Bed-Exit Prediction Applying Neural Network Combining Bed Position Detection and Patient Posture Estimation.

Area of Science:

• Nursing Technology
• Geriatric Care
• Computer Vision

Background:

• Nursing work involves high-stress tasks, with potential for serious accidents from errors or omissions.
• Patient falls, especially among the elderly with reduced muscle strength, are a significant concern, often occurring during toileting.
• Existing nursing workloads are high, necessitating innovative solutions to mitigate risks and improve patient safety.

Purpose of the Study:

• To develop a sensing system aimed at preventing fall accidents in healthcare settings.
• To specifically address fall risks associated with elderly patients during toileting activities.
• To utilize camera imaging for early detection of fall-initiation postures.

Main Methods:

• Employed a camera image-based sensing system to detect the patient's end position, indicative of fall initiation.
• Combined patient skeletal position detection with bed position detection to accurately identify the sitting posture.
• Constructed a simulation environment to evaluate the system's performance using real-world captured images.

Main Results:

• Successfully detected the patient's sitting position by integrating skeletal and bed position data.
• Evaluated the estimation accuracy of the patient's end sitting position within a simulated environment.
• Demonstrated the feasibility of using camera vision for fall risk assessment in active patient and nurse scenarios.

Conclusions:

• The developed sensing system shows promise in preventing fall accidents by detecting critical pre-fall postures.
• Integrating skeletal and bed position detection offers a viable method for identifying at-risk patient positions.
• This technology can contribute to reducing the burden on nursing staff and enhancing elderly patient safety.