Tongue body self-diagnosis health cloud service system based on deep convolution neural network
A convolutional neural network, neural network technology, applied in the field of self-service health care
Active Publication Date: 2017-01-04
汤一平
9 Cites 138 Cited by
AI-Extracted Technical Summary
Problems solved by technology
[0031] To sum up, there are still several thorny problems in using the deep learning-based convolutional neural network for "tongue diagnosis" as follows: 1) How to accurately segment the overall image of the tongue from the complex background; 2) How to use as little tagged tongue image data as possible to accurately obtain various characteristic data of the tongue; 3) How to build...
Method used
What shown in Fig. 3 is convolutional neural network figure, is divided into eight layers altogether, convolutional neural network is the depth structure that is formed alternately by convolution layer, activation layer and down-sampling layer, and this depth structure can effectively reduce Compute time and establish invariance on the spatial structure. The input image is mapped layer by layer in the network, and finally different representations of each layer for the image are obtained, and the depth representation of the image is realized. The method of convolution kernel and downsampling directly determines the mapping method of the image.
[0195] In addition, Chinese medicine has the saying of "prolonged disease and kidney", which means that when the functional decline of each viscera reaches a certain level, the function of the kidney will be damaged. Kidney deficiency can also be seen in elderly and infirm normal people. People with kidney deficiency have different syndromes such as kidney yin deficiency, kidney yang deficiency, and deficiency of kidney essence. Those who have a pale tongue and see tenderness beside the tongue usually belong to the syndrome of deficiency of kidney yang. Insufficient yang qi makes the body afraid of cold, so it is necessary to thicken the clothes and quilts to keep warm, and avoid places with large temperature differences to prevent colds. Exercising regularly in a sunny place can invigorate the body and improve Yang Qi; bathing in hot springs is also a fitness method to warm the Qi, activate blood and dredge collaterals. Winter is the season when all things in nature are stored. Drinking ginseng in winter can warm the yang and replenish qi, support righteous qi, get rid of pathogenic factors, and strengthen the body's immunity. Choose the appropriate variety according to different physiques. Red ginseng is warm in nature, suitable for the elderly and those with yang deficiency. Young people or those without obvious chills can choose white ginseng, ...
Abstract
The invention discloses a tongue body self-diagnosis health cloud service system based on a deep convolution neural network. The system mainly comprises a convolution neural network used for deep learning and training recognition, a tongue body segmentation method based on the deep convolution neural network, the deep convolution neural network used for tongue image classification and a health cloud service platform used for conducting self-diagnosis according to a recognized tongue image type. According to the tongue body self-diagnosis health cloud service system based on the deep convolution neural network, the automation and intelligent level of watching the tongue for diagnosis on the basis of a mobile internet can be effectively improved, more people can understand and participate in self-service health detection, evaluation and guidance, the health consciousness of the public is improved, and the self-health management ability is increased.
Application Domain
Medical automated diagnosisSpecial data processing applications +2
Technology Topic
CrowdsTongue body +12
Image
Examples
- Experimental program(2)
Example Embodiment
[0107] Example 1
[0108] Reference Figure 1~24 The technical solutions adopted by the present invention to solve its technical problems are:
[0109] The tongue self-diagnosis health cloud service system based on deep convolutional neural network includes a convolutional neural network for deep learning and training recognition, a tongue segmentation method based on full convolutional neural network, and a tongue A deep convolutional neural network for image classification and a cloud service platform for self-diagnosis and health based on the recognized tongue type; the block diagram of the self-diagnosis and health cloud service platform for tongue is as follows figure 1 Shown
[0110] The use and preparation of the tongue self-diagnosis health cloud service system: When using a mobile phone or other mobile device to take pictures of the tongue, the user first cleans the mouth with drinking water to prevent food and medicine from contaminating the tongue coating, and then opens the camera software on the smartphone , And turn on the lighting on the phone; then naturally extend the tongue out of the mouth, relax the tongue, flatten the tongue with the tip of the tongue slightly downward, fully exposing the tongue; take the image of the tongue; finally pass the tongue image on the phone Send your WeChat or MMS or QQ to the health cloud service platform;
[0111] (1) About designing a convolutional neural network for deep learning and training recognition
[0112] Convolutional neural network is essentially a deep-mapped network structure, such as figure 2 As shown, the input signal is continuously decomposed and expressed through layer-by-layer mapping in the network, and finally forms a multi-layer expression about the tongue image. Its main feature is that it does not need to manually select and construct the tongue image features, but through the machine Automatically learn to get a deep representation of the tongue image.
[0113] The first layer: such as Figure 4 As shown, the input image data is a 224×224 pixel image, which is divided into 3 components in the RGB color space, the filling value is 3, and the output data is 227×227×3; then after 96 filters, the window size is 11×11 , Convolutional layer 1 with a step size of 4 is processed, and [(227-11)/4]+1=55 features are obtained. The subsequent layers are divided into two groups of processing, and the output features are 55×55×96, and then proceed ReLU activation layer 1 processing, the output feature is 55×55×96, after the pooling layer 1, the largest pooling 3×3 core, the step size is 2, we get [(55-3+1)/2]+1= 27 features, the total number of features is 27×27×96, then regularization is performed, the number of channels used for summation is 5, and finally 27×27×96 data is obtained;
[0114] The second layer: such as Figure 5 As shown, the input data is 27×27×96, the filling value is 2, 256 filters, the window size is 5×5, and [(27-5+2×2)/1]+1=27 features are obtained, and output The feature is 27×27×256, and then the ReLU activation layer 2 is processed, and the output feature is 27×27×256. After the pooling layer 2, the maximum pooling 3×3 core is obtained, and the step size is 2, and the [(27- 3)/2]+1=13 features, the total number of features is 13×13×256, and then regularization is performed, the number of channels used for summation is 5, and finally 13×13×256 data is obtained;
[0115] The third layer: such as Image 6 As shown, the input data is 13×13×256, the filling value is 1,384 filters, the window size is 3×3, and [(13-3+1×2)/1]+1=13 features are obtained, and output The feature is 13×13×384, and then the ReLU activation layer 3 is processed, and finally 13×13×384 data is obtained;
[0116] The fourth layer: such as Figure 7 As shown, the input data is 13×13×384, the filling value is 1,384 filters, the window size is 3×3, and [(13-3+2×1)/1]+1=13 features are obtained, and output The feature is 13×13×384, and then the ReLU activation layer 4 is processed, and finally 13×13×384 data is obtained;
[0117] The fifth layer: such as Picture 8 As shown, the input data is 13×13×384, the filling value is 1,256 filters, the window size is 3×3, and [(13-3+2×1)/1]+1=13 features are obtained, and output The feature is 13×13×256, and then the ReLU activation layer 5 is processed, and the output feature is 13×13×256. After the pooling layer 5, the largest pooling 3×3 core is performed, and the step size is 2, and the [(13- 3)/2]+1=6 features, the total number of features is 6×6×256, and finally 6×6×256 data is obtained;
[0118] Sixth layer: such as Picture 9 As shown, the input data is 6×6×256, fully connected, and 4096 features are obtained, and then the ReLU activation layer 6 processing is performed, and the output feature is 4096. After dropout6 processing, 4096 data is finally obtained;
[0119] Seventh layer: such as Picture 10 As shown, the input data is 4096, fully connected, and 4096 features are obtained, and then the ReLU activation layer 7 processing is performed, and the output feature is 4096. After dropout7 processing, 4096 data is finally obtained;
[0120] The eighth layer: such as Picture 11 As shown, the input data is 4096, fully connected, and 1000 characteristic data are obtained;
[0121] The prediction process of the convolutional neural network is a forward propagation process. The output of the previous layer is the input of the current layer and is passed layer by layer through the activation function. Therefore, the actual calculation output of the entire network is expressed by formula (1),
[0122] O p = F n (...(F 2 (F 1 (XW 1 )W 2 )...)W n ) (1)
[0123] In the formula, X represents the original input, F l Represents the activation function of the lth layer, W l Represents the mapping weight matrix of the lth layer, O p Represents the actual calculation output of the entire network;
[0124] The output of the current layer is represented by (2),
[0125] X l = F l (W l X l-1 +b l ) (2)
[0126] In the formula, l represents the number of network layers, X l Represents the output of the current layer, X l-1 Represents the output of the previous layer, that is, the input of the current layer, W l Represents the already-trained mapping weight matrix of the current network layer, b l Is the additive paranoia of the current network, f l Is the activation function of the current network layer; the used activation function f l In order to correct the linear unit, namely ReLU, it is expressed by formula (3),
[0127] f l = m a x ( ( W l ) T X l , 0 ) = ( W l ) T X l ( W l ) T X l 0 0 ( W l ) T X l ≤ 0 - - - ( 3 )
[0128] In the formula, l represents the number of network layers, W l Represents the already-trained mapping weight matrix of the current network layer, f l Is the activation function of the current network layer; its function is to let it be 0 if the convolution calculation result is less than 0; otherwise, keep its value unchanged.
[0129] Convolutional neural network training is a back propagation process, similar to the BP algorithm, through error function back propagation, using stochastic gradient descent method to optimize the convolution parameters and bias until the network converges or reaches the maximum number of iterations to stop.
[0130] The neural network training is a back propagation process, through the error function back propagation, using the stochastic gradient descent method to optimize the convolution parameters and bias adjustments, until the network converges or reaches the maximum number of iterations to stop;
[0131] Backpropagation needs to compare labeled training samples and use a square error cost function. For c categories and N training samples, multiple categories are recognized. The final output error function of the network uses formula (4) to calculate the error.
[0132] E N = 1 2 X n = 1 N X k = 1 c ( t k n - y k n ) 2 - - - ( 4 )
[0133] Where E N Is the square error cost function, Is the kth dimension of the label corresponding to the nth sample, The nth sample corresponds to the kth output predicted by the network;
[0134] When backpropagating the error function, a calculation method similar to the traditional BP algorithm is used, as shown in formula (5),
[0135] δ l = ( W l + 1 ) T δ l + 1 X f ′ ( u l ) u l = W l x l - 1 + b l - - - ( 5 )
[0136] Where δ l Represents the error function of the current layer, δ l+1 Represents the error function of the previous layer, W l+1 Is the upper level mapping matrix, f'represents the inverse function of the activation function, that is, upsampling, u l Indicates the output of the previous layer that failed the activation function, x l-1 Represents the input of the next layer, W l Mapping the weight matrix for this layer.
[0137] The algorithmic thinking of convolutional neural network learning and training is: 1) First, build a single layer of neurons layer by layer, so that a single layer network is trained each time; 2) After all layers are trained, use the wake-sleep algorithm to adjust excellent.
[0138] The deep learning training process is as follows:
[0139] STEP21: Use bottom-up unsupervised learning, that is, start from the bottom layer and train layer by layer to the top layer to learn tongue image features: first train the first layer with unlabeled tongue image data, and learn the first layer during training. The parameters of one layer, due to the limitation of model capacity and sparsity constraints, enable the obtained model to learn the structure of the data itself, thereby obtaining features that are more expressive than the input; after learning the l-1th layer, change l The output of the -1 layer is used as the input of the first layer, and the first layer is trained to obtain the parameters of each layer; the specific calculation is shown in formulas (2) and (3);
[0140] STEP22: Top-down supervised learning, that is, training through labeled tongue image data, the error is transmitted from top to bottom, and the network is fine-tuned: the specific calculation is shown in formulas (4) and (5);
[0141] Based on the parameters of each layer obtained by STEP21, further fine-tune the parameters of a multi-layer model. This step is a supervised training process; STEP21 is similar to the random initialization process of neural networks, because the STEP21 of deep learning is not random initialization, but through learning input The structure of the data is obtained, so this initial value is closer to the global optimum, which can achieve better results.
[0142] The tagged tongue image data here is the key to "seeing the tongue for diagnosis". It is necessary for a senior TCM doctor to screen the various tongue pictures collected. Experts check the color of the tongue, the thickness of the tongue and the thickness of the tongue coating. Identify and classify the number, red tongue and petechiae; the specific method is that three senior TCM doctors with more than 20 years of clinical diagnosis experience are responsible for determining the category label of each sample; this is based on the expert’s tongue identification experience and opinions Perform analysis and synthesis to obtain more scientific and accurate classification basis and diagnosis results of tongue features; provide training and learning tongue image data for deep convolutional neural networks;
[0143] In the present invention, the tongue image pictures after tongue identification by experts are labeled, and then these labeled tongue image images are learned by the deep convolutional neural network, and the labeled tongue image features are automatically extracted; the tongue image features include The color of the tongue, the color of the tongue, the thickness of the tongue, the moisture (dryness), the texture (soily), the shape, the state of the tongue, the veins of the tongue;
[0144] Experimental studies have shown that the larger the tongue data set, the richer the tongue image sample categories, the more accurate the tongue image classification; therefore, it is a key to make a labeled tongue image data set;
[0145] The preparation of the tongue data set; one type of data obtains the tongue image data with labels through the professional books of Chinese medicine. This kind of data can be directly used as the data in the tongue data set, such as "Tongue Diagnosis of Chinese Medicine" published by Beijing College of Traditional Chinese Medicine, Chen Zelin "Research on Tongue Diagnosis" and "Tongue Coating Atlas of Traditional Chinese Medicine" compiled by Song Tianbin; the other is that we have collected various tongue image data through crawling software. The search results are basically with keywords. The labeled tongue image data, we use it as training data; another type of data is that we use the cooperation with senior Chinese medicine practitioners to label the collected tongue image data samples;
[0146] On the basis of the above-mentioned tongue image data, the amount of input data is increased by one or a combination of the following data enhancement and transformation techniques; ①Rotation|Reflection transformation: Randomly rotate the image by a certain angle to change the orientation of the image content; ②Flip transformation: Flip the image along the horizontal or vertical direction; ③Zoom transformation: zoom in or reduce the image according to a certain ratio; ④Translation transformation: translate the image in a certain way on the image plane; ⑤You can use random or artificially defined methods to specify the translation range Pan step length, pan in the horizontal or vertical direction to change the position of the image content; ⑥Scale transformation: zoom in or out the image according to the specified scale factor; or refer to the SIFT feature extraction idea and use the specified scale factor to Image filtering constructs scale space; changes the size or blur degree of the image content; ⑦contrast transformation: in the HSV color space of the image, changes the saturation S and V brightness components, keeping the hue H unchanged; for the S and V components of each pixel Carry out exponential calculation, the exponent factor is between 0.25 and 4, increasing the light change; ⑧ noise disturbance: random disturbance to each pixel RGB of the image; common noise modes are salt and pepper noise and Gaussian noise; ⑨ color transformation: in the training set PCA is performed on the RGB color space of the pixel value to obtain 3 main direction vectors and 3 eigenvalues of the RGB space, p1, p2, p3, λ1, λ2, λ3; each pixel of each image Ixy=[IRxy,IGxy, IBxy] T Add the following changes: [p1,p2,p3][α1λ1,α2λ2,α3λ3] T;
[0147] In the tongue image data set, the color of the tongue, the color of the tongue, the thickness of the tongue, the humidity, the texture, the shape, the state of the tongue, and the vein of the tongue should be reflected in the tongue image of the sample label; the color category of the tongue Including pale tongue, pale red tongue, red tongue, dark red tongue, crimson tongue, dark purple tongue, thin white coating, white coating, white thick coating, thin yellow coating, yellow coating, yellow thick coating, gray coating, brown coating, black coating Coating, etc.; the thickness of the tongue coating; tongue types include oval tongue, square tongue, rectangular tongue, round tongue, sharp triangle tongue, blunt triangle tongue and hammer tongue, etc.; tongue tip redness, petechiae, and petechiae; these samples label tongue The quality of the body image will directly affect the accuracy of "Tongue Diagnosis";
[0148] Strictly speaking, everyone’s tongue image is different. With the expansion of the self-diagnosis health cloud service platform, the tongue image data will be a very large amount of data, which can be summarized through the processing of big data. For some new types of tongue signs, of course, senior Chinese medicine practitioners must participate in this process;
[0149] (2) Regarding the design of a tongue segmentation method based on a fully convolutional neural network;
[0150] Since when using a mobile phone to take images of the tongue, the acquired images often include the tongue, lips, and the skin around the mouth. Segmenting the tongue from the collected tongue images is an important prerequisite for tongue diagnosis, so it must be designed A tongue segmentation algorithm based on fully convolutional neural network;
[0151] First, design a tongue segmentation algorithm based on a fully convolutional neural network, that is, to select and locate the tongue object in the image;
[0152] In order to locate the position of the tongue object; since the tongue object may appear in any position of the image, and the size and aspect ratio of the tongue object are also uncertain, the original technique is to initially use the sliding window strategy to adjust the whole frame The image is traversed, and different scales and different aspect ratios need to be set; although this exhaustive strategy includes all possible positions of the tongue target, the disadvantage is also obvious: the time complexity is too high and redundancy is generated. There are too many windows, which also seriously affects the speed and performance of subsequent feature extraction and classification; therefore, how to use semantic concepts to locate and segment tongue objects is very important;
[0153] An important advantage of deep convolutional neural network is to extract information layer by layer from pixel-level raw data to abstract semantic concepts, which makes it have outstanding advantages in extracting global features and context information of images, and brings solutions to image semantic segmentation. Breakthrough; the higher the number of layers of the convolutional neural network, the more it can express the global features and semantic concepts of the image, but the deep convolutional neural network through multi-layer downsampling makes the image with the higher the number of convolutional neural network layers smaller than the original image Times, if the highest layer of the convolutional neural network is used as segmentation prediction, the segmented objects will be rougher and generally have rough outlines. The tongue object obtained in this way will seriously affect the accuracy of subsequent tongue diagnosis; The invented tongue segmentation algorithm based on the full convolutional neural network is based on the convolutional neural network. The convolutional neural network is first introduced below;
[0154] image 3 Shown is the convolutional neural network diagram, which is divided into eight layers. The convolutional neural network is a deep structure composed of convolutional layers, activation layers, and downsampling layers. This deep structure can effectively reduce calculation time and establish space Structural immutability. The input image is mapped layer by layer in the network, and finally the different representations of each layer for the image are obtained, and the depth representation of the image is realized. The convolution kernel and the down-sampling method directly determine the image mapping method.
[0155] In order to accurately segment the tongue object, the main idea of the present invention is to change the deep convolutional neural network to a fully convolutional neural network, namely FCN. After inputting an image, the dense prediction is directly obtained at the output end, that is, each pixel belongs to The class of to obtain an end-to-end method to achieve semantic segmentation of tongue object images;
[0156] After the image including the tongue body is convolved multiple times by the deep convolutional neural network, the resulting image is getting smaller and smaller, and the resolution is getting lower and lower. So how does FCN get the category of each pixel in the image? In order to recover from this low-resolution rough image to the original resolution, FCN uses upsampling. For example, after 5 times of convolution, the resolution of the image is reduced by 2, 4, 8, 16, and 32 times. For the output image of the last layer, 32 times of upsampling is required to get the original Figure one Kind of size like Figure 14 As shown, in the present invention, a step size of 32 is used to upsample the output image of the last layer; for the output image of the last second layer, 16 times of upsampling is required to obtain the original Figure one Kind of size like Figure 15 As shown, in the present invention, a step size of 16 is used to up-sample the output image of the last second layer; for the output image of the last third layer, an 8-fold up-sampling is required to obtain the original Figure one Kind of size like Figure 16 As shown, the present invention uses a step size of 8 to upsample the output image of the last third layer; the upsampling operation here can be regarded as deconvolution, and the parameters of the convolution operation are the same as those of CNN in training FCN In the process of the model, it is learned through BP algorithm;
[0157] In order to accurately predict the segmentation result of each pixel, the present invention divides the tongue object positioning and segmentation algorithm into large to small (that is, from the input large image to the small image after positioning and classification), and then from small to large ( The two processes consistent with the original input image size); from large to small is caused by the action of the down-sampling layer in the deep convolutional neural network, and from small to large, it needs to be realized by the up-sampling layer; in the up-sampling process In the present invention, the method of increasing in stages is adopted, and in each stage of upsampling, the features of the down-sampling corresponding layer are used for assistance; the so-called assistance is to adopt the method of layer jump to reduce the step of upsampling at the shallow layer. Long, the fine layer obtained is fused with the coarse layer obtained from the high layer, and then up-sampled to get the output; this layer-jumping method takes into account both local and global information;
[0158] First put image 3 The fully connected layers of the convolutional neural network shown in the figure, the sixth, seventh and eighth layers in the figure, here are used as convolutional layers, the convolutional template size is the size of the input feature map, that is to say Think of the fully connected network as convolving the entire input image. The fully connected layer has 4096 1×1 convolution kernels, 4096 1×1 convolution kernels, and 1000 1×1 convolution kernels. ;
[0159] Figure 13 The output shown is 1000 1×1 convolution kernels, the last two stages are fully connected, and the parameters are discarded;
[0160] Figure 14 As shown, the 7th layer 1×1×4096 feature map is predicted to be divided into 16×16×3 small images, and then directly up-sampled to a 500×500×3 large image; here 500×500 is the size of the original image In the present invention, the same size of the original image can be recovered according to the size of the original image; 3 is the depth value, here means 2 types of targets + 1 background; the step size of the deconvolution is 32, and this network is called FCN- 32s;
[0161] Figure 15 As shown, the up-sampling is divided into two completions; before the second up-sampling, the prediction results of the fourth pooling layer are merged in, and then up-sampled to a large image of 500×500×3; the skip structure is used to improve accuracy Sex; the second deconvolution step size is 16, this network is called FCN-16s;
[0162] Figure 16 As shown, the up-sampling is completed in three times; the prediction results of the third pooling layer are further integrated, and then up-sampling is a large image of 500×500×3; the third deconvolution step is 8, which is recorded as FCN-8s.
[0163] The network structure is summarized as follows; the input can be an image color image of any size; the output is the same size as the input, and the depth is: 2 types of targets + background = 3; 2 types of targets are tongue and upper lip, and the background uses facial skin color; Use the FCN-8s full convolutional neural network to segment the tongue object; what I want to emphasize here is the first use Figure 14 Training FCN-32s full convolutional neural network as shown, and then use Figure 15 Shown to train FCN-16s fully convolutional neural network, and finally use Figure 16 The shown training FCN-8s full convolutional neural network;
[0164] After segmenting the tongue object with the FCN-8s full convolutional neural network, it is necessary to classify the tongue image through a deep convolutional neural network.
[0165] (3) Regarding the design of a deep convolutional neural network for tongue image classification;
[0166] Deep convolutional neural network for tongue image classification and image 3 The convolutional neural network shown is exactly the same, except that a Softmax classifier is connected after the eighth fully connected layer;
[0167] The Softmax classifier uses the learning results in the deep neural network as the input data of the softmax classifier; Softmax regression is a logistic regression for multi-class classification problems, a general form of logistic regression, and is suitable for mutual exclusion between categories Situation; suppose for the training set {(x (1) ,y (1) ,...,X (m) ,y (m) )}, with y (1) ∈{1,2,...,k}, for a given sample input x, output a k-dimensional vector to indicate that the probability of each classification result is p(y=i|x), assuming the function h(x )as follows:
[0168] h θ ( x ( i ) ) = p ( y ( i ) = 1 | x ( i ) , θ ) p ( y ( i ) = 1 | x ( i ) , θ ) · · · p ( y ( i ) = k | x ( i ) , θ ) = 1 X j = 1 k e θ j T x ( i ) e θ 1 T x ( i ) e θ 2 T x ( i ) · · · e θ k T x ( i ) - - - ( 11 )
[0169] θ 1 ,θ 2 ,...Θ k Are the parameters of the model, and the sum of all probabilities is 1. The cost function after adding the rule item is:
[0170] J ( θ ) = - 1 m [ X i = 1 m X j = 1 k 1 { y ( i ) = j } log e θ j T x ( i ) X l = 1 k e θ l T x ( i ) ] + λ 2 X l = 1 k X j = 0 n θ i j 2 - - - ( 12 )
[0171] The partial derivative of the cost function with respect to the lth parameter of the jth category is:
[0172] ▿ θ j J ( θ ) = - 1 m X i = 1 m [ x ( i ) ( 1 { y ( i ) = j } - p ( y ( i ) = j | x ( i ) ; θ ) ) } ] + λθ j - - - ( 13 )
[0173] Where j is the number of categories, m is the number of categories in the training set, p(y (i) =j|x (i);Θ))} is the probability that x is divided into category j, and λ is the rule item parameter, also called the weight attenuation item, the rule item parameter is mainly to prevent overfitting;
[0174] Finally, by minimizing J(θ), the softmax classification regression is realized, and the classification regression results are saved in the feature library;
[0175] In the tongue recognition and classification, such as Figure 17 As shown, the extracted input data features are compared with the tongue image feature library data obtained by learning and training, the probability of each classification result is calculated, and the result with the highest probability is then output.
[0176] (4) Regarding the construction of a cloud service platform for self-diagnosis and health based on the recognized tongue type;
[0177] The first is the working principle of the self-diagnosis health cloud service platform: figure 1 As shown, the user uses a mobile phone to take an image of his tongue, and then sends the image of his tongue to the health cloud service platform through WeChat or MMS or QQ. The health cloud service platform automatically reads the image sent from WeChat, MMS or QQ At the same time, generate a WeChat or MMS or QQ number folder, and save the original image in this folder; on the other hand, the health cloud service platform first uses a tongue segmentation method based on a full convolutional neural network to segment the tongue image Perform segmentation to obtain the segmented tongue image; then use the deep convolutional neural network for tongue image classification to classify the segmented tongue image to obtain the tongue image type; finally access the health cloud service platform database according to the tongue image type Etiology analysis table, syndrome differentiation and treatment table, and life guidance table, to obtain three groups of information reflecting the tongue image type: cause analysis, syndrome differentiation and treatment, and life guidance. The tongue image and type, etiology analysis, syndrome differentiation and treatment, and life guidance Group information automatically generates a health consultation file. The name of the health consultation file is named after the user’s WeChat account, or mobile phone number, or QQ number transmitted to the health cloud service platform. Finally, the health consultation file is named after the user’s WeChat account, or mobile phone number, Or the QQ account is fed back to the visiting user and stored in the server, or the user is notified to visit the health cloud service platform to obtain the user's self-service health test result report;
[0178] In the present invention, according to the method of tongue diagnosis in traditional Chinese medicine, the type of tongue is first divided into two categories, namely tongue body and tongue coating; then the tongue body is subdivided into 5 categories, namely tongue spirit, tongue color, tongue shape, and tongue state. And sublingual collaterals; the tongue coating is subdivided into two major categories: coating color and tongue quality, as shown in 19;
[0179] Tongue God is divided into prosperous tongue and dry tongue; tongue color is divided into pale red tongue, pale white tongue, red crimson tongue, and blue-purple tongue; Picture 20 Shown
[0180] The tongue shape is divided into old and tender, fat and thin, prickly and cracked; Figure 21 Shown
[0181] Tongue is further divided into soft, hard, skewed, trembling, chattering and shortening; such as Figure 22 Shown
[0182] The color of the moss is divided into white moss, yellow moss, gray moss and black moss; Figure 23 Shown
[0183] The tongue is divided into thick and thin, moisturizing, greasy and rotten, and peeling; Figure 24 Shown
[0184] Figure 20 ~ Figure 24 Listed in the etiology analysis of various tongue types; tongue body and tongue coating is a whole, and each has its own bias. The tongue is mainly used to detect the insufficiency of the viscera and the ups and downs of the qi, blood and body fluid, but also the nature of the evil energy; the tongue coating focuses on identifying the nature of the pathogen and the ebb and flow of the evil, but it can also detect the existence and death of the stomach; the tongue is unilaterally abnormal, indicating the condition Relatively single. Or the tongue quality is normal but the tongue coating is different, or the tongue coating is normal but the tongue quality is different; if the tongue body and tongue coating are both abnormal, or the two changes are the same, the pathogenesis is the same, and the disease is often a combination of the two; the tongue body and tongue coating changes are inconsistent , There are often two or more pathogenesis, the condition is more complicated, and the main disease of the tongue picture is also a combination of the two, and attention should be paid to the relationship of the specimens;
[0185] Dynamic analysis of tongue picture: Tongue picture has corresponding changes with the development of the disease, and dynamic analysis should also be made with the development of the disease during observation; such as changes in tongue picture in exogenous diseases, the development of internal injuries and miscellaneous diseases, and diseases Advancing and retreating, and adversity of the pathological situation, based on this, provide an important basis for early diagnosis and early treatment; the present invention records in detail the results of the user’s visit to the health cloud service platform from the clinic and the time of the visit. This information is helpful Dynamic analysis of tongue image;
[0186] For example: Exogenous disease, the tongue coating changes from thin to thick, which is caused by evil qi from the outside; the tongue coating turns from white to yellow, which turns the disease into heat; the tongue changes from pale red to red and crimson, which is flooded by evil heat, and the air is burnt; see The tongue coating is peeling off, and the tongue is red and crimson. The heat enters the blood and hurts both Qi and Yin.
[0187] The same is true for internal injuries and diseases, such as stroke patients, if the tongue is pale red and the coating is thin and white, it means that the condition is mild and the prognosis is good; if the tongue changes from pale red to red, then it turns dark red, red crimson, purple and dark, and the tongue coating turns to Yellow greasy or scorched black, or irritability in the sublingual collaterals, it indicates wind phlegm to dissipate heat, blood stasis is blocked; if the tongue changes from dark red, dark purple to pale red, and the tongue coating gradually turns, it often indicates that the condition is improving .
[0188] The dialectic results of "cold-heat deficiency and excess" are rich in connotation. Here are just two examples:
[0189] Regarding pale tongue and yellow greasy coating: pale tongue, mainly deficiency and cold, yellow and greasy tongue coating, mainly dampness and heat. The former reflects the yang deficiency due to lack of righteous qi, and the latter suggests feeling the evil of damp heat. Patients with yang deficiency (such as spleen and stomach deficiency and cold) who have relapsed to damp-heat evil;
[0190] About red and crimson tongue, white and greasy coating: the tongue is red and crimson, the main internal heat is blazing, and the coating is white and slippery, it is cold and wet and internal depression, and its disease is both cold and hot. Clinically, it can be seen in exogenous disease, which is caused by heat in Yingfen, while dampness in Qi. It can also be seen in internal injury and disease, which is the body of yin deficiency and fire, regaining cold and dampness, or stagnation of phlegm and food; It is seen in patients with damp-heat disease, both of which are positive and damp;
[0191] If the identified tongue picture belongs to the pale tongue color type, the etiology analysis, syndrome differentiation and treatment, and life guidance of this type are stored in the database as follows:
[0192] Cause analysis: pale tongue color is lighter than normal tongue color, which mostly indicates deficiency of Qi and blood or Yang deficiency. A pale tongue first indicates blood deficiency. As the hemoglobin in the body is reduced, hemoglobin is reduced and the blood is thinned. Therefore, the blood color of the tongue mucosa becomes pale, and the tongue is pale. Accompanied by symptoms of anemia such as dizziness, palpitation, dizziness, and pale complexion. Secondly, qi deficiency can also be seen in pale white. Pale tongue is the same as mental fatigue, forgetfulness, sweating when moving, and weakness in vocalization. Mostly due to qi deficiency, pale tongue also appears in yang deficiency physique. Due to insufficient yang qi, the body's water fluid runs slowly. Excess water stays in the body (it has been called water poisoning, water dampness) and cannot be excreted in time, causing edema of the tissue mucosa, thickening of the mucosa due to edema, reduced transparency, and blood color cannot be revealed, thus making the tongue appear fat and tender and pale in color Characteristics. If the waist and legs are cold, the lower body seems to be sitting in cold water. In addition, diseases that belong to the deficiency of kidney essence (such as aplastic anemia) are also one of the reasons for the pale white.
[0193] Dialectics and treatment: Patients with pale tongue, weak and cold, sore waist and knees, short urination, and frequent lower extremity edema, usually take Jinkuishenqi Pills or Fugui Bawei Pills, which can warm Yang and diuresis, invigorate the kidney and relieve turbidity. In patients with chronic anemia, in addition to symptomatic treatment for the cause of blood loss and stop bleeding as soon as possible, they usually use more methods to adjust the spleen and stomach, invigorate qi and promote blood. Bazhen Decoction, Danggui Buxue Decoction, Shiquan Dabu Decoction are commonly used. Replenishing qi and blood. At the same time, iron supplementation is also necessary. If it is a disease of bone marrow hematopoietic dysfunction (such as aplastic anemia), it is advisable to add kidney essence supplements for effective. Need to use Guilu Erxian Ointment, Ginseng Rongbuxue Pill and other prescriptions. These drugs can not only supplement nutrition, but also promote hematopoietic function and gradually improve the anemia.
[0194] Life guidance: pale tongue is a common tongue picture, mostly caused by insufficient blood and yang. Tonic Qi and blood is the most effective measure to improve tongue color. Therefore, people tend to focus on choosing appropriate medicines and health foods, and there are two prerequisites for negligence to replenish qi and blood, otherwise it will be twice the result. One is the digestion and absorption capacity; the other is whether the hematopoietic function of the body has reached the most vigorous state. Therefore, in the nursed back to nursery, the first thing to do is not to eat too much, and to have a sense of hunger to ensure that the digestive function of the stomach is in the best state. Secondly, "aerobic exercise" can promote hematopoietic function to become vigorous. Such as aerobic gymnastics, walking briskly until the body is hot, sweating, swimming, dancing, etc. The amount of exercise is controlled by the degree of rapid breathing without panting. Maintain an activity time of 2 hours a day. Among fruits, peaches are known as the "king of beauty and beauty" because peaches are rich in iron and have a better blood-tonifying effect. Regular use makes the complexion ruddy, the skin smooth, and the tongue color also improves. The iron content of cherries is twenty times that of apples, and can also be used as a blood-tonifying fruit for patients with pale tongue.
[0195] In addition, traditional Chinese medicine has the term "long-term illness and kidneys", which means that the function of the kidneys will be impaired when the function of various organs and organs is degraded to a certain degree. Kidney deficiency can also be seen in elderly and weak normal people. Patients with kidney deficiency have different syndromes such as kidney yin deficiency, kidney yang deficiency, and kidney essence deficiency. Those who have pale tongue and see the tender side of the tongue are mostly of kidney-yang deficiency. Insufficient yang and cold body, you should thicken your clothes and keep warm, and avoid entering and leaving places with large temperature differences to prevent colds. Regular exercise in a sunny place can invigorate the body and promote Yang Qi. Bathing in hot springs is also a fitness method for warming luck and promoting blood circulation. Winter is the season for all things in nature to be sealed off. Drinking ginseng tonic in winter can warm the yang and nourish qi, support righteousness, eliminate pathogens, and strengthen the body’s immunity. Choose appropriate varieties according to different physiques. Red ginseng is warm in nature, suitable for the elderly and those with yang deficiency, young people or those who have no obvious chills, white ginseng or processed bowling ginseng is a good health product. Add some green onions, ginger, leeks, garlic and the like in the diet to warm the internal organs and help digestion. Ejiao, jujube, and angelica mutton porridge are also good medicinal diets for warming essence and blood.
[0196] If the identified tongue picture belongs to the normal tongue coating type, the etiology analysis, syndrome differentiation and treatment, and life guidance for this type as it are stored in the database are as follows:
[0197] Etiological analysis: thin white coating is normal tongue coating, suggesting that the stomach has qi to grow, and internal organs have normal physiological functions. It can also appear in the initial stage of the disease when the condition is relatively shallow and the visceral function has not been damaged. The thin white fur is produced by steaming on the stomach gas and moisturizing the stomach yin on the tongue. According to histology, the body's nutrition and metabolism are normal, and the tongue mucosal papillae, especially the filamentous papilla, grow normally, which is a normal tongue coating. However, thin white fur can also indicate that wind-cold and wind-heat have first attacked the human body, and the pain evil is in the light and superficial stage of the skin. If the tongue is reddish with red spots, it indicates that the wind and cold have a tendency to dissipate heat. One possibility is that the physique is too hot to feel the wind chill, which can dissolve the heat. For example, if you are greedy for cold drinks in summer, or the indoor air conditioner temperature is too low, or you are caught in the rain, the body will be afraid of cold and sweating, headache, nasal discharge, sore limbs, fever, and thirst. Symptoms of heat. Otherwise, it may be caused by external heat. At the beginning of the disease, symptoms such as fever, fear of wind, swelling and sore throat (tonsillitis), etc. appear, which are symptoms of wind-heat invading the muscle surface, and thin white coatings are generally seen in the initial cases of the disease.
[0198] Dialectics and treatment: Thin white coating and light red tongue are normal tongue appearances, reflecting the body's qi and blood congestion, the coordination of internal organs, and the qi of the stomach. The human body is in a normal physiological state. The tongue is pale red, the coating is thin and white, accompanied by fear of wind and cold, no sweating, mild fever, and a short course of disease, indicating that the external cold is felt, and the treatment is mainly to relieve the surface. Jingfangbaidu powder, afternoon tea, ginger tea, etc. can be used to dispel cold and relieve table. Such as fever, fear of wind, headache, slight thirst, sweating, throat swelling and sore throat, mostly cold and cold, Yinqiaosan and Sangjuyin should be used as the main prescription to clear away heat and dissolve. Radix isatidis and dandelion have the effect of clearing away heat and detoxification. For patients with swollen throat (such as acute tonsillitis), the main prescription can be added to enhance the effect, or bezoar detox tablets or Liushen pills can be used. If the cold does not heal for several days, accompanied by cough, yellow phlegm, chest pain and other symptoms, it indicates that the external evil has entered the interior to dissipate heat, and it is necessary to combine the Qianjin Weijing Decoction for clearing the lungs and resolving phlegm. Add houttuynia cordata tablets, expectorant oral liquid, etc. If you have a cold and diarrhea, you often choose Pueraria lobata soup or Mushroom Xiangzhengqi powder as the main prescription, which can not only relieve the surface but also adjust the stomach and intestines. After years of clinical and pharmacological experimental research, Yinqiao powder and Huoxiang Zhengqi powder (pills) have obvious curative effects on viral colds and pathological gastroenteritis, because these drugs have self-inhibition of virus. For those who have poor resistance and are prone to catching colds, you can often take Yupingfeng San to increase the body's immune function and reduce upper respiratory diseases.
[0199] Life guidance: Under pathological conditions, thin white fur is more common in the initial stage of illness caused by wind-cold or wind-heat. The "external evil" mentioned here refers to the abnormal climate change, also known as the evil of the "six evils". According to the characteristics of the season, it is susceptible to cold evil in winter, wind evil in spring, and dry evil in autumn. There is a saying in Chinese medicine that "the six qi (evil qi) are all from cremation", that is, all kinds of pathogens that are not cleared in the body in time are called severe febrile diseases (huahuo). Therefore, people with poor body resistance, especially the elderly and children, are very susceptible to colds when the season changes. During the day, there is a big temperature difference between morning and night, so pay special attention. The human body can adapt to the weather and reduce diseases. The ancient teachings of "Evil (pathogenic factor), its qi (resistance) must be weak", "Void evil and thieves have time to avoid the wind" are meaningful for maintaining health. At the same time, you should also participate in appropriate sports activities, and insist on sunbathing or cold water bathing exercises, which will greatly help enhance the body's resistance and adapt to climate change. Sufficient intake of various essential nutrients such as vitamins, proteins, fats and carbohydrates is also the main aspect of maintaining a healthy body.
[0200] If the identified tongue picture belongs to the tongue tip petechia type, the etiology analysis, syndrome differentiation and treatment, and life guidance of this type are stored in the database as follows:
[0201] Etiological analysis: petechiae on the tongue suggests congestion in the internal organs. The reasons for its formation are roughly as follows: increased blood viscosity, increased external vascular resistance, or decreased cardiac pumping function, causing circulatory dysfunction, especially the slowing of microcirculation blood flow, tissue stagnation, insufficient oxygen exchange, and blood oxygen The partial pressure is reduced, making the blood purple and dark. Hypoxia and nutrient metabolism obstacles can also cause capillary wall degeneration and increased permeability. Red blood cells leak from the tube wall and deposit under the mucosal tissues of the tongue and in the fungal papilla, forming stasis, so ecchymosis or ecchymosis can be seen on the tongue Petechiae (large and visible purple-black fungal papillae). Seeing ecchymosis tongue is not just a local pathological change of the tongue, but a reflection of body tissue or visceral lesions on the tongue. It's just that the change in tongue picture is obvious. Therefore, lavender ecchymosis tongue is also one of the important signs of visceral tissue congestion. Such as liver cirrhosis, tumors of various organs, bruises and stasis in the body, menstrual pain, stillbirth, chronic limb bruising, etc., you can see different degrees of petechiae or petechiae on the tongue.
[0202] Syndrome differentiation and treatment: There are many reasons for congestion, and the clinical symptoms are different. Insufficiency of Yang or the invasion of cold pathogens can cause blood to stagnate, and lavender is more common in tongue. Insufficiency of yin and fire or heat evil invade, can make blood concentrate, the tongue is often dark red or purple. The internal resistance of turbid phlegm can make the blood sticky, the tongue is fat and dark purple, and the tongue coating is thick and greasy. Blue and purple tongue with ecchymosis, or fat and tender tongue with dark ecchymosis, mostly caused by imbalance of Qi and stagnant blood circulation, resulting in congestion in the corresponding internal organs. Treatment must be based on clinical symptoms, analysis of the cause of the disease and prescribing the right medicine. For example, long-term depression, anxiety, stagnation of liver qi, and poor circulation of qi and blood cause congestion in the organs and various parts of the body. The treatment uses Yueju Pills and Xuefu Zhuyu Decoction. Congestion block the meridian and cause pain and numbness of the body, you can choose Shufeng Huoxue Decoction and Danggui Shaoyao San for flavor. For Qi deficiency and blood stasis syndrome, a large number of Qi-tonifying drugs must be used to promote blood flow. Huang (Caoshi) Guizhi Wuwu Decoction and Buyang Huanwu Decoction are the preferred prescriptions. For those with blood stasis due to traumatic injury, it is necessary to strengthen the elimination of blood stasis and swelling. Sanqi injury pills and Yunnan Baiyao are commonly used orally, and externally rubbed with Chinese traditional Chinese medicine oil on the injured area can improve local circulation and promote the regression of hematoma.
[0203] Life Guidance: The main reasons for the formation of bruises and ecchymosis tongue are qi stagnation and blood stasis, qi deficiency and blood stasis, cold coagulation and blood stasis, phlegm and dampness blocking blood vessels, etc., causing local lesions of the meridians, internal organs or tissues. The pathological concept of the so-called "qi stagnation and blood stasis" in Chinese medicine is related to the mental and emotional activities of the human body and the state of autonomic nervous function. For example, under the market economy situation, enterprises are facing severe competition. Many entrepreneurs, operators and managers are under heavy mental pressure, and lofty ambitions are difficult to realize for a while, often falling into an unhealthy mental state of anxiety and depression. Among. Some people are busy with banquets and entertainment, their intestines and stomach are overwhelmed, and their rhythm of life is disordered. They cannot get the rest and relaxation they should have. Over time, stagnation of Qi and blood stasis, phlegm and blood stasis will occur. In some weak links of the body, blood stasis will gradually change. This is one of the understandings of tumor incidence in Chinese medicine. Therefore, we must pay attention to adjusting our mental state, cultivate a cheerful personality, and be good at finding pleasure on our own. At the same time, we must adjust our daily life and participate in sports appropriately, which is of great benefit to disease prevention and treatment. In terms of diet, safflower oil can be used to cook dishes. It is not only delicious, but also has a good blood circulation effect. Salt and pepper peach kernels, sweet and sour garlic, buckwheat, onions, etc. have a good blood circulation and collateral function, which can be used as an auxiliary dish and is often eaten.
Example Embodiment
[0204] Example 2
[0205] The tongue self-diagnosis health cloud service system based on the deep convolutional neural network of the present invention can be directly applied to hospitals and health centers at all levels. Like blood pressure measurement, it is done through the tongue self-diagnosis health cloud service platform before going to the doctor. A pre-diagnosis; on the one hand, this can help the patient quickly find the department corresponding to his condition; on the other hand, it can provide a reference for the doctor for further examination and diagnosis.
PUM


Description & Claims & Application Information
We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.