Abstract
Accurate segmentation of the tongue body is an important prerequisite for computer-aided tongue diagnosis. In general, the size and shape of the tongue are very different, the color of the tongue is similar to the surrounding tissue, the edge of the tongue is fuzzy, and some of the tongue is interfered by pathological details. The existing segmentation methods are often not ideal for tongue image processing. To solve these problems, this paper proposes a symmetry and edge-constrained level set model combined with the geometric features of the tongue for tongue segmentation. Based on the symmetry geometry of the tongue, a novel level set initialization method is proposed to improve the accuracy of subsequent model evolution. In order to increase the evolution force of the energy function, symmetry detection constraints are added to the evolution model. Combined with the latest convolution neural network, the edge probability input of the tongue image is obtained to guide the evolution of the edge stop function, so as to achieve accurate and automatic tongue segmentation. The experimental results show that the input tongue image is not subject to the external capturing facility or environment, and it is suitable for tongue segmentation under most realistic conditions. Qualitative and quantitative comparisons show that the proposed method is superior to the other methods in terms of robustness and accuracy.
Highlights
Tongue diagnosis is one of the important diagnostic methods of traditional Chinese medicine, while for a long time, tongue diagnosis relied on the doctor’s clinical experiences by short-term visual observation, which causes the subjective and uncertain diagnosis results
Osher and Sethian [31, 32] proposed a level set method based on the important idea of fluid, which solved the problem that the topological structure is not easy to change during image segmentation. e level set method implicitly represents the closed active contour as a zero level set of a higher dimensional level set function and uses the curve evolution to locate the edge of the target
The tongue image dataset contains 550 tongue images, part of which is from GitHub’s open-source dataset, with a total of 300 tongue images; the other part is provided by the teachers of the University of traditional Chinese medicine, with a total of 250 tongue images. e images in the dataset are different in size, shape, angle, and position, but they all contain the complete tongue body, which is suitable for this experiment
Summary
Tongue diagnosis is one of the important diagnostic methods of traditional Chinese medicine, while for a long time, tongue diagnosis relied on the doctor’s clinical experiences by short-term visual observation, which causes the subjective and uncertain diagnosis results. Traditional methods often focus on segmentation based on image features and variable models, and the level set model is one of the most representative methods of the active contour model. Yang [23] presented a gradient vector active contour model based on the original tongue edge detection method and color gradient and obtained a good segmentation effect. There are other tissues such as peri lip in the image, and the color features of these parts are very close to the tongue itself, which results in the slow change of the gradient of the tongue edge. Is leads to problems such as incomplete segmentation and boundary leakage in level set methods that rely on active contour models or gradient information to extract edges, and the accuracy of segmentation results is difficult to guarantee. It is proved by experiments that this method can complete automatic precise tongue segmentation suitable for most real situations
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
