Project #3: Detecting Falls Among Elderly Patients in Nursing Homes Using Wireless Sensors Networks

Abstract:

Accidental falls among the senior population are the leading cause for seniors’ admission to hospitals.  Wireless Sensor Networks (WSNs) can be used to efficiently detect falls of senior patients in nursing homes. While some fall detection methods use only on the acceleration of the patient, others register acceleration and body position to detect falls.  In this paper, we describe a novel alert system created using WSNs capable of detecting falls based on the body position and bed occupancy. The system was created according to data collected in MoteView from seven test subjects.  Accordingly, it was designed to sense three possible conditions:  (1) patient being active; (2) patient lying in bed; or (3) if the patient has fallen down.  The experimental portion of this research was performed at a nursing facility to further validate measurements previously collected in the laboratory.  Furthermore, the system has been tested on three subjects for different types of falls and was found to detect all types of falls with high accuracy. In order to provide caregivers with constant alerts regarding patients’ conditions a graphic user interface was created in LabView. Design of this system maximizes the capabilities of Memsic’s Wireless Sensor Network Developmental Kit consisting of MICAz and MIB520 base station.  Overall, the system provides a very simple and effective solution that yields high accuracy for detecting falls.  

Approach:

• Developed an algorithm that detects falls based on the body position of the patient and bed occupancy 
• Tailored the design of the system to the technical capabilities that were available from Crossbow Wireless Sensor Network Development Kit and LabView 

Algorithm For Detecting Falls:

Two separate devices were designed and built. One to detect the patient's body position-if the patient's torso is upright or in a horizontal position and the other to determine whether the bed is occupied or not. These situations resulted in three possible outcomes

1. Patient is active

2. Patient is lying in bed

3. Patient has fallen down

Table 1: Fall Detection Depending on 4 events 

IF the position of patient torso IS	AND IF the patients bed IS	Then the situation IS
Upright	Occupied	PATIENT ACTIVE
Uptight	Unoccupied	PATIENT ACTIVE
Horizontal	Occupied	PATIENT IN BED
Horizontal	Unoccupied	PATIENT HAS FALLEN DOWN

Axis orientation of the bi-axial accelerometer which is placed on the patient

Note: The sensor records the acceleration due to gravity

Hardware Implementation:

•    MIB520 Base Station

• Set of MICAz Motes equipped with MTS400CC sensor board

• MoteView Software

Data Collection:

The data was collected from seven male subjects performing various activities. The activities in question were walking, walking with a cane, jogging, tying shoes, standing straight, lying on the back, lying on the left side, lying on the right side, sitting, eating on a table.

The analysis of the data led to the establishment of the threshold acceleration (Accy):

• If Accy > 7m/S² then the patient is in upright position (patient active)
• If Accy < 4m/s² then the patient is in a horizontal position (patient lying)

Experimental Results:

Data collected from one test subject was graphed for various activities. The resulting graph was in agreement with the established threshold for the acceleration.

The established threshold was further tested for three additional subjects performing different activities and the results were satisfying.

Bed Mechanism

A black box was built for the bed mechanism. Essentially, a switch placed on the patient’s bed is used to detect the occupancy of the bed. Once the patient lies on the bed, the switch will be depressed; the switch is connected in a DC circuit with 4.5V battery and 8 LED lights. Led lights are placed in a box that is constructed from opaque material to prevent light coming from outside. The only light allow inside the box is the light from the LEDs. MICAz node is placed in the box and it is sending light intensity data to MIB520 base station. If the light intensity is under 1Lux, that is a signal LED’s are not emitting light and that bed is not occupied. If light intensity is above 10Lux that is a signal that LED’s are emitting light and that bed is occupied.

  

User's Interface-LabView Graphical

There are currently two graphical user interfaces created. In the first interface LABVIEW receives data from Memsic WSN and MICAz node that is placed on patient chest, and then it compares values from acceleration in y-axis with our set thresholds. Depending on those values, two alerts are possible: patient is lying down (y_{acc <}  4m/s² ) or patient is active (y_acc>  7m/s² ). Furthermore these values have been coupled in two consecutive time intervals ( y_acc (T), y_acc (T+1)) in order to get more accurate information. Second graphical user interface has been created for data received from node that detects bed occupancy. If flux is under 1lux that there will be an alert that the bed is unoccupied. If Light flux is above 10 Lux second alert will prompt that bed is occupied. 

Flow Chart

Conclusion

Accidental falls represent serious issue among the senior population. Timely detection of these events in nursing homes can significantly improve treatment of injuries and reduces costs to healthcare system. The aforementioned paper is a clear demonstration of a simple and effective Wireless Sensor Network (WSN) system which detects senior patients falling in nursing homes. Behavior of senior patients was observed in nursing home and used to create list of possible activities that senior patients engage in daily activities.  Then the system was created based on data from seven test subjects, and then system capabilities were verified on three test subjects. System has detected every fall, and didn't have false alarms.