Automated Patient Orientation Detection via Motion Hotspot Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current patient monitoring methods, such as wrist-worn accelerometers, are limited in detecting changes in motoric behavior and vital signs, particularly for delirium, due to their disturbance to patients and inability to capture movements from other body parts, and existing video analysis methods struggle with dynamic hospital environments like scene variations and occlusions, leading to delayed detection of critical changes in motoric behavior.
Innovation Solution
A device and method for automated detection of a person's orientation and location using image data, which identifies motion hotspots and body parts through motion detection and intensity analysis, allowing for continuous and unobtrusive monitoring of patient movements and vital signs, even in challenging environments like hospitals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If wrist-worn accelerometer techniques are used to monitor patient movements, then activity levels can be measured, but the sensor disturbs the patient and cannot capture movements of other body parts
Solution Approach 1:
The patent replaces mechanical contact sensors (accelerometers worn on the wrist) with an optical detection system using video cameras and image processing. This substitution eliminates the need for physical contact with the patient, thereby removing the disturbance caused by wearable sensors while maintaining movement detection capability through non-contact visual monitoring.
Solution Approach 2:
The video monitoring system serves multiple functions: it detects movements of any body part (not limited to the wrist), monitors vital signs, tracks orientation and location, and provides comprehensive patient behavior analysis. This multi-functional approach overcomes the limitation of wrist accelerometers that can only measure activity at a single location.
2Reliability
If medical staff observe patient motoric behavior during visits, then critical situations can be detected, but detection is delayed and relies on periodic inspection
Solution Approach 1:
The system implements continuous automated video monitoring that operates without interruption between medical staff visits. The image processing algorithm continuously analyzes patient movements, orientation changes, and behavior patterns in real-time, ensuring that critical situations are detected immediately as they occur rather than being missed during periods between periodic observations.
Solution Approach 2:
The monitoring system operates autonomously without requiring active intervention by medical staff. The automated image processing and movement analysis algorithms independently detect and classify patient behaviors, freeing medical staff from manual observation tasks while ensuring continuous monitoring coverage.
3Measurement precision
If typical body segmentation methods are used in dynamic hospital environments, then body parts can be identified, but scene variations and occlusions make segmentation difficult
Solution Approach 1:
The system employs dynamic adaptation to changing hospital environments. The image processing algorithm continuously adjusts to scene variations such as changing lighting conditions, moving objects (TV screens, medical equipment), and varying patient positions in bed. This dynamic capability allows accurate body part segmentation despite the non-stationary nature of hospital settings.
Solution Approach 2:
The patent introduces motion detection as an intermediary step between raw video input and body segmentation. By first identifying moving regions and tracking their temporal consistency, the system creates a robust intermediate representation that helps distinguish patient body parts from background elements and occluding objects, improving segmentation accuracy in challenging environments.
Data Source
Figure 1
Figure 2~7
Figure 3
AI summary
The present invention relates to a device, system and method for automated detection of orientation and/or location of a person. To increase the robustness and accuracy, the proposed device comprises an image data interface (20) for obtaining image data of a person (110), said image data comprising a sequence of image frames over time, a motion detector (21) for detecting motion within said image data, a motion intensity detector (22) for identifying motion hotspots representing image areas showing frequently occurring motion, and a person detector (23) for detecting the orientation and/or location of at least part of the person (110) based on the identified motion hotspots.