Facial paralysis grade evaluation system based on depth video data analysis

Abstract

The invention discloses a facial paralysis grade evaluation system based on depth video data analysis, which comprises a training set establishment module, an evaluation model establishment module, aninput module, a judgment and an output module. The evaluation model establishment module establishes an evaluation network, and trains the evaluation network according to the training set to obtain an evaluation model. The identification network comprises three LSTM networks arranged in parallel, each LSTM network is used for extracting motion characteristic information of a face region, Then, the motion feature information extracted from different face regions is weighted by a one-dimensional linear convolution kernel, and the weighted feature is fused by a feature vector addition method toobtain the fused feature, and then the corresponding classification result is obtained by classifying the fused feature; The extracted features can largely contain the dynamic information of facial muscle movement, which can greatly improve the classification accuracy of facial paralysis.

Description

technical field [0001] The invention relates to the technical fields of medical treatment and neural network, in particular to a facial paralysis grade evaluation system based on in-depth video data analysis. Background technique [0002] Facial paralysis is a kind of common and frequently-occurring disease that facial muscle motor function is hindered, and its main symptom is that facial expression muscle group can't carry out normal functional movement, and clinical often manifests as unilateral peripheral facial paralysis (sickness on one side, The other side is normal), with a wide range of incidence and no age limit. Facial paralysis can produce great harm to the patient's physical and mental health, both can affect the normal work and life of the patient and can seriously hinder the social activities of the patient and others, bring heavier mental burden to the patient. Therefore, it is extremely important to accurately diagnose and evaluate the disease, which is rela...

AI Technical Summary

Problems solved by technology

[0004] First, the algorithm that uses regional blocks for grade evaluation often only considers the local area of ​​​​a single organ, which is very easy to destroy the overall information of the patient's face, resulting in a lack of correlation between the blocks

[0005] Second, the traditional regional block algorithm needs to train each block area separately, and it needs to repeat the training as many times as there are block areas. This is not only time-consuming but also inefficient, and it cannot guarantee relevance

[0006] Third, the traditional regional block algorithm generally adopts a weighted method when performing feature fusion between regions, but the selection of its weighting coefficient is obtained through multiple manual experiments, and its coefficient is a fixed value that cannot be determined according to Adaptive adjustment and optimization are made for practical problems, resulting in the fusion features often not well adapted to the actual situation

[0007] Fourth, most of the traditional regional segmentation algorithms use facial static images for facial palsy grading assessment, only considering the facial asymmetry information of facial paralysis patients and ignoring the movement information of their facial muscles

[0008] In addition, most traditional facial paralysis assessment methods use traditional machine learning algorithms, which make it difficult to extract more effective feature information from massive data

Even if some methods can preserve the patient's muscle movement information, it is difficult to show the complete process of the patient's movement changes

Images