Facial paralysis recognition and grade evaluation method based on expression recognition technology

Abstract

The invention belongs to the field of medical diagnosis, and discloses a facial paralysis recognition and grade evaluation method based on an expression recognition technology, and the method specifically comprises the steps: 1, applying the expression recognition technology to the facial paralysis recognition and grade evaluation of a clinical patient, and building a quantitative index evaluation model; 2, researching dynamic changes of facial muscle functions before and after drug therapy, acupuncture, manipulation and TMS intervention, and perfecting and optimizing a facial paralysis recognition and grade evaluation model of a facial paralysis patient based on an expression analysis technology; from the perspective of practice, compared with a traditional method, the method has higher efficiency and precision, beneficial guidance can be provided for related problems in practice, facial paralysis can be objectively and quantitatively evaluated, facial pictures and videos of facial paralysis patients are researched on the basis of the expression analysis technology and aiming at clinical requirements, the research design is reasonable, and the method is suitable for popularization and application. The interdisciplinary advantage of medical and industrial combination is exerted, and computer and medical clinical practice can be combined for cross-professional research.

Description

Technical Field

[0001] The present invention relates to the field of medical diagnosis, and in particular to a facial paralysis recognition and grade assessment method based on expression recognition technology. Background Art

[0002] Facial paralysis is a common disease in clinical medicine. The main symptoms are that the facial nerve is damaged, which causes the facial expression muscles of the patient to be unable to perform certain normal functional movements like normal people. The specific manifestations are weakness of muscles on one side of the face, inability to close the eyes, inability to close the corners of the mouth, and a series of other conditions. It is a common and frequently occurring disease, and there are often patients with facial paralysis at all ages. Affected by work pressure and other factors such as irregular rhythm of life and work, the number of people suffering from facial paralysis is increasing year by year.

[0003] At present, there are a large number of effective means for the treatment of facial paralysis in China, and good results have been achieved, which have enabled a large number of patients with facial paralysis to recover and reduced the incidence rate. However, the existing facial paralysis identification and facial paralysis grade classification still do not meet the medical clinical practice standards in terms of accuracy, and there are still the following deficiencies in their application in medical clinics: (1) Due to the lack of medical expertise, the number of patients with facial paralysis is currently far greater than the number of doctors, which means that many patients with facial paralysis cannot be discovered early and receive timely treatment, which further aggravates their condition and increases the annual prevalence of facial paralysis.

[0004] (2) Currently, most doctors rely on their own experience to diagnose facial paralysis symptoms in patients with facial paralysis, which is greatly influenced by their own subjective factors.

[0005] (3) Currently, there is no unified standard for the diagnosis of facial paralysis symptoms and the assessment of facial paralysis grades in China. Affected by the subjective factors of medical personnel, there are inconsistencies in the evaluation of the efficacy of facial paralysis patients during the treatment and recovery process. Summary of the invention

[0006] The purpose of the present invention is to solve the problems in the above-mentioned background technology and to provide a method for facial paralysis recognition and grade assessment based on expression recognition technology.

[0007] The facial paralysis recognition and level assessment method based on expression recognition technology includes the following specific steps: Step 1: Apply facial expression recognition technology to the recognition and grade assessment of facial paralysis in clinical patients, and establish a quantitative index evaluation model; Research subjects: Patients with facial paralysis in the rehabilitation department of a hospital were selected as research subjects; Information collection: The HB system groups and collects information: divided into group A with central facial paralysis and group B with peripheral facial paralysis, and collects electromyograms, static photos and dynamic videos of patients. ① Static photos: The examinee sits facing the examiner and remains relaxed. Mark points are made on the nasal root point, nasal tip point, upper alveolar seat point, lower alveolar seat point, extraorbital point, Daycron point, zygomatic protrusion point, and mandibular tubercle point on the face, and a static facial picture is taken. ② Dynamic video: The examinee is asked to raise eyebrows, frown, close eyes, open eyes, hold nose, puff cheeks, purse lips, show teeth, and grin in turn, and the movements are recorded; Facial feature quantification program architecture: ①Analysis of static photos: pre-process the image, segment the facial area, determine the facial symmetry axis, classify the input vector calculation, and select the final parameters; ②Analysis of dynamic videos: determine the coordinates of feature points, track the displacement and speed of feature points, calculate coefficients and output the results. The above two analyses are combined to obtain the primary program; Step 2: Study the dynamic changes of facial muscle function before and after drug treatment, acupuncture, manipulation, and TMS intervention, and improve and optimize the facial paralysis recognition and grade assessment model based on expression analysis technology for patients with facial paralysis; After the two groups of patients received drug treatment, acupuncture, manipulation, and TMS intervention, they were input into the primary program for analysis, and the program was optimized and improved. The information of a new group of facial paralysis patients was input into the program to output whether they had facial paralysis and its type and stage to test the program operation results.

[0008] Beneficial effects of the present invention: From a practical perspective, the present invention has higher efficiency and accuracy than traditional methods, can provide useful guidance and instruction for relevant issues in practice, and can objectively and quantitatively evaluate facial paralysis. The present invention is based on expression analysis technology and studies facial images and videos of patients with facial paralysis in response to clinical needs. The research design is reasonable, and the cross-disciplinary advantages of combining medicine and engineering are brought into play. It can combine computers with medical clinical practice for cross-disciplinary research, and use the research results of predecessors to carry out some related research contents, making the research have good feasibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a structural diagram of the facial paralysis and non-facial paralysis analysis data model of the present invention. DETAILED DESCRIPTION

[0010] See also Figure 1 As shown, the facial paralysis recognition and level assessment method based on expression recognition technology includes the following specific steps: Step 1: Apply facial expression recognition technology to the recognition and grade assessment of facial paralysis in clinical patients, and establish a quantitative index evaluation model; Establish the data model framework for facial paralysis and non-facial paralysis analysis Input layer: The input of CNN is a 4D array with shape (batch_size, height, width, channels). The image size is 128x128 and it is a grayscale image (1 channel), so the shape of the input layer is (None, 128, 128, 1).

[0011] The first convolutional layer (Conv2D): Number of filters: 32 Convolution kernel size: 3x3 Activation function: ReLU (Rectified Linear Unit) Input shape: (128, 128, 1) (Since this is the first convolutional layer, the input shape needs to be specified) Purpose: Detect local features in images The first maximum pooling layer (MaxPooling2D): Pooling window size: 2x2 Step size: Default is 2 Function: Reduce the spatial dimension (i.e. height and width) of the feature map while maintaining important feature information; The second convolutional layer (Conv2D): Number of filters: 64 Convolution kernel size: 3x3 Activation function: ReLU Function: Further extract more complex features, allowing the network to learn richer feature representations Second maximum pooling layer (MaxPooling2D): Pooling window size: 2x2 Function: Same as the first pooling layer, further reducing the spatial size of the feature map.

[0012] The third convolutional layer (Conv2D): Number of filters: 64 Convolution kernel size: 3x3 Activation function: ReLU Function: Based on the previous convolutional layer, continue to extract higher-level features.

[0013] Flatten: Function: Flatten the multi-dimensional feature map into a one-dimensional array so that it can be input into the fully connected layer.

[0014] The first fully connected layer (Dense): Number of neurons: 64 Activation function: ReLU Function: Perform advanced processing and classification of features, which maps the flattened features to a higher-dimensional space; Output layer (Dense): Number of neurons: 2 (corresponding to two categories: patients and healthy people) Activation function: softmax Function: Generate the final classification decision and use the softmax activation function to convert the output into a probability distribution, indicating the predicted probability of each category.

[0015] Step 2: Study the dynamic changes of facial muscle function before and after drug treatment, acupuncture, manipulation, and TMS intervention, and improve and optimize the facial paralysis recognition and grade assessment model based on expression analysis technology for patients with facial paralysis; After the two groups of patients were treated with medication, acupuncture, manipulation, and TMS, they were input into the primary program for analysis and the program was optimized and improved. A new group of facial paralysis patients were input into the program to output whether they had facial paralysis and its type and stage to test the program operation results.

Claims

1. A facial paralysis recognition and grade assessment method based on expression recognition technology, characterized in that: The specific steps include: Step 1: Apply facial expression recognition technology to the recognition and grade assessment of facial paralysis in clinical patients, and establish a quantitative index evaluation model; Research subjects: Patients with facial paralysis in the rehabilitation department of a hospital were selected as research subjects; Information collection: The HB system groups and collects information: divided into group A with central facial paralysis and group B with peripheral facial paralysis, and collects electromyograms, static photos and dynamic videos of patients. ① Static photos: The examinee sits facing the examiner and remains relaxed. Mark points are made on the nasal root point, nasal tip point, upper alveolar seat point, lower alveolar seat point, extraorbital point, Daycron point, zygomatic protrusion point, and mandibular tubercle point on the face, and a static facial picture is taken. ② Dynamic video: The examinee is asked to raise eyebrows, frown, close eyes, open eyes, hold nose, puff cheeks, purse lips, show teeth, and grin in turn, and the movements are recorded; Facial feature quantification program architecture: ①Analysis of static photos: pre-process the image, segment the facial area, determine the facial symmetry axis, classify the input vector calculation, and select the final parameters; ②Analysis of dynamic videos: determine the coordinates of feature points, track the displacement and speed of feature points, calculate coefficients and output the results. The above two analyses are combined to obtain the primary program; Step 2: Study the dynamic changes of facial muscle function before and after drug treatment, acupuncture, manipulation, and TMS intervention, and improve and optimize the facial paralysis recognition and grade assessment model based on expression analysis technology for patients with facial paralysis; After the two groups of patients received drug treatment, acupuncture, manipulation, and TMS intervention, they were input into the primary program for analysis, and the program was optimized and improved. The information of a new group of facial paralysis patients was input into the program to output whether they had facial paralysis and its type and stage to test the program operation results.

Images