Noise-Corrected Patient Fall Risk State Prediction System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current patient fall prediction systems generate a high number of false alarms due to noise interference and are unable to accurately distinguish between non-risky and risky patient movements, leading to decreased vigilance among healthcare professionals and reduced effectiveness in preventing falls.
Innovation Solution
A noise correcting patient fall risk state prediction system that evaluates video frames to define discrete fall risk states by subdividing the patient's image into body parts and using fall risk state transition rules, excluding noise from processing and postponing alerts until confirmed movement is detected, thereby reducing false alarms and improving the accuracy of fall risk assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video monitoring is used to detect patient movements, then fall detection capability is improved, but false alarm rate increases due to noise interference
Solution Approach 1:
The patient's body is segmented into multiple body parts (head, torso, limbs) and the video frame is divided into predetermined areas. Motion detection is performed separately for each body part rather than treating the entire patient as a single object, enabling more precise identification of meaningful movements versus noise
Solution Approach 2:
Different fall risk state transition rules are applied to different body parts based on their local characteristics. For example, movement of the head or upper body may indicate different fall risks compared to lower body movement, allowing customized detection thresholds and rules for each body part
2Measurement precision
If motion detection sensitivity is increased to capture all patient movements, then detection capability is improved, but false alarm rate increases due to non-dangerous motions
Solution Approach 1:
The system performs preliminary classification of detected movements by evaluating them against fall risk state transition rules before generating alerts. Movements are assessed in context of the patient's current fall risk state, and only movements that represent actual transitions to higher risk states trigger alerts, filtering out non-dangerous motions
Solution Approach 2:
The system continuously updates the patient's fall risk state based on detected movements and uses this state information to modulate future detection sensitivity. The current fall risk state serves as feedback that adjusts how subsequent movements are interpreted, allowing the system to be more sensitive when risk is low and more selective when risk is already high
3Reliability
If continuous monitoring is implemented to improve patient safety, then patient safety is improved, but healthcare costs increase due to additional staffing
Solution Approach 1:
The video monitoring system operates autonomously to perform continuous patient surveillance, automatically detecting movements, evaluating fall risk states, and generating alerts without requiring constant human observation. The system serves itself by processing video data and making detection decisions, replacing the need for dedicated monitoring staff
Solution Approach 2:
The patent replaces the mechanical system of human staff monitoring with an automated video processing system using image analysis algorithms. The electronic automated detection system substitutes for human visual monitoring, providing continuous surveillance without the associated labor costs and limitations
Data Source
AI summary
A patient fall prediction system from noise corrected surveillance video by identifying patient fall risk states. A hierarchy of discrete patient fall risk states, from no risk, to intermediate risk to critical risk, describe a patient fall risk. The system transitions from state to state based on changes detected in corresponding areas between a current video frame and a background frame. A set of fall risk state transition rules govern the entry into new fall risk states. A video frame is subdivided into multiple predetermined areas, at least two contain images of the patient. The number of false alarms are reduced by accurately defining fall risk state transition rules and by reducing the opportunity for noise to impact the state transition results. Frames that contain new changes are excluded from fall risk state processing, i.e., the first video frame that might cause an erroneous elevated fall risk state is culled.