Stroke Detection via Video-Based Joint Weakness Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing stroke in emergency settings are often inaccurate and delayed, leading to missed or incorrect diagnoses, which can result in fatal consequences and increased healthcare costs due to the narrow treatment window for clot dissolving agents like tPA.
Innovation Solution
A computer-implemented method using machine learning models to detect bodily joint weaknesses by analyzing video recordings of patients, extracting spatial coordinates, and applying machine learning processes like support vector machines to identify hemiparesis and other stroke symptoms, enabling early and accurate stroke detection without requiring expert neurologists.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subjective evaluation by stroke specialists is used to diagnose stroke, then diagnostic accuracy may be improved, but diagnosis time is delayed and expert availability is limited
Solution Approach 1:
The patent creates a digital copy of the neurologist's diagnostic expertise through machine learning models trained on video data of patients performing motor tasks. The system captures the neurologist's acumen by learning from labeled video examples, then applies this learned knowledge to automatically detect hemiparesis and diagnose stroke in new patients without requiring the actual neurologist's presence, thereby eliminating time delays while maintaining diagnostic accuracy.
Solution Approach 2:
The patent replaces the mechanical system of human neurologist evaluation with an automated computer vision system. Instead of relying on subjective human assessment, the system uses machine learning algorithms to objectively analyze video data of motor movements, detect abnormalities, and diagnose stroke. This substitution eliminates the time loss associated with scheduling and performing manual evaluations while preserving diagnostic precision.
2Measurement precision
If expert neurologists are required for stroke diagnosis, then diagnostic accuracy is improved, but healthcare accessibility deteriorates in underserved areas
Solution Approach 1:
The patent creates a replicable digital model of neurologist expertise that can be deployed in any location with video recording capability. By capturing and encoding the diagnostic knowledge in a machine learning model, the system makes expert-level stroke diagnosis accessible in underserved areas without requiring actual neurologists, thereby improving healthcare accessibility while maintaining diagnostic accuracy through the copied expertise.
Solution Approach 2:
The patent creates a universal diagnostic system that can be deployed across diverse settings - from well-equipped hospitals to resource-limited communities. The machine learning model serves multiple functions: it diagnoses stroke, detects hemiparesis, and provides clinical decision support in various healthcare environments. This universal application eliminates the disparity in access to expert diagnosis between served and underserved areas.
3Device complexity
If crude stroke deficit scales are used for outcome prediction, then simplicity is maintained, but prediction accuracy deteriorates
Solution Approach 1:
The patent replaces crude manual stroke deficit scales with an automated computer vision-based assessment system. The system captures detailed video data of patients performing motor tasks and uses machine learning to extract multiple features including movement quality, speed, coordination, and range of motion. This substitution provides comprehensive, objective measurements that significantly improve prediction accuracy for stroke outcomes while maintaining ease of use through automated analysis.
Solution Approach 2:
The patent transforms the assessment from crude single-parameter scales to a multi-parameter evaluation system. By analyzing multiple video features (spatial coordinates, movement trajectories, temporal patterns, coordination metrics), the system captures nuanced information about motor function. This parameter enrichment enables accurate prediction of stroke outcomes while the automated processing maintains simplicity in execution.
Data Source
AI summary
A method, a system, and a computer program product for detecting and/or determining presence and/or absence of a weakness in one or more bodily joints of a subject. One or more video recordings of one or more subjects are received. One or more features from each of the video recordings are extracted. The features correspond to one or more bodily joints of the subjects. One or more models are trained based on the extracted features to identify a weakness in the bodily joints. The trained model is applied to a first video recording of a first subject, and, using the applied model, a determination is made whether a weakness is present in one or more bodily joints of the first subject.