Static-Dynamic Facial Expression Analysis for Parkinson’s Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing PD diagnosis methods relying on static facial images have low accuracy due to insufficient consideration of dynamic facial expression characteristics, and traditional assessment scales are subjective and inconsistent.
Innovation Solution
An auxiliary diagnosis method and system that extracts both static and dynamic features from facial expression videos, using generative networks to synthesize healthy expressions and analyze key point movements, followed by a balanced classification network to integrate these features for improved diagnosis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional assessment scales (NMSQuest, NMSS, UPDRS) are used for PD diagnosis, then the evaluation process is simple and convenient, but the diagnosis accuracy is low due to subjective medical ability and patient state variability
Solution Approach 1:
The patent replaces the manual, subjective mechanical assessment process with an automated computer vision system that captures facial videos and uses deep learning models to objectively extract and analyze facial expression features, eliminating human subjectivity while maintaining operational simplicity
Solution Approach 2:
The patent introduces an intermediate automated analysis system that acts as a mediator between the patient's facial expressions and the diagnosis, using algorithms to process and interpret facial features objectively, thereby improving measurement precision without complicating the evaluation process
2Device complexity
If only static facial images are used for PD diagnosis, then the data processing is simple, but the diagnosis accuracy is limited due to insufficient dynamic expression characteristics
Solution Approach 1:
The patent transitions from static image analysis to dynamic video analysis, extracting temporal features and motion patterns from facial expressions over time. This captures the dynamic characteristics of facial stiffness and expression deficiency that are critical for PD diagnosis, significantly improving measurement precision
Solution Approach 2:
The patent adds the temporal dimension to facial expression analysis by processing video data across multiple frames and time points. This transforms the analysis from a single static snapshot to a multi-dimensional temporal-spatial analysis, capturing expression dynamics without excessive complexity
Data Source
AI summary
An auxiliary diagnosis method for Parkinson's disease (PD) based on static and dynamic features of facial expressions is provided. Video data of various facial expressions performed by a to-be-tested patient is acquired and pre-processed to extract a plurality of optimal facial expression images corresponding to the various facial expressions. A similarity discrimination is performed on a synthesized happy facial expression image of the to-be-tested patient in a healthy state and an extracted happy facial expression image to obtain similarity features. Distances between multiple facial key points in the various facial expression images are calculated to obtain multiple key features, which are spliced with the plurality of key features to form static features. Coordinate change degrees of multiple facial key points of eyelids and mouth are calculated to obtain dynamic features. A classification prediction result of PD is output based on the spliced features.