Systems and methods for patient fall detection
A patient monitoring system to help manage a patient that is at risk of falling is disclosed. The system includes a patient-worn wireless sensor that senses the patient's motion and wirelessly transmits information indicative of the sensed motion to a patient monitor. The patient monitor receives, stores, and processes the transmitted information to determine whether the patient has fallen or is about to fall. Upon such detection, the system can notify the patient's caretakers that the patient has fallen or is about to fall and therefore, is in need of immediate attention.
1. A method of detecting whether a monitored patient has fallen by processing signals indicative of movement by the patient, the method comprising:
receiving, with a processing device from a wireless sensor, signals responsive to a linear acceleration of said patient with respect to a roll axis, a pitch axis, and a yaw axis, said roll axis corresponding to a longitudinal axis extending along a length of said patient's body, said pitch axis corresponding to said patient's hips, and said yaw axis corresponding to a horizontal plane in which said patient is located, wherein said wireless sensor is configured to be worn by said patient and comprises an accelerometer, a gyroscope, and a first wireless transceiver, and wherein said processing device comprises a processor, a memory device, a storage device, a display, and a second wireless transceiver configured to communicate with the first wireless transceiver;
receiving, with said processing device from said wireless sensor, signals responsive to an angular velocity of said patient with respect to said roll axis, said pitch axis, and said yaw axis; and
processing said received signals responsive to said linear acceleration and said angular velocity of said patient with said processor of said processing device, including electronically:
forming a sensor data vector comprising a plurality of data elements, each of the plurality of data elements responsive to one of said linear acceleration or said angular velocity with respect to one of said roll axis, said pitch axis, and said yaw axis;
normalizing said plurality of data elements to form a normalized sensor data vector;
generating a feature vector from said normalized sensor data vector, the feature vector comprising a plurality of features indicative of a patient fall, wherein each of the plurality of features are generated based on one or more of the data elements from said normalized sensor data vector;
deriving an activation value based on an application of weight vector to said feature vector, wherein said weight vector comprises a plurality of weights, each of the plurality of weights corresponding to one of the plurality of features of said feature vector;
determining whether a patient fall has occurred based on the derived activation value; and
alerting a caregiver based on said determination, said alert indicating that said patient fall has occurred.
2. The method of claim 1 , wherein said normalizing said plurality of data elements comprises normalizing each of said plurality of data elements to have zero-mean and unit-variance.
3. The method of claim 1 , wherein said plurality of features comprise:
an acceleration magnitude;
an angular velocity magnitude;
a jerk magnitude;
a fall duration;
a pitch change; and
vertical velocities.
4. The method of claim 1 , wherein said application of said weight vector to said feature vector comprises computing an inner product of said feature vector with said weight vector.
5. The method of claim 1 , wherein said application of said weight vector to said feature vector comprises:
presenting, to a supervised learning algorithm, training data that include example inputs and known outputs; and
mapping, by said supervised learning algorithm, said example inputs to said known outputs to derive said weight vector.
6. The method of claim 5 , wherein said mapping said example inputs to said known outputs to derive said weight vector is performed by Fishers' linear discriminant.
7. The method of claim 1 , wherein said determining whether a patient fall has occurred comprises identifying a sign attribute of said derived activation value, wherein a positive sign attribute of said derived activation value indicates that said patient fall has occurred.
8. The method of claim 1 , wherein one of said plurality of features comprises a fall duration, said fall duration determined by evaluating an acceleration profile of said patient over a first time period.
9. The method of claim 8 , further comprising determining a starting point of the patient fall by comparing a first magnitude of acceleration during the acceleration profile to a first threshold.
10. The method of claim 9 , wherein the first threshold is 0.6 g.
11. The method of claim 9 , further comprising determining an impact point of the patient fall by comparing a second magnitude of acceleration during the acceleration profile to a second threshold.
12. The method of claim 11 , wherein the second threshold is 0.8 g.
13. The method of claim 1 , wherein said plurality of features comprises a pitch change, said pitch change determined by comparing a first pitch orientation of the patient with a second pitch orientation of the patient, said first pitch orientation occurring prior to said second pitch orientation.
14. A method of detecting a patient fall, the method comprising:
receiving, with a processing device, a first set of signals responsive to an acceleration of a patient from a wireless sensor, wherein said wireless sensor is configured to be worn by said patient and comprises a accelerometer, a gyroscope, and a first wireless transceiver, and wherein said processing device comprises a processor, a memory device, a storage device, and a second wireless transceiver configured to communicate with the first wireless transceiver of the wireless sensor;
receiving, with said processing device, a second set of signals responsive to an angular velocity of said patient from said wireless sensor; and
generating a sensor data vector comprising a plurality of data elements, each of the plurality of data elements responsive to one of the first set of signals or one of the second set of signals;
generating a feature vector from said sensor data vector, the feature vector comprising a plurality of features indicative of a patient fall, wherein each of the plurality of features are generated based on one or more of the data elements from said sensor data vector;
deriving an activation value based on an application of a weight vector to said feature vector, wherein said weight vector comprises a plurality of weights, each of the plurality of weights corresponding to one of the plurality of features of said feature vector;
determining whether the patient fall has occurred based on the derived activation value; and
alerting a caregiver based on said determination.
15. The method of claim 14 , wherein the first set of signals is responsive to the acceleration of said patient with respect to a roll axis, a pitch axis, and a yaw axis, said roll axis corresponding to a longitudinal axis extending along a length of said patient's body, said pitch axis corresponding to said patient's hips, and said yaw axis corresponding to a horizontal plane in which said patient is located.
16. The method of claim 14 , wherein said plurality of features comprises a fall duration, said fall duration determined by evaluating an acceleration profile of said patient over a first time period.
17. The method of claim 16 , further comprising determining a starting point of the patient fall by comparing a first magnitude of acceleration during the acceleration profile to a first threshold.
18. The method of claim 17 , further comprising determining an impact point of the patient fall by comparing a second magnitude of acceleration during the acceleration profile to a second threshold.
19. The method of claim 14 , wherein said plurality of features comprises a pitch change feature, said pitch change feature determined by comparing a first pitch orientation of the patient with a second pitch orientation of the patient, said first pitch orientation occurring prior to said second pitch orientation.
20. The method of claim 19 , wherein said first pitch orientation is determined one second prior to said second pitch orientation.