《外文翻译-- Sublingual Veins Extraction Method Based on Hyperspectral Tongue Images》由会员分享,可在线阅读,更多相关《外文翻译-- Sublingual Veins Extraction Method Based on Hyperspectral Tongue Images(3页珍藏版)》请在金锄头文库上搜索。
1、Sublingual Veins Extraction Method Based on Hyperspectral Tongue Images Qingli Li, Yiting Wang, Hongying Liu Key Laboratory of Polor Materials and Devices East China Normal University Shanghai 200241, China Zhen Sun Department of Traditional Chinese Medicine Second Military Medical University Shang
2、hai, China AbstractSublingual vein is one of the important features on tongue surface, which may have pathological relationship with some diseases. Extracting sublingual veins accurately is the primitive work of computer-aided tongue disease diagnosis. Most existing sublingual veins extraction metho
3、ds are using sublingual images captured by traditional CCD cameras. However, these conversional methods impede the accurate analysis on the subjects of sublingual veins because of the limited information of the images. To solve these issues, a hyperspectral tongue imaging system instead of a digital
4、 camera is used to capture sublingual images. Then an improved spectral angle mapper (ISAM) algorithm for automatic sublingual veins extraction was presented. In this algorithm, the spectral of sublingual veins were extracted and the spectral angles of all bands and partial bands were calculated res
5、pectively. Finally, the sublingual veins were extracted according to the spectral angles. The experimental results demonstrate that this algorithm can extract the sublingual veins more accurately. Keywords- biomedical imaging; hyperspectral imaging; image segmentation; sublingual veins; spectral ang
6、le mapper algorithm I. INTRODUCTION Human tongue is one of the important organs of the body, which carries abound of information of the health status 1. Among the various information of tongue, Sublingual veins is the most important clinical symptoms. Sublingual veins distributing over lower surface
7、 of tongue directly connects with viscera organs and blood through channels 2. Therefore, Inspection of sublingual veins can provide valuable insights into the healthy condition of humans. However, the subjective characteristic of traditional method impedes this objective in that sublingual vein dia
8、gnosis is usually based on detailed visual discrimination, which mainly depends on the subjective analysis of the examiners 3. Nowadays, the rapid progress of information technology promotes the automatization of tongue disease diagnosis based on modern image processing and pattern recognition appro
9、aches 4, 5. Among these techniques, automatic extraction of sublingual veins out of complex scenes should be foremost solved in computerized tongue diagnosis. This is due to the qualities of segmentation directly influencing on the subsequent feature extraction and recognition. There have been some
10、experiments implemented on sublingual images acquired by ordinary camera under visible light source or infrared ones for sublingual veins extraction 6-8. Although many issues of standardization and quantification have been resolved, there are still some difficulties because of the limitations of the
11、se kinds of images. For example, the thickness and transparence of sublingual mucosa which covered on sublingual veins may change due to different degree of varicosity. This change may lead to that in some sublingual images the sublingual veins are clear, but blurry in others and the contours of sub
12、lingual veins are difficult to be extracted out 7. To acquire high-quality sublingual images that retain more invariable information of sublingual veins, a hyperspectral tongue imaging system was developed and used in this paper. Then a sublingual veins extraction algorithm based on hyperspectral su
13、blingual images was proposed. Unlike existing approaches, our new method can recognize sublingual veins using their spectral signatures rather than their gray values. The results of experiment show that this is an effective method for sublingual veins extraction. II. MATERIALS AND METHODS A. Hypersp
14、ectral tongue imaging system The hyperspectral tongue imaging system is specially designed according to the pushbroom hyperspectral imager commonly used in remote sensing, which has high spectrum resolution and high spatial resolution 9. The new system represents a hybrid modality for optical diagno
15、stics, which obtains spectroscopic information and renders it in image form. As shown in figure 1, the hyperspectral images provided by the new system can be visualized as a 3D cube or a stack of multiple 2D images because of its intrinsic structure, where the cube face is a function of the spatial
16、coordinates and the depth is a function of wavelength. In this case, each spatial point on the face is characterized by its own spectra. The spectrum range of the system is 400800 nm and the spectral resolution is better than 5 nm. Consequently, we can analyze sublingual veins from both the spatial
17、and spectral angle. B. Sublingual veins extraction algorithm In the area of computer-assisted medical tongue diagnosis, research on the feature extraction of tongue surface has achieved considerable progress 10, 11. However, the 978-1-4244-4713-8/10/$25.00 2010 IEEEsublingual vein diagnosis extracti
18、on, which is also one important part of tongue diagnosis, is rarely referred. As the extraction accuracy can directly influence the results of classification and recognition, algorithm should be designed to extract sublingual veins from hyperspectral sublingual images by using both spectral and spat
19、ial information. In this paper, we use the improved spectral angle mapper (ISAM) algorithm to segment the sublingual veins. Spectral angle mapper (SAM) algorithm is a tool that permits rapid mapping of spectral similarity of one image spectrum to another spectrum 12-14. The algorithm determines the
20、spectral similarity between two spectra by calculating the angle between them. The angle between the endmember spectra vector and each pixel vector in N-dimensional space is compared. Smaller angles represent closer matches to the reference spectra. The spectral angle (SA) between two spectral vecto
21、rs T and R can be calculated by the following formula =NiiNiiNiiirtrtRTRTRT1212111cos,cos),(? (1) where N is the total number of bands. This algorithm can extract the target from hyperspectral images effectively by mapping the spectral similarity. However, the wavelengths often shift several bands w
22、ith the influence of noise in the real hyperspectral tongue imaging system which lead to some extraction errors. To overcome this disadvantage, we used an improved spectral angle mapper (ISAM) algorithm in this paper. The ISAM algorithm calculates the SA not only according to formula (1), but also b
23、etween the reference spectral vector and the testing spectral vector with shift forward and backward 2 bands, respectively. Then the maximum was selected as the real SA value between the two vectors as the following ()=+=+NiiNijiNiijijrtrtMaxRTMaxSA121211cos),(? (2) where j = -2, -1, 0, 1, 2. This m
24、ethod is insensitive to illumination since the ISAM algorithm uses only the vector direction and not the vector length. It also can reduce the wavelengths shift errors effectively. The result of the ISAM sublingual vein extraction is an image showing the best match at each pixel. III. RESULTS The hy
25、perspectral tongue images can provide not only spatial but also spectral information to automatic tongue diagnosis system. Figure 2 illustrates a representative subtotal of spectral images captured at various wavelengths by using the hyperspectral tongue imaging system. From the figure we can see th
26、at different tongue features would be obvious at corresponding wavelength band. As we have presented, automatic sublingual veins extraction is very important for computer-aided tongue disease diagnosis. After capturing the hyperspectral sublingual images, we can extract sublingual veins accurately b
27、y the new algorithms which can utilize both spatial and spectral information. Figure 3 shows the extraction results by the traditional SAM algorithm and the improved SAM algorithm with SA = 0.1. From the figures it can be seen that the ISAM algorithm can segment the cells more accurately than the tr
28、aditional method. The experimental results also show that the hyperspectral imaging technology has good application prospect in optical diagnostics. IV. CONCLUSIONS Recent research demonstrates that some features on tongue are pathological forms and thereby have clinical significance, for example, t
29、he sublingual veins 15. In this paper, we use the hyperspectral imaging system to capture sublingual images, which can provide both spectral and spatial information of sublingual veins to assist computerizing the sublingual vein diagnosis. Furthermore, the corresponding extraction method of sublingu
30、al veins in the captured hyperpsectral sublingual images is proposed. Experimental results show that, the proposed ISAM algorithm for hyperspectral sublingual images Figure 1. hyperspectral sublingual image data cube. The face of the 3D data cube is a false color image with the 626.9 nm, 541.3 nm, a
31、nd 510.4 nm single-band images. Figure 2. Single-band images at various wavelengths did indeed segment the sublingual veins with an acceptable degree of accuracy. ACKNOWLEDGMENT This work is supported in part by the National Natural Science Foundation of China (Grant No. 60807035, 60976004), the Spe
32、cialized Research Fund for the Doctoral Program of Higher Education of China (Grant No. 200802691006), and the Project supported by the Shanghai Committee of Science and Technology, China (Grant No. 09JC1405300). REFERENCES 1 Q. L. Li and Z. Liu, Tongue color analysis and discrimination based on hyp
33、erspectral images, Computerized Medical Imaging and Graphics, vol. 33, pp. 217-221, 2009. 2 Q. Chen, Z. Xu, and Y. Chai, General situation on the study of inspection of sublingual veins, Journal of Traditional Chinese Medicine, vol. 24, pp. 133-136, 2004. 3 Z. Yan, M. Yu, K. Wang, and N. Li, Subling
34、ual vein segmentation from near infrared sublingual images, Journal of Computer Aided Design & Computer Graphics vol. 20, pp. 1569-1574, 2008. 4 B. Pang, Z. David, and K. Q. Wang, Tongue image analysis for appendicitis diagnosis, Inf. Sci., vol. 175, pp. 160-176, 2005. 5 B. L. Pham and Y. Cai, Visua
35、lization techniques for tongue analysis in traditional Chinese medicine, in roceedings of the SPIE, 2004, pp. 171-180. 6 C.-C. Chiu, C.-Y. Lan, and Y.-H. Chang, Objective assessment of blood stasis using computerized inspection of sublingual veins, Computer Methods and Programs in Biomedicine, vol.
36、69, pp. 1-12, 2002. 7 Z. Yan, K. Wang, and N. Li, Segmentation of sublingual veins from near infrared sublingual images, in Segmentation of sublingual veins from near infrared sublingual images Athens, Greece: IEEE, Piscataway, NJ, USA, 2008, p. 5. 8 Z. Yan, K. Wang, and N. Li, Computerized feature
37、quantification of sublingual veins from color sublingual images, computer methods and programs in biomedicine, vol. 93, pp. 192-205, 2009. 9 Q. L. Li, G. H. Xiao, and Y. Q. Xue, Hyperspectral tongue imaging system used in tongue diagnosis, in 2nd International Conference on Bioinformatics and Biomed
38、ical Engineering, iCBBE 2008 Shanghai, China: Institute of Electrical and Electronics Engineers Computer Society, Piscataway, NJ 08855-1331, United States, 2008, pp. 2579-2581. 10 S. Lukmana, Y. Heb, and S.-C. Huic, Computational methods for Traditional Chinese Medicine: A survey, Computer Methods a
39、nd Programs in Biomedicine, vol. 88, pp. 283-294, 2007. 11 B. Pang, D. Zhang, and N. M. Li, Computerized tongue diagnosis based on bayesian networks, IEEE Trans. On Biomedical Eng., vol. 51, pp. 1803-1810, 2004. 12 F. A. Kruse, A. B. Lefkoff, and J. W. Boardman, The spectral image processing system
40、(SIPS)-software for integrated analysis of AVIRIS data, in Summaries of the 4th Annual JPL Airborne Geoscience Workshop, Pasadena, 1992, pp. 23-25. 13 F. A. Kruse, A. B. Lefkoff, J. W. Boardman, K. B. Heidebrecht, A. T. Shapiro, P. J. Barloon, and A. F. H. Goetz, The spectral image processing system
41、 (SIPS) - interactive visualization and analysis of imaging spectrometer data Remote Sensing of Environment, vol. 44, pp. 145-163, 1993. 14 B. Park, W. R. Windhama, K. C. Lawrencea, and D. P. Smitha, Contaminant classification of poultry hyperspectral imagery using a spectral angle mapper algorithm,
42、 Biosystems Engineering, vol. 96, pp. 323-333, 2007. 15 J. Chen and C. Hu, On factors related to abnormality of sublingual vein in 530 cancer patients, Chin J Integr Med, vol. 8, pp. 590-592, 1988 Figure 3 Extraction results. (a)(b) false color image of three single bands; (c)(d) SAM results; (e)(f) ISAM results.