A statistical method for evaluation quality of medical images: a case study in bit discarding and image compression
Introduction
Medical images are often deteriorated by noise due to various interference sources and other phenomena that affect the measurement process in imaging acquisition systems. If the small difference between normal and abnormal tissues is contaminated by noise, direct analysis of the acquired images is often more difficult [1]. Noise in medical images also degrades the correlation between pixels. As a result, the reversible image compression ratio can only achieve a maximum compression ratio of about 2–4 [2].
Steward et al. [3] suggested that 5–6 least significant bits (LSBs) are spectrally white in an image by applying the power spectrum method to verify each bit plane in a magnetic resonance (MR) image. They concluded that the gray level dynamic range of a MR image spanned only six or seven bits at most. Chan et al. [4] showed that the detection accuracy was independent of pixel depth for discarded up to three LSB planes from a digitized mammography with 12 bits depth.
Olsen [5] estimated the standard deviation (STD) for Gaussian and uniform distributed white additive noise in images. He found that the use of STD to estimate noise in the residual images was most reliable. The residual image was obtained by subtracting the pre-filtered image from the original image, where smoothed image is obtained by average or median filters. Chuang et al. [6], [7], [8] reported that 8–9 information bits exist in medical images and 4–5 bits contain only noise. Discarding these noisy bits would not visually deteriorate image quality, as indicated in their study. However, the true effect on the image has not been verified.
With the evolution of teleradiology, the use of lossy compression technique is a requirement for reducing the image volume and speeding up communication [9]. However, greater compression is obtained at the expense of image quality. Efforts have been devoted to evaluate the subtle differences between images using visual or numerical methods. Many studies had tried to build a visual threshold or acceptance level of lossy image compression. However, the radiological tasks is widely varied and the building of visual threshold is task dependent and time consuming [2], [10], [11], [12], [13]. The normalized root mean square error (NMSE) or mean square error (MSE) for numerical estimation was most commonly used to measure the degradation in reconstructed images. The guidance document for Picture Archiving and Communication System (PACS) also requires that manufacturers report to FDA, the NMSE of their lossy compression techniques [14]. Nevertheless, NMSE and MSE are found to neither correlate nor provide any quality information [2], [10], [13], [15], [16].
The objective image quality metrics, visual discrimination model (VDM) and visible differences predictor (VDP), has been developed in past years. These metrics combined perceptual quality measurement by taking into account human visual system (HVS) features [16], [17], [18], [19], [20], [21], [22]. Since a human observer is the end user of image quality measurement, image quality model based on HVS seemed to be more appropriate for user perceptual. Recently, the detection threshold of artifact in medical image subject to JPEG and JPEG 2000 compression was reported [23]. They mentioned that the VDM predictions of artifact visibility were highly correlated with observer performance. However, it was reported [21], [22], [24] that none of the objective quality model outperformed others by comparing nine proponent objective quality models and all of them were statistically indistinguishable with the peak signal to noise ratio (PSNR).
Recently, Wang and Bovic [25] suggested a Q index to estimate image quality. The image spatial information is estimated from a local region with a 8×8 window size instead of from pixel, as indicated by their study. They showed that the variation of the Q index for a manipulated image was accordant with subjective image quality metrics. Chen et al. [26], [27] applied the Moran statistics in a local region to extract image spatial information. The sharpness or smoothness in image is indicated well by this statistics [27]. In this study we propose to evaluate spatial information in image by using Moran statistics and then a statistical method is applied to determine if this subtle difference between images is due to chance.
Section snippets
Method and material
The image was first manipulated to generate subtle differences images from the original. The image manipulation includes discarding LSBs as denoising, lossy image compression at low compression ratio and average filter blurring. The Z values on images were then estimated pixel-by-pixel using the Moran statistics with an 11×11 window [7], [8] and centered on it. Two cumulative Z value distributions were produced by collecting all Z values of image and then integrated. The Kolmogorov–Smirnov (KS)
LSBs manipulation
The LSBs in the image were considered as noisy bits and discarded using the whole-bit plane or adaptive methods. With the adaptive method, pixels of original image were discarded nb(x,y) or a conservative level nb(x,y)−1 bits. The mean discarding bits and of the image could be obtained by averaging nb(x,y) and nb(x,y)−1 on images, as listed in Table 1, Table 2, Table 3, Table 4 for the T1, T2, CT and CR images. On average, the highest noising level is the CR images followed by the CT,
Denoising
On average the order of STD estimation is CR>CT>T2>T1, but the KS test shows different aspects. In spite of highest STD, the CR image hardly passed the KS test even at 1 LSB plane manipulation. However, CT and T2 images accepted the null hypothesis with 2-bit planes manipulation. The D value of T1 images fell in the critical region with (with mean=1.58 bits) bits discarding. The order of discarding level of noisy bits should be CT>T2>T1>CR. Although the result contradicts the STD value of
Summary
In this report, we proposed a method to prove indistinguishableness statistically for subtle differences between images. The proposed method has shown that the discrepancy between an original and bits-manipulated image could be attributed to chance, meaning statistically that the LSBs could be discarded without any image degradation. With lossy image compression, this method has shown to be consistent with the results of human observer and model observer. The Moran Z values were very sensitive
Acknowledgements
The medical images and SMPTE test patterns used in this work were obtained with the help of Jim-Chao Chuang from the Department of Medical Imaging and Alex Hsu from the Department of Radiation Oncology, Chung Shan Medical and Dental College Hospital, Tai Chung, Taiwan.
Tzong-Jer Chen currently is a Postdoctoral Researcher at Department of Nuclear Medicine, National PET/Cyclotron Center, Taipei Veterans General Hospital, Taipei, Taiwan. He received his PhD in Medical Physics and MS in Health Physics both from the National Tsing-Hua University, Taiwan. His research interests include image analysis, medical image quality evaluation and inverse planning for radiation therapy.
References (31)
Estimation of noise in images: an evaluation
CVGIP
(1993)Teleradiology technologies and some service models
Comput Med Imaging Graphics
(1996)Handbook of medical imaging processing and analysis
(2000)Irreversible compression of medical images
(2000)- et al.
Gray level dynamic range in magnetic resonance imaging
SPIE Proc
(1986) - et al.
Digitization requirements in mammography: effects on computer-aided detection of microcalcifications
Med Phys
(1994) - et al.
Comparison of chi-square and joint-count methods for evaluating digital image data
IEEE Trans Med Imaging
(1992) - et al.
Assessment of noise in a digital image using the join-count statistic and Moran test
Phys Med Biol
(1992) - et al.
Noise content analysis in clinical digital images
Radiographics
(1994) - et al.
Methods of measurement of image quality in teleultrasound
Br J Radiol
(2000)
Evaluation of JPEG and wavelet compression of body CT images for direct digital teleradiologic transmission
Radiology
Assessment of visually lossless irreversible image compression: comparison of three methods by using an image-comparison workstation
Radiology
Radiologic image compression—a review
Proc IEEE
What's wrong with mean square error?
Cited by (18)
A neural network based framework for effective laparoscopic video quality assessment
2022, Computerized Medical Imaging and GraphicsCitation Excerpt :Monitoring and assurance of a good video quality is a very critical task in various applications like video streaming, video surveillance, medical procedures and underwater exploration (Wang, 2011). Any loss of useful information in these videos, resulting from some kinds of video distortion, may not only affect the visual experience but could also cause fatal consequences like in the fields of video surveillance (Beghdadi et al., 2018) and medical imaging (Amirrashedi et al., 2021; Chen et al., 2004). For instance, one of the most common medical procedures, where having a good video quality becomes highly desirable, is the laparoscopic/endoscopic surgery.
Characterisation of heterogeneity and spatial autocorrelation in phase separating mixtures using Moran's I
2018, Journal of Colloid and Interface ScienceCitation Excerpt :As shown in the images in Fig. 1, increased clustering leads to higher values of Moran’s I. Where pixel intensities are randomly distributed, Moran’s I is equal to 0, and more alternating features lead to lower, more negative, values of Moran’s I [34,36]. In the MRI studies, Moran’s I has been used to assess noise levels [30,31,35], study neural networks [16] and distribution of fat in muscles [34]. In the study by Derado et al. [16], Moran’s I was used to investigate neural networks, which were identified using segmentation techniques.
Protecting patient privacy against unauthorized release of medical images in a group communication environment
2005, Computerized Medical Imaging and GraphicsAn FCM-Based Image De-Noising with Spatial Statistics Pilot Study
2023, Applied Sciences (Switzerland)A FCM-based image de-noising with spatial statistics pilot study
2022, Research SquareCompressed medical image quality determination using the kolmogorov- smirnov test
2017, Current Medical Imaging Reviews
Tzong-Jer Chen currently is a Postdoctoral Researcher at Department of Nuclear Medicine, National PET/Cyclotron Center, Taipei Veterans General Hospital, Taipei, Taiwan. He received his PhD in Medical Physics and MS in Health Physics both from the National Tsing-Hua University, Taiwan. His research interests include image analysis, medical image quality evaluation and inverse planning for radiation therapy.
Keh-Shih Chuang received both his BS in Nuclear Engineering and MS in Health Physics from the National Tsing-Hua University, Taiwan, and his PhD in Medical Physics from UCLA. He is currently an Associate Professor at the Department of Nuclear Science in the National Tsing-Hua University, Taiwan. His research interests are quality evaluation of medical image, scatter correction in PET image, and inverse planning for radiation therapy.
Yuang-Chin Chiang received his PhD degree in Statistics in 1988 from Florida State University, USA. He is currently an Associate Professor in the Institute of Statistics of the National Tsing-Hua University (Taiwan). His major research interest at present is in speech recognition, especially that of the Taiwanese language.
Jen-Hao Chang received his BS in Radiological Science and Technology from the National Yang-Ming University, Taiwan. Then he received his MS in Medical Physics from Department of Nuclear Science, National Tsing-Hua University, Taiwan. He is currently an Associate Researcher of the Biomedical Engineering Center, Industrial Technology Research Institute (ITRI), Taiwan. His interests are medical imaging, medical engineering, bio-photonics, radiation applications in biomedical technology.
Ren-Shyan Liu received his MD degree in medicine from National Defense Medical Center in Taiwan in 1997. From September 1997, he works at the Department of Nuclear Medicine, Taipei Veterans General Hospital and also was a faculty at National Defense Medical Center and National Yang-Ming University Medical School. He is now a staff physician at the Department of Nuclear Medicine, Taipei Veterans General Hospital, and the Associate Professor and Chief of Division of Nuclear Medicine at National University Medical School. He is also in charge of PET Gene Probe Core for National Research Program of Genomic Medicine in Taiwan. He is on the Editorial Board of Journal of Nuclear Medicine, European Journal of Nuclear Medicine, and Annals of Nuclear Medicine. His current research interests include molecular image, gene expression imaging, and application of PET in oncology, especially tumor hypoxia and C-11 acetate tumor imaging.