Tongue Coat Information Extraction of the Traditional Chinese Medicine with Hyperspectral Image

Tongue coat information extraction plays an important role in disease diagnosis for the traditional Chinese medicine. For the purpose of quantifying the tongue coat properties in traditional Chinese medicine, most of the existing methods are based on computerized image analysis, carrying out RGB color space in a tongue imaging captured with digital camera. However, those methods cannot meet the requirements of clinical medicine. To explore more information about the tongue objectively, a new approach to analyze tongue information based on hyperspectral images is presented. Hyperspectral images are acquired using the hyperspectral imaging system in the spectral range of 370. 200 0 to 992. 956 0 nm (343bands), and the traditional Chinese medicine clinical diagnosis information is recorded. The main region of interest (ROI) in the samples is extracted while the background is removed from the tongue image, then tongue information of ROI is analyzed. The largest different spectral characteristics between tongue proper and tongue coat are found in the wavelength range of about 525 to 600 nm. Nine wavelengths tongue spectral images from 382. 108 0 to 963. 668 0 nm are extracted, then comparing with the actual tongue situation, we find that the spectral image at 527. 548 0 nm band can better reflect the actual tongue situation than others. The experiment results show that hyperspectral imaging technique is very helpful for tongue coat information extraction of the traditional Chinese medicine, and this new analysis approach can provide a fast and simple non-invasive detection method for tongue coat segmentation and tongue coat information extraction.