1 is acceptable for x-ray images. Keywords: Medical Image, Wavelet Transform, Telemedicine.

minimizing storage requirement and speeding transmission time. The primary goal of medical image compression is to achieve the best possible fidelity for the available communication and storage channels, therefore the objective of compression is to reduce the data volume and to achieve a low bit rate in the digital representation of radiological images without perceive loss of image quality [2]. Image data compression can be classified into two broad categories: Lossy and Lossless (information preserving) [3]. Lossy compression schemes have not been widely used for both clinical and legal reasons. However standard and newer Lossless compression algorithms such as JPEG2000 and wavelet-based compression can yield images statistically identical diagnostic results compared with using the original images without any loss [4,5], therefore lossless image coding is important for medical image compression because any information loss or error caused by the image compression process could affect clinical diagnostic decision [6]. The aim of this paper is to evaluate a set of compressed medical images using wavelet transform technique for an acceptable degree of the reconstructed CT, MRI, and X-ray images for different compression levels. Both objective and subjective methods are applied for this evaluation.

2. Image Quality Measures 1. Introduction Remote medical monitoring is a telemedicine application in which dynamic fluoroscopy images during a radiological interventional procedure are transmitted in real time or near real time to another location, where the physician or specialist can advice regarding the diagnostic and therapeutic strategies. [1]. To represent such large raw medical images with smallest possible number of bits, image data compression is essential and plays an important role in

Methods for image quality evaluation can be classified as objective and subjective measures. By objective measures some statistical indices are calculated to indicate the reconstructed image quality and by subjective measure viewers read images directly to determine their quality .

2.1 Objective Measures A widely used measure of reconstructed image for an N x M size image is the mean square error (MSE) as given by [6].

INFOS2008, March 27-29, 2008 Cairo-Egypt © 2008 Faculty of Computers & Information-Cairo University

HBI-15

MSE =

1 NM

N −1 M −1

∑ ∑ ⎡⎢⎣ f (i, j ) − f (i, j ) *

1= 0 j = 0

2

⎤ ⎥⎦

(1)

Where f (i, j ) the original is image data and

Signal-to-Noise Ratio (SNR) is widely used in the signal processing literature (since it is related to the signal power and noise power), and is perhaps more meaningful because it gives 0 dB for equal signal and noise power. SNR is used more commonly in the

f * (i, j ) is the compressed image data. image-coding field. So, the SNR that is used corresponding to the above error is defined as N −1 M −1 ⎧ 2 f (i , j ) ∑ ∑ ⎪ ⎪ i =0 j =0 SNR = 10 log ⎨ N −1 M −1 ⎪ ∑ ∑ f (i , j ) − f * (i , j )2 ⎪⎩ i = 0 j = 0

[

⎫ ⎪ ⎪ dB ⎬ ⎪ ⎪⎭

]

(2)

Another quantitative measure is the peak signal-tonoise ratio (PSNR), based on the mean square error of the reconstructed image. The formula for PSNR is given by :

⎛ 2B −1⎞ ⎟⎟ dB PSNR = 10 log⎜⎜ ⎝ MSE ⎠

(3)

Where B is the bit depth of the image. For an 8-bit image, the PSNR is computed by:

⎛ (255)2 PSNR = 10 log⎜⎜ ⎝ MSE

⎞ ⎟ dB ⎟ ⎠

(4)

2.2 Subjective measure Subjective evaluation by viewers is still a method commonly used in measuring image quality. The subjective test emphatically examines fidelity and at the same time considers image intelligibility. When taking subjective test, viewer's focus on the difference between reconstructed image and the original image, they notice such details where information loss cannot

be accepted. The representative subjective method is Mean Opinion Score (MOS) [7, 8, and 9]. It has two kinds of scores: one is absolute and another is relative. Two examples are shown below in Table 1. In our experiment, we use absolute score in order to seek the consistency between subjective and objective measures. Each viewer compares the reconstructed image with the original one to decide which level it belongs to and gives the score.

3. Results and Discussion Three different medical image modalities which are CT, MRI, and digitized x-ray as shown in figure 1 have been used in this study. The test results obtained by both objective and subjective measures are shown in figures 2-4. Table 2 summarizes the result for MSE and PSNR for these images and Figure 2 illustrates the PSNR values versus compression ratio. For the subjective evaluation results, Table 3 represents the average score of 10 non specialists' student observers from the National University of Malaysia. A score of 5 is no distortion (Excellent), score of 4 represents a little distortion which can be ignored (Good), score of 3 shows distortion which can be seen evidently but it can be accepted (Fair), score 2 shows a lot of distortion, which can not be accepted (Bad), and finally score of 1 shows too much distortion, therefore can not be tolerated (Very Bad).These results have been illustrated in figure 3 and a comparison between the original and reconstructed images is illustrated in figure 4.

Table 1. Mean Opinion Score (MOS) method used for subjective evaluation Absolute Score Relative Score 5 Excellent 5 The best in the group 4 Good 4 Better than the average 3 Fair 3 The average of the group 2 Bad 2 Worst than the average 1 Very Bad 1 The worst in the group

HBI-16

Table 2. The MSE and PSNR results for CT, MRI, and X-Ray images CT Modality MRI Modality X-Ray Modality Compression Ratio MSE PSNR(dB) MSE PSNR(dB) MSE PSNR(dB) 10:1 0.1 58.0 0.2 55.0 0.5 51.0 15:1 1.1 47.7 0.7 49.7 1.9 45.3 20:1 3.9 42.2 1.3 47.0 4.0 42.1 25:1 7.0 39.7 3.1 43.2 9.0 38.6 30:1 10.8 37.8 4.2 41.8 13 37.9 35:1 14.5 36.5 6.5 40.0 20 35.1 40:1 19.2 35.3 8.9 38.6 23 34.5

Table 3. Subjective evaluation results for CT, MRI, and X-Ray images The Average score of all readers Compression Ratio Score for CT Image Score for MR Image Score for X-Ray Image Original Image 5 5 5 10:1 4 5 4 20:1 4 4 4 30:1 3 4 3 40:1 3 4 2

A

B

C

Figure 1. Original Test Images: (A) CT (B) MRI (C) X-Ray

PSNR for CT

PSNR for MRI

PSNR for Xray

PSNR (dB)

80 60 40 20 0 10

15

20

25

30

35

Compression Ratio

Figure 2. PSNR and compression ratio for CT, MRI, and X-ray images

HBI-17

40

MOS

CT Image

MRI

X-ray

6 5 4 3 2 1 0 Original Image

10

20

30

40

Compression Ratio

Figure 3. Subjective score results for CT, MRI, and X-ray images

A

B

30:1

40:1

C

20:1

Figure 4. Comparison between original and compressed images

4. Conclusion In this study; three different modalities of medical images which are CT, MRI, and X-Ray have been compressed and reconstructed using wavelet transform. Objective and subjective evaluation has been done to evaluate the diagnostic quality of the reconstructed images. For the objective evaluation, the results show

that the PSNR which indicates the quality of the reconstructed image is ranging from (35.3dB to58.0dB, 38.6dB to 55.0dB, and 34.5dB to 51.0dB) for CT, MRI, and X-Ray respectively. For the subjective evaluation test, the results show that the compression ratio of 30:1 was acceptable for CT image, and a compression ratio of 40:1 was acceptable for MRI whereas for X-Ray image 20:1 was acceptable for clinical applications.

HBI-18

References [1] B. R. Sanders, J. H. Shanon, " Telemedicine: Theory and Practice", Springfield, Illionis, 1997. [2] B. J. Abdullah, K. H. Ng, R. Pathmanathan, “The Impact of Teleradiology in Clinical Practice: A Malaysian Perspective”, Med. Journal Malaysia, 1999, vol. 54, pp. 169-174 [3] Rafael C. Gonzalez, Richard E. Woods "Digital Image Processing", 2nd edition ,2002, Pearson Prentice Hall. [4] Persons K., Pallison P., Patrice M., Manduca A., Willian J., Charboneau. "Ultrasound grayscale image compression with JPEG and Wavelet techniques ", Journal of Digital Imaging, 2000; 13: 25-32. [5] Bradley J. and Erickson M.D, (2000). "Irreversible Compression of Medical Images", Department of Radiology, Mayo Foundation, [www.scarnet.org/pdf/SCAR%20Whitepaper.pdf] [6] Sayre J., Aberle D., and Boechat I., "The Effect of Data Compression on Diagnostic Accuracy in Digital Hand and Chest Radiography" 1992, Proceedings of SPIE, 1653: 232-240.

[7] Ahmet M., Paul S., "Image quality measures and their performance", IEEE Transactions on communications, 1995, 43: 2959-2965. [8] Lee H., Haynor D., and Kim Y. "Subjective evaluation of compressed image quality" Proceedings of SPIE, Image Capture, Formatting and Display, 1992; 1653: 241-245 [9] Pamela, C., Robert, M., Richard, A. "Evaluating quality of compressed medical images: SNR, Subjective Rating, and Diagnostic Accuracy", Proceedings of the IEEE, 1994; 82: 919-932

HBI-19