assessment method and system for Parkinson's disease gait dyskinesia severity, and equipment

Abstract

The invention provides an automatic assessment method and an automatic assessment system for Parkinson's disease gait dyskinesia severity, and equipment. The assessment method comprises the followingsteps of: acquiring a gait video of a Parkinson's disease patient; processing the data of the gait video, and dividing the data into a training set and a test set so as to train and test a neural network model; and analyzing a gait video of a to-be-assessed Parkinson's disease patient by using the neural network model, so as to obtain an assessment result of the gait dyskinesia severity of the to-be-assessed Parkinson's disease patient. The assessment method for automatically assessing the severity of the Parkinson's disease gait dyskinesia by analyzing the gait video of the Parkinson's disease patient through utilizing the neural network model is provided for the first time, and compared with the prior art, the assessment method has the advantages of being convenient to operate, high in assessment efficiency and the like.

Description

technical field [0001] The invention relates to the field of gait movement analysis of Parkinson's disease patients, in particular to an automatic assessment method, system and equipment for the severity of Parkinson's disease gait movement disorder. Background technique [0002] Parkinson's disease (PD) is a progressive neurodegenerative disorder with four cardinal symptoms of resting tremor, stiffness, dyskinesia, and postural instability. Among them, dyskinesia is the most typical clinical feature and one of the most easily recognizable symptoms of PD [1]. At present, the main basis for the evaluation of motor function in Parkinson's disease is the evaluation scale. The Unified Parkinson's Disease Rating Scale (UPDRS) [2] is the most complete standard evaluation scale for evaluating PD, and was updated in 2007 by the movement disorder Association (MDS) revised version, known as MDS-UPDRS [3]. The third part of the scale evaluates PD motor symptoms, and the rater is requ...

AI Technical Summary

Problems solved by technology

First of all, in the sensor-based mode, most of the sensors need to be in direct contact with people, and after wearing them, they will more or less affect the patient's movements, and the sensors often need to add additional overhead, making it difficult to popularize in daily mobility assessments

Secondly, in the vision-based mode, there are mainly the following three problems: 1) The traditional feature engineering method needs to extract and select important features through image preprocessing, which puts forward high requirements for video shooting. Shooting from the side view provides conditions for extracting human body contour features. Factors such as the appearance and clothing of the patient, the background of the shooting environment, and lighting have a great impact on the effect of image preprocessing, making the model lack of robustness; 2) Studies have shown that [9], when judging the gait movement of PD patients, in addition to the feet, other parts of the patient's body can also provide useful feature information. The gait evaluation rules in MDS-UPDRS also clearly point out that the patient's Rotation and arm swing, so all body parts of the patient should be considered in the assessment

For example, complex environmental factors in traditional RGB frames may interfere with fine-grained classification to a certain extent, and in [14] it is necessary to specifically track the human body bounding box

Secondly, the multi-branch and multi-scale information in the deep learning network is often served for the same final task. For example, the three-branch attention module proposed in [15] serves the final action recognition task. However, multi-branch , Different feature information extracted at multiple scales often lacks strong correlation constraints

Images