Medical device for fall detection

Abstract

A medical device is configured to generate an accelerometer signal and detect a patient fall from the accelerometer signal. The device generates a body posture signal and a body acceleration signal from the accelerometer signal and detects a patient fall in response to determining that the body posture signal and the body acceleration signal satisfy fall detection criteria. The medical device is configured to receive a fact signal from another device that is not the medical device. The fact signal can indicate that the detected patient fall is a false detected patient fall, and in response to receiving the fact signal, the medical device adjusts at least one fall detection control parameter.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to a medical device including an accelerometer for detecting that a patient has fallen. BACKGROUND

[0002] A variety of medical devices for monitoring a patient and / or delivering therapy are available or have been proposed. Such devices can be implantable, wearable, or external monitoring devices, and can include one or more sensors for sensing signals related to a condition of the patient or be coupled thereto. Some medical devices can detect an adverse condition from the sensor signals and deliver therapy or generate an alert to alert the patient or a caregiver. Accelerometers can be included in some medical devices, such as pacemakers, for determining physical activity of the patient, for example, in order to provide cardiac pacing at a pacing rate that is a function of the activity level of the patient. Accelerometers can also be used to determine the posture of the patient.

[0003] Some patients, such as elderly or infirm patients or patients experiencing a neurological or other medical condition, can be prone to falling. A fall can result in injury. Complications can arise and become more serious to the patient if not treated as a result of the fall-related injury. In some cases, a fall can not result in injury, but can be an indication of a worsening health condition. The cause of the fall can be related to an undiagnosed medical condition or a worsening medical condition. SUMMARY

[0004] The technology of the present disclosure generally relates to a medical device system including an accelerometer that generates signals related to a patient body posture and patient body acceleration. In some examples, the medical device can be an implantable medical device (IMD) or a wearable device and be configured to detect a fall of the patient based on detecting a change in patient body posture and detecting an accelerometer signal spike corresponding to an impact of a fall. The medical device can be programmed with various thresholds for detecting a change in body posture and an accelerometer signal spike corresponding to acceleration and / or impact during a fall. A fall can be detected when both a change in body posture and an accelerometer signal spike corresponding to an impact during a fall occur within a predetermined fall time window of each other. As disclosed herein, the medical device can adjust one or more fall detection control parameters to avoid false fall detections and to avoid undetected falls. The medical device can determine a measure of fall severity that can be used to prioritize storage of data capturing a fall event relative to storage of other detected fall data.

[0005] In one example, a medical device includes an accelerometer configured to produce an accelerometer signal; and control circuitry configured to generate a body posture signal and a body acceleration signal from the accelerometer signal, determine that the body posture signal and the body acceleration signal satisfy a fall detection criterion, and detect a patient fall in response to the body posture signal and the body acceleration signal satisfying the fall detection criterion. The medical device can also include telemetry circuitry configured to receive a fact signal from another device, the fact signal can indicate that the detected patient fall is a falsely detected patient fall. In response to receiving the fact signal, the control circuitry is further configured to adjust at least one fall detection control parameter used in detecting patient falls. The adjusted fall detection control parameter can be used by the control circuitry to control at least one of: generating the body posture signal from the accelerometer signal; generating the body acceleration signal from the accelerometer signal; or determining that the fall detection criterion is satisfied.

[0006] In another example, a method of detecting a patient fall by a medical device includes producing an accelerometer signal and generating a body posture signal and a body acceleration signal from the accelerometer signal. The method also includes determining that the body posture signal and the body acceleration signal satisfy a fall detection criterion, and detecting a patient fall in response to the body posture signal and the body acceleration signal satisfying the fall detection criterion. The method can also include receiving a fact signal from another device, the fact signal can indicate that the detected patient fall is a falsely detected patient fall, and adjusting at least one fall detection control parameter in response to receiving the fact signal. The fall detection control parameter can be used by the medical device to control generating the body posture signal from the accelerometer signal, generating the body acceleration signal from the accelerometer signal, or determining that the fall detection criterion is satisfied.

[0007] In yet another example, a non-transitory computer-readable medium stores instructions that, when executed by a control circuit of a medical device, cause the medical device to generate an accelerometer signal, generate a body posture signal and a body acceleration signal from the accelerometer signal, and determine that the body posture signal and the body acceleration signal satisfy a fall detection criterion. The instructions also cause the medical device to detect a patient fall in response to the body posture signal and the body acceleration signal satisfying the fall detection criterion. The instructions can also cause the device to receive a fact signal from another device that can indicate that the detected patient fall is a false detected patient fall. The instructions can cause the device to adjust at least one fall detection control parameter in response to receiving the fact signal. The adjusted fall detection control parameter can be used by the medical device to control at least one of: generating the body posture signal from the accelerometer signal; generating the body acceleration signal from the accelerometer signal; or determining that the fall detection criterion is satisfied.

[0008] The details of one or more aspects of the disclosure are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of the disclosure will be apparent from the description and drawings, and from the claims. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 is a schematic diagram of an IMD implanted in a patient and including an accelerometer for detecting whether the patient has fallen.

[0010] Figure 2 is a conceptual diagram of a three-dimensional output vector of a three-axis accelerometer relative to gravity.

[0011] Figure 3 is a block diagram of circuitry that can be encapsulated within a housing of an IMD of Figure 1 to provide fall detection, according to one example.

[0012] Figure 4 is a flowchart of a method performed by a medical device to detect a patient fall, according to one example.

[0013] Figure 5 is a flowchart of a method performed by a medical device to establish a reference vector corresponding to an upright body posture, according to some examples.

[0014] Figure 6 is a flowchart of a method performed by a system of Figure 1 to establish a reference vector corresponding to an upright posture, according to another example.

[0015] Figure 7 is a flowchart of a method for establishing a reference vector corresponding to an upright posture by a system of Figure 1a flowchart of a method of a medical device system to establish a reference vector.

[0016] Figure 8 a flowchart of a method performed by a medical device to determine a recovery time after a fall detection.

[0017] Figure 9 a flowchart of a method performed by a medical device to set or adjust a fall detection control parameter according to some examples.

[0018] Figure 10 a flowchart of a method that can be performed by a medical device to detect a change in body posture according to some examples. DETAILED DESCRIPTION

[0019] An implantable or wearable medical device is configured to detect a patient fall based on a change in body posture and an accelerometer signal spike indicative of a fall impact. According to the techniques disclosed herein, the medical device can automatically adjust the criteria for fall detection in response to a signal indicative of the fact that a detected fall was a false fall detection and / or an actual occurring fall was not detected by the medical device. The medical device can also determine a fall recovery time, e.g., based on a change in body posture after a fall detection, for selecting an appropriate fall detection response, which can include prioritizing data and information related to a detected fall stored in a memory of the medical device. The techniques disclosed herein can be implemented in various medical devices, which can include a neurostimulator, a cardiac pacemaker or defibrillator, a drug pump, a cardiac signal monitor, or other medical monitoring device that can or can not include a therapy delivery capability.

[0020] Figure 1 is a schematic diagram of an IMD 10 implanted in a patient 8 and including an accelerometer 12 for detecting whether the patient has fallen. While the examples described herein relate to an IMD, it should be understood that the disclosed techniques can be implemented in a wearable medical device, which can be included in or on a watch, bracelet, bar, belt, or other wearable substrate. A medical device system performing the techniques disclosed herein includes at least an accelerometer that produces a signal related to a patient body posture relative to gravity and the patient body posture for detecting a patient fall. The medical device system can include additional sensors for monitoring other conditions or physiological signals of the patient, which can be used in some examples to corroborate a fall detection based on the accelerometer signal.

[0021] As used herein, the term "has fallen" is distinguished from "is falling" in that if a patient "has fallen," the patient has collided with a surface, such as the ground or floor. When a patient "is falling," the patient is still in the process of falling and has not yet collided with a surface. In other words, a patient "is falling" is still moving through the air and has not yet hit the ground or another surface; a patient "has fallen" has collided with a surface and is likely in a non-upright position at least initially after the collision, e.g., until recovery from the fall.

[0022] The term "body posture" as used herein refers to the position of a patient's body relative to the force of gravity of the Earth. As used herein, "upright posture" generally refers to a body position in which at least the upper body of the patient is aligned relative to gravity. Examples of upright postures include, but are not limited to, standing, walking, running, and sitting. "Non-upright posture" generally refers to a body position in which at least the upper body of the patient is not aligned relative to gravity. Examples of non-upright postures include, but are not limited to, prone, supine, reclining, and lying on one's side positions. While patient falls can often occur from an upright or vertical posture to a non-upright or horizontal posture, the techniques disclosed herein include methods for detecting falls from non-upright or horizontal postures, e.g., when a patient rolls off a bed from a lying posture and lands on the floor in a non-upright posture.

[0023] IMD 10 includes a housing 25 that defines a hermetically sealed interior cavity in which internal components of IMD 10 reside. Housing 25 encases the electrical circuitry that is part of the functionality of IMD 10 herein, including, for example, signal sensing circuitry, therapy delivery circuitry, control circuitry, memory, telemetry circuitry, accelerometer 12, and a power source, as described in greater detail in connection with FIGS. 2-4 below. Figure 3 Housing 25 can be formed from a conductive material including titanium or a titanium alloy, stainless steel, MP35N (a non-magnetic nickel-cobalt-chromium-molybdenum alloy), platinum alloy, or other biocompatible metal or metal alloy. In other examples, housing 25 is formed from a non-conductive material including ceramic, glass, sapphire, silicone, polyurethane, epoxy, acetyl copolymer plastic, polyether ether ketone (PEEK), liquid crystal polymer, or other biocompatible polymer.

[0024] In some examples, IMD 10 can be a leadless device that does not require medical leads to extend away from housing 25. In other examples, a connector assembly 27, sometimes referred to as a "header," can be coupled to housing 25. Connector assembly 27 can have one or more connector holes that are used to receive medical electrical leads and a desired number of electrical feedthroughs for electrically coupling medical electrical lead connectors to electrical circuitry encapsulated within housing 25. For example, when IMD 10 embodies a pacemaker, implantable cardioverter-defibrillator, or neurostimulator, IMD 10 can be coupled to one or more leads that carry electrodes that are deployed at an anatomical location that is remote from an implant site of IMD 10. While Figure 1 The leads are not shown, but it should be understood that the fall detection techniques and hardware, firmware, and / or software for performing the fall detection techniques disclosed herein can be implemented in various IMDs that operate in conjunction with one or more medical leads as well as leadless IMDs that do not require leads to be coupled to the IMD.

[0025] IMD 10 includes an accelerometer 12, which can be encapsulated within housing 25 (or in some cases mounted along housing 25 or in connector block 27) or carried by leads extending from IMD 10. Accelerometer 12 can be included in an accelerometer circuit 90, as described below in connection with Figure 3 The accelerometer circuit 90, which can include one or more processing units that generate one or more accelerometer signals indicative of acceleration. In some examples, the processing units of the accelerometer circuit can also be capable of processing or manipulating the one or more accelerometer signals and transmitting the processed (or unprocessed) signals to another component of IMD 10. Thus, accelerometer 12 can be included in an accelerometer circuit having components for generating accelerometer signals and / or components for processing accelerometer signals. Further, while certain functionality is described herein as being performed by the accelerometer circuit 90 shown in FIG. 1 or specifically by accelerometer 12 for conceptual purposes, it should be understood that such functionality can additionally or alternatively be performed by another component of IMD 10, such as control circuitry included in IMD 10 or processor 44 of external device 40 described below. Figure 3 Figure 3 as described below in connection with FIG. 2, or processor 44 of external device 40 described below.

[0026] ​For example, the accelerometer 12 can include a three-axis (three-dimensional) accelerometer that generates a signal through a respective accelerometer element along each of three different accelerometer axes (labeled x, y, and z). The x, y, and z signal components generated by each respective accelerometer element define a resultant output vector signal that is related to the body posture of the patient relative to gravity. Assuming that the device orientation within the patient's body does not change significantly over the same time period, a change in the direction of the resultant output vector over a specified time interval, e.g., over a few seconds or less, is indicative of a change in the patient's posture. The orientation of the three-axis accelerometer 12 relative to gravity 14 can vary between patients when the patient 8 is in an upright posture, and can change over time as the implant orientation of the IMD 10 relative to gravity 14 can change, e.g., as device migration or rotation occurs. However, the relative change in the direction of the resultant output vector of the three-dimensional accelerometer that occurs as the patient's body posture or orientation changes relative to gravity can be identified for use in detecting a patient fall.

[0027] In some examples, the three-axis accelerometer 12 includes three single-axis accelerometer elements arranged orthogonally, each generating a respective signal component corresponding to acceleration in the x-direction 20, the y-direction 22, or the z-direction 24. In other examples, the three single-axis accelerometer elements can be arranged in a non-orthogonal configuration, with each accelerometer element positioned along a unique plane relative to the other two accelerometers.

[0028] In the illustrated example, the IMD 10 is a subcutaneous leadless device, however in other examples one or more leads can extend from the IMD 10 for positioning electrodes or other sensors, including the accelerometer 12 in some examples, at desired anatomical locations remote from the IMD housing 25. The IMD 10 is shown implanted along a chest location, but can also be implanted in other subcutaneous or submuscular locations or sub-sternal implantation, e.g., implanted within the patient's thoracic cavity or intracardiac within a ventricle, or within another organ.

[0029] IMD 10 is configured to communicate with external device 40 via communication link 42. External device 40 can be referred to as a "programmer" for use in a hospital, clinic, or physician's office for retrieving data from IMD 10 and for programming operational parameters and algorithms in IMD 10 to control the functioning of the IMD, including fall detection control parameters as well as other IMD sensing and / or therapy delivery functions. External device 40 can alternatively be embodied as a home monitor for retrieving data from IMD 10 for transmission to a central database or computer to enable a clinician to remotely monitor patient 8. External device 40 can include a processor 44, a memory 45, a display unit 46, a user interface 47, and a telemetry unit 48. Processor 44 controls the operation of the external device and processes data and signals received from IMD 10. Display unit 46, which can include a graphical user interface, displays data and other information to a user for reviewing IMD operation and programmed parameters, as well as can display accelerometer signals retrieved from IMD 14 related to fall detection, to the user.

[0030] User interface 47 can include a mouse, touch screen, keypad, etc. to enable a user to interact with external device 40 to initiate a telemetry session with IMD 10 for retrieving data from and / or transmitting data to IMD 10, including programmable parameters for controlling fall detection as described herein. Telemetry unit 48 includes a transceiver and antenna configured for bidirectional communication with telemetry circuitry included in IMD 10 and is configured to operate in conjunction with processor 44 for transmitting and receiving data related to IMD functionality, which can include data related to fall detection, via communication link 60.

[0031] A wireless radio frequency (RF) link such as Wi-Fi or Medical Implant Communication Service (MICS) or other RF or communication frequency bandwidth or communication protocol can be used to establish communication link 42 between IMD 10 and external device 40. Data stored or acquired by IMD 10, including accelerometer signals or associated data derived therefrom, results of device diagnostics, and history of detected falls and sensing and / or therapy delivery functions of IMD 10, can be retrieved by external device 40 upon interrogation commands. It is recognized that in some examples, IMD 10 can communicate with external device via a relay device, which can be another implantable medical device that receives signals from IMD 10 and transmits the signals to external device 40. For example, when IMD 10 is implanted in the abdominal or thoracic cavity, which can be external or internal to the heart or another organ, IMD 10 can communicate with another implantable medical device via RF communication, tissue conductance communication, or other implemented communication protocol for relaying data to external device 40.

[0032] IMD 10 is also shown in communication with a personal communication device 50, which can be a smart phone, tablet, or other device used by the patient and configured to communicate via a communication link 56, such as a Bluetooth® link, with IMD 10. The personal communication device 50 can be configured to communicate via a Bluetooth® (or other) communication link 54 with another device or apparatus 52 while the patient is performing their daily activities. For example, the apparatus 52 can be a bathroom scale, a car audio system, a computer mouse or keyboard, a fitness tracker, etc. As described herein, IMD 10 can receive wireless signals from the patient's external personal device 50 via link 56 that are indicative of patient activity that can involve the apparatus 52 in communication with the personal device 50. IMD 10 can use the indicated patient activity, such as standing on a bathroom scale, driving or sitting in a car, sitting in front of a computer, or walking or jogging, to facilitate determining or confirming a reference upright body posture. The personal communication device 50 can be configured to communicate via a Bluetooth® (or other) communication link 54 with another device or apparatus 52 while the patient is performing their daily activities. For example, the apparatus 52 can be a bathroom scale, a car audio system, a computer mouse or keyboard, a fitness tracker, etc. As described herein, IMD 10 can receive wireless signals from the patient's external personal device 50 via link 56 that are indicative of patient activity that can involve the apparatus 52 in communication with the personal device 50. IMD 10 can use the indicated patient activity, such as standing on a bathroom scale, driving or sitting in a car, sitting in front of a computer, or walking or jogging, to facilitate determining or confirming a reference upright body posture.

[0033] Figure 2 is a conceptual diagram of the resultant output vectors 30 and 34 of a tri-axial accelerometer relative to gravity 14. To detect a fall, acceleration signal spikes are detected from a high pass or band pass filtered accelerometer signal related to patient body motion. In addition, changes in patient body posture are detected from a low pass filtered accelerometer signal related to changes in patient body position relative to gravity 14. The acceleration signal spikes detected from the band pass filtered accelerometer signal and the changes in patient body posture detected from the low pass filtered accelerometer signal can need to fall within a time window of each other in order to satisfy a fall detection criterion. When a fall occurs, the change in body posture and the acceleration signal spikes are coincident.

[0034] In one example, Figure 2 The output vector 30 shown in FIG. 3 can represent a resultant body acceleration signal vector at a sample time of a body acceleration signal acquired by band pass filtering the output signals of each of the x-axis, y-axis, and z-axis accelerometer elements. The magnitude of the resultant output vector 30 can be calculated as a mathematical combination of the x, y, and z components of the accelerometer signal. In some examples, the magnitude of the output vector 30 is calculated as the Euclidean norm, i.e., the square root of the sum of the squares of the three orthogonal x, y, and z component vectors:

[0035] ∥v∥ = (x 2 +y 2 +z 2 ) 1 / 2

[0036] Generally, if the tri-axial accelerometer is not accelerating / decelerating, the magnitude of the output vector is 1.0 g (9.8 m / s 2 ​). The magnitude of the output vector 30 is expected to exceed the threshold magnitude when the patient collides with the ground or other surface as a result of a fall. In other examples, the magnitude of the output acceleration vector at a given sample time can be determined by summing the magnitudes of the three acceleration signal components at the sample time.

[0037] As described herein, the magnitude, slew rate, and / or frequency of the body acceleration signal can be determined to detect acceleration signal spikes in the bandpass (or high pass) filtered accelerometer signal. High magnitude and high frequency spikes can be associated with a fall, particularly when reaching maximum acceleration during a fall and when hitting the floor or ground upon impact. The magnitude of the body acceleration signal can be determined at regular sampling intervals or in response to a triggering event (e.g., a detected change in body posture) to determine whether the magnitude satisfies a spike detection threshold magnitude. In some examples, in addition to or instead of determining whether the magnitude of the accelerometer signal reaches a threshold magnitude, an acceleration signal spike corresponding to a patient fall can be detected by detecting that a positive slew rate of the accelerometer signal satisfies a positive slew rate threshold. In some examples, the acceleration signal spike detection can require detecting a positive slew rate followed by a negative slew rate, both of which satisfy respective positive and negative slew rate thresholds, with both slew rates occurring within a specified time interval of each other such that the slew rates correspond to a single signal spike. In this manner, high frequency body acceleration signals corresponding to a fall can be detected in the time domain.

[0038] To capture acceleration signal spikes resulting from a fall, the three-dimensional (or two-dimensional or single-axis) accelerometer signal can be bandpass filtered to remove DC components and high frequency noise. As an example, the accelerometer signal can be filtered through a 1 to 30 Hz bandpass filter or a 1 to 10 Hz bandpass filter to obtain an output signal that includes accelerations resulting from body motion during a fall and upon impact. The filter can be an adjustable filter with a set of cutoff frequencies that are customized for individual patients to optimize fall detection performance, e.g., to reduce false fall detection rates and / or to reduce the likelihood of failing to detect a fall. For example, the IMD 10 can monitor the accelerometer signal over a period of time (e.g., over a day, a week, or longer) to determine a frequency range of the accelerometer signal during the patient’s normal daily physical activity (e.g., associated with activities of daily living (ADLs)). The control circuit or processing circuit of the IMD 10 can use the frequency range of the patient’s ADLs to determine and set the high pass and / or low pass cutoff frequencies of the bandpass filter used to generate the body acceleration signal from the signal produced by the accelerometer 12. Acceleration signal spikes can be detected from the body acceleration signal to detect patient falls.

[0039] The output signals of the three-dimensional accelerometer can also be used to determine changes in body posture or position relative to gravity. In this case, low-pass filtered signals from each accelerometer axis are taken to obtain body posture signals. As one example, the accelerometer signals from each axis can be filtered through a 1 Hz low-pass filter, where higher frequency acceleration signals due to body motion are removed.

[0040] In Figure 2 In the illustrated example, two output vectors 30 and 34 can be obtained from the low-pass filtered signals (rather than from band-pass filtered body acceleration signals as described in the previous example to obtain vector 30). In this case, the two body posture vectors 30 and 34 are determined at two different points in time, for example ranging from 2 to 12 seconds apart, for detecting changes in body posture relative to gravity 14. The angle 36 between the output vector 30 and the output vector 34 represents a change in direction from the body posture vector 30 to the body posture vector 34 from one point in time corresponding to vector 30 to another point in time corresponding to vector 34. The angle 36 can be determined by taking the inverse cosine of the product of the dot product of the two output vectors 30 and 34 divided by their magnitudes. In some examples, to simplify the calculation, the dot product of the two vectors 30 and 34 can be determined and the resulting scalar can be used as a measure of the angle 36 between the two output vectors. The angle 36, the scalar result of the dot product, or another measure of the change in direction between the two body posture vectors obtained from the low-pass filtered accelerometer signals is related to changes in the patient's body posture relative to gravity 14. The measure of the change in direction between the two body posture vectors obtained at two different points in time can be compared to a change threshold to determine whether the measure meets the fall detection criteria.

[0041] Figure 3 is a block diagram 100 of circuitry that can be enclosed within the housing 25 of an IMD 10 to provide the functionality disclosed herein to perform fall detection according to one example. The IMD 10 can include control circuitry 80, memory 82, therapy delivery circuitry 84, sensing circuitry 86, telemetry circuitry 90, accelerometer circuitry 90, and a power source 98. In some examples, the IMD 10 is capable of delivering therapy, which can be electrical stimulation therapy such as cardiac pacing therapy, cardioversion / defibrillation shocks, neural stimulation, or muscle stimulation. In this case, the therapy delivery circuitry 84 can be coupled to electrodes, which can be carried on the housing 25 and / or by medical electrical leads extending away from the housing 25. In other examples, the therapy delivery circuitry 84 can be a drug pump for delivering a medication or biological agent, such as an insulin pump. In this case, the therapy delivery circuitry can include a reservoir and a port or catheter coupled thereto for holding and delivering the fluid agent.

[0042] IMD 10 can include sensing circuitry for sensing mechanical, electrical, optical, or chemical signals related to a physiological condition of a patient. For example, electrodes can be coupled to the sensing circuitry to sense cardiac electrical signals or electrical nerve or muscle signals. IMD 10 can include various sensors in the medical device for monitoring a patient or detecting a physiological or pathological condition, such as a pressure sensor, an acoustic sensor, a temperature sensor, an impedance sensor, an oxygen sensor, a blood perfusion sensor, a blood flow sensor, etc.

[0043] Control circuit 80 can receive signals from sensing circuitry 86 and accelerometer circuit 90 for detecting physiological conditions and for controlling therapy delivered by therapy delivery circuit 84. Control circuit 80 can include a processor configured to execute various monitoring and therapy delivery protocols stored in memory 82 and to control sensing circuitry 86 and therapy delivery circuit 84 accordingly. Power source 98 can include one or more rechargeable or non-rechargeable batteries, and provide power to each of control circuit 80 and the other circuits 82, 84, 86, 88, and accelerometer circuit 90, as needed.

[0044] Accelerometer circuit 90 includes an accelerometer 12, which can be encapsulated by the housing 25 of IMD 10. However, it is recognized that when IMD 10 is coupled to one or more medical electrical leads, the accelerometer 12 can be carried by the leads, e.g., along a distal portion of the leads, and coupled to circuit 90 within the housing 25 via electrical conductors. Accelerometer 12 Figure 1 may include a one-axis, two-axis, or three-axis accelerometer as described above. Each axis of accelerometer 12 can be defined by a piezoelectric element, a microelectromechanical system (MEMS) device, or other sensor element capable of generating an electrical signal in response to changes in acceleration imparted on IMD 10 and subsequently on the sensor element, e.g., by converting the acceleration into a force or displacement of the accelerometer sensor element that is converted into an electrical signal by the accelerometer sensor element.

[0045] Each accelerometer sensor element generates an acceleration signal corresponding to a vector component aligned with the axis of the sensor element. Each accelerometer sensor element generates a DC component corresponding to a vector component of the force of gravity or other force exerted on the patient along the respective accelerometer axis. Each accelerometer sensor element generates an AC component related to acceleration vector components due to patient motion along the respective axis. The accelerometer circuit 90 can include a low pass filter having a relatively low cutoff frequency (e.g., less than 2 Hz or 1 Hz or less) for generating a patient body posture signal related to the body position of the patient relative to gravity. As an example for generating a body acceleration signal, the accelerometer circuit 90 can also include a band pass filter having a band pass from 1 to 30 Hz, 1 Hz to 20 Hz, or 1 to 10 Hz. The band pass filtered body acceleration signal is related to acceleration due to patient body motion, e.g., during patient body activity and during a fall and fall impact. Each of the low pass filtered body posture signal and the band pass filtered body acceleration signal are received by the control circuit 80 for detecting patient falls as disclosed herein. The accelerometer circuit 90 can include one or more analog to digital converters (ADCs) for providing multi-bit digital signals to the control circuit 80 for analysis and processing to detect patient falls.

[0046] A band pass filtered patient body activity signal can also be generated by the accelerometer circuit 90 for use by the control circuit 90 in determining a measure of patient body activity. The band pass filtering cutoff frequencies used to generate the patient body activity signal can be different than the band pass filtering cutoff frequencies used to generate the body acceleration signal used to detect acceleration signal spikes for fall detection. The patient body activity measure can be determined based on changes in direction of the body posture vector determined from the low pass filtered body posture signal and high amplitude and / or high slope or frequency body acceleration signals to confirm fall detection criteria. For example, a low constant level of patient body activity measure following a detected acceleration signal spike supports a positive detection, while a relatively high patient activity measure can indicate that the patient is engaged in body activity and the fall detection is false.

[0047] Fragments of patient physical activity metrics and / or patient activity signals prior to a fall detection can be included in the fall detection data stored in memory 82, e.g., along with the raw accelerometer signals, generated body posture signals, and / or generated body acceleration signals, to provide useful information to understand what the patient was doing prior to the fall or what might have caused the fall. Fragments of patient activity metrics and / or patient activity signals after a fall detection can be stored in memory 82 with the fall detection data. In some cases, the patient's physical activity after a fall can indicate a false fall detection or a fall recovery, but in other cases, the patient's physical activity can be due to a seizure or a reaction to pain. Storage of patient activity signals after a fall detection can be used to verify future fall detections, e.g., based on analysis of the post-fall patient activity signal morphology, as described below in connection with FIG. 6. Figure 4

[0048] For example, when IMD 10 is embodied as a pacemaker, control circuit 80 can determine patient physical activity metrics from accelerometer signals at a desired frequency for use in determining a sensor indicated rate (SIR). The physical activity metrics can vary between a minimum rest level and a maximum activity level associated with maximum motion. In some examples, the activity metric is determined as an activity count. Control circuit 80 can include a counter for tracking the activity count during an activity count interval (e.g., a 2-second interval) when the patient physical activity signal from accelerometer sensor 90 crosses a threshold value a number of times. The count at the end of each activity count interval is related to patient physical motion during the activity count interval, and thus to patient physical activity. The threshold value applied to the accelerometer signal, which causes the activity count to increase when crossed by the motion sensor signal, can be a default threshold value or a programmable threshold value, or can be an automatically adjusted threshold value. An example method for obtaining an activity count over an n-second interval is generally disclosed in U.S. Patent No. 5,720,769 (van Oort).

[0049] In other examples, an activity metric can be obtained from a motion sensor signal by integrating or summing the motion signal samples over an activity count interval (e.g., a two-second interval, although longer or shorter time intervals can also be used to determine the activity metric). The activity metric can be used to identify an ADL range corresponding to normal daily activities, such as walking around the house, driving a car, light tasks, etc. Other example methods for determining patient physical activity metrics are generally disclosed in U.S. Patent No. 6,449,508 (Sheldon et al.).

[0050] ​Other types of sensors can be included in the sensing circuit 86, which can generate signals related to patient physical activity or other patient physiological conditions. Such sensors include heart rate, respiratory activity (such as minute ventilation), or blood or tissue oxygen saturation sensors. Other types of sensors can be used to provide signals related to patient physical activity or other physiological conditions to the control circuit 80 for use in confirming a detected fall or providing pre-fall or post-fall data or information stored with detected fall data to the external device 40 for review by a clinician.

[0051] The control circuit 80 can receive body acceleration signals, body posture signals, and patient physical activity signals from the accelerometer circuit 90. Each of these signals can include one or more single-axis signal components and / or any combination of two- or all three-axis signal components. As described above, each signal is filtered over a bandwidth that includes signal content related to fall acceleration, body posture changes relative to gravity, and patient physical activity, respectively. The control circuit 80 can include processing circuitry for sampling, averaging, and analyzing each of the body acceleration signals, body posture signals, and in some examples, patient physical activity signals generated by the accelerometer circuit 90 for fall detection in accordance with the techniques disclosed herein.

[0052] The control circuit 80 performs responses to fall detection, which can include storing fall-related data in the memory 82, generating an alarm or notification (which can be transmitted by the telemetry circuit 88), adjusting sensing performed by the sensing circuit 86 to monitor other physiological signals of the patient, and / or adjusting therapy delivered by the therapy delivery circuit (which can include turning therapy on or off).

[0053] The telemetry circuit 88 can include a transceiver and antenna for communicating with the external device 40 via the communication link 42, as described above in connection with Figure 1 The telemetry circuit 88 can also include a transceiver and antenna for communicating with the external device 40 via the link 56, e.g., via Wi-Fi or Bluetooth®, as described above in connection with Transmitters and receivers in communication with the personal device 50. Communication with the personal device 50 can be performed to facilitate detection and verification of patient body posture based on patient body posture signals and information received from the personal device 50. As described below, the control circuit 80 can control the telemetry circuit 88 to transmit interrogations or pings to the external device 40 or the personal device 50 to request confirmation or a data log related to times when the patient was most likely in an upright position and to transmit notifications or alerts related to detected patient falls to the external device 40 or the personal device 50. The telemetry circuit 88 can receive "fact" data from the external device 40 and / or the personal device 50 to confirm fall detections and, in some cases, to learn of missed fall detections. The control circuit 80 can use the fact data to adjust fall detection control parameters and methods used in fall detection to increase the sensitivity and / or specificity of the fall detection technique.

[0054] Figure 4 is a flowchart 200 of a method performed by the IMD 10 to detect patient falls according to one example. At block 252, the control circuit 80 determines a reference vector from body posture signals. At block 254, the control circuit 80 samples the body posture signals to obtain a current body posture signal vector. The current body posture vector and the reference vector are analyzed to detect a directional change from the reference vector to the current body posture vector that meets fall detection criteria. The directional change in the body posture vector is related to a change in the patient body posture relative to gravity. In some examples, the reference vector determined at block 252 is a body posture vector corresponding to an upright body posture and is based on body posture signals acquired at times when the patient was verified to be in an upright position, e.g., as instructed during an office visit or established based on signals from the external device 40 or the personal device 50 indicating an upright posture. In some examples, the control circuit 80 can verify that other requirements were met to establish the reference vector from a stable, reliable body posture signal. Such requirements can include verifying that the patient body activity was low (e.g., a patient body activity metric was below a predetermined threshold level) based on a body activity signal generated from the accelerometer signals and / or verifying that the reference vector magnitude was about 1g due to the force of gravity acting on the patient without other acceleration forces acting on the patient. The upright reference vector can be stored in the memory 82 and retrieved from the memory 82 for comparison to the currently sampled body posture vector. Examples of techniques for establishing the upright or vertical reference vector at block 252 are described below in connection with FIGS. 3-5. Figures 5 to 7 Examples of techniques for establishing the upright or vertical reference vector at block 252 are described.

[0055] In other examples, the reference vector determined at block 252 is a sampled body posture vector from the low-pass filtered accelerometer signal at a point in time prior to the sampled body posture vector obtained at block 254. By determining a reference vector at a prior point in time, the control circuit 80 is able to detect a relative change in direction of the body posture vector from the prior point in time to the current point in time. The reference vector determined at block 252 can be a body posture vector determined six seconds prior to the current sampled body posture vector obtained at block 254, for example.

[0056] At block 254, the control circuit 80 samples the body posture signal received from the accelerometer circuit 90 to detect changes in the patient’s body posture in accordance with a fall detection protocol. For example, the body posture signal can be sampled at selected times during the day (or night) or continuously at a desired sampling rate. The sampling can include sampling the body posture signal at a desired sampling rate and then averaging the sampled signals over a predetermined averaging time interval to obtain an average body posture vector. In one example, the body posture signal is sampled at 32 Hz or less and the sampled signals taken over a one to two second time interval are averaged to obtain an average body posture vector for an n second interval. When a three-dimensional accelerometer signal is received, each of the sampled x, y, and z components of the body posture signal can be averaged over the predetermined time interval to obtain an average x, y, and z axis component of the body posture signal. The average body posture vector at the n second point in time can then be determined as a currently sampled body posture vector defined by the average x, y, and z components. When the averaging time interval is two seconds, the reference vector can be an average body posture vector obtained six seconds prior to the current body posture vector in one example.

[0057] At block 256, a measure of the change in direction between the reference vector and the current body posture vector is determined. As discussed in connection with Figure 2 the change in direction can be computed as the angle between the reference vector and the current body posture vector by determining the inverse cosine of the product of the dot product of the current body posture vector and the reference vector divided by their magnitudes. This angle is related to the change in the patient’s body position relative to gravity. Alternatively, the scalar result of the dot product of the reference vector and the current body posture vector can be determined as the measure of change in direction related to the angle between the reference vector and the body posture vector. In other examples, other methods can be used to determine a measure related to the change in direction (or angle) between the current body posture vector and a previously determined reference vector, for example, by determining and analyzing the differences between the respective x, y, and z components. Another method of determining a measure of the change in direction of the body posture signal is described below in connection with Figure 10

[0058] ​At block 258, the determined angle or other measure of directional change between the current body posture vector and the reference signal is compared to a fall detection threshold. It is expected that when a patient falls, the body posture vector direction will change by a threshold amount, e.g., corresponding to a change from an upright to a non-upright posture, within a predetermined time interval, e.g., within two seconds or less, within four seconds or less, or within six seconds or less. The threshold change can be an angle of at least 45 degrees, at least 50 degrees, at least 60 degrees, or other predetermined threshold. In one example, if the angle between the current body posture vector and the reference vector is at least 57 degrees, the control circuit 80 can determine that the body posture vector direction change satisfies the fall detection criteria. When the scalar result of the dot product divided by the vector magnitudes of the current body posture vector and the reference vector is used as the measure of directional change, the threshold can be set equal to the cosine of the corresponding threshold angle, e.g., the cosine of 57 degrees or approximately 0.54. In some examples, the product of the vector magnitudes can be assumed to be 1 g, thereby eliminating the division step.

[0059] However, the fall detection criteria used to detect body posture vector direction changes is not limited to detecting threshold changes from an upright or vertical body position. Some patients can roll off a bed, in which case the body posture vector direction can change as a result of the patient rolling from a substantially horizontal position on the bed to the floor during the fall. In some cases, the patient can land in a substantially horizontal position, but can also land in a non- or semi-horizontal position, e.g., ending up in a substantially seated position on the floor. Thus, in some cases, the reference vector can be a body posture vector four to six seconds earlier than the current body posture vector, and the direction change between the reference vector and the current body posture vector corresponds to the patient's roll.

[0060] Using Figure 2 Using the example axis orientation shown, in which the longitudinal axis, which is the x-axis of the accelerometer 12, is aligned substantially perpendicular to the direction of gravity when the patient is standing upright, the two radial axes, e.g., the y-axis and the z-axis, can contribute the most to the directional change when the patient rolls off the bed. Thus, as one example, determining the directional change from the reference vector can include analyzing the change in each vector component (each accelerometer axis signal) over a one to six second time period, or analyzing the directional change of a composite vector of the three vector components to determine if a threshold directional change is reached, which can correspond to the patient's roll motion or rotation. The threshold directional change from a reference vector determined a predetermined time interval earlier than the current body posture vector can be detected without requiring or verifying that the reference vector corresponds to an upright patient posture. In some examples, the current body posture vector can be compared to both an upright reference vector established previously when the patient was in a known upright position and a previous body posture vector determined four to six seconds earlier than the current body posture vector.

[0061] When the change in body posture vector direction at block 258 does not satisfy the fall detection criteria, control circuit 80 continues to sample the body posture signal at block 254. In some examples, the current body posture signal becomes the reference vector for future comparisons with body posture vectors for detecting a change in body posture vector direction that is greater than the change threshold and occurs within a predetermined time interval, which can be a multiple of the body posture signal averaging time interval. Thus, the current body posture vector (i) can be stored in memory 82 as the reference vector for the determination of the i+3 body posture vector six seconds later, as an example, when a new body posture vector is determined every two seconds. In other examples, the current body posture vector (i) can be stored as the reference vector to be compared to the i+1 body posture vector, the i+2 body posture vector, and / or the i+n body posture vector.

[0062] When the change in body posture vector direction relative to the reference vector at block 258 satisfies the fall detection requirements, control circuit 80 samples the bandpass filtered body acceleration signal received from accelerometer circuit 90 (block 260). The body acceleration signal can be sampled at a relatively higher frequency than the sampling rate of the body posture signal to enable detection of relatively high frequency acceleration spikes that occur with a fall. The body acceleration signal can be sampled over a fall window, which is a time window that encompasses the body posture change detected at block 258, during which a possible fall associated with the body posture change can be detected. For example, the body acceleration signal can be buffered in memory 82, and in response to the change in direction of the body posture vector, the body acceleration signal can be sampled at a rate between 64 Hz and 256 Hz for up to 10 seconds before and 5 seconds after the body posture change detected at block 258. The body acceleration signal can be sampled at 128 Hz over a drop window that extends one second before and four seconds after the time of the body posture vector that resulted in the threshold change in direction detection at block 258.

[0063] In some examples, a single axis body acceleration signal is sampled at block 262 and compared to spike detection criteria. The single axis can be sensitive enough to detect high amplitude spikes caused by a fall, and sampling a single axis can reduce current draw from the power source. For example, the single axis signal used for spike detection that is sampled during the fall window can be the accelerometer axis that is the radial axis or the minor axis, which can experience less variation due to patient ADLs and IMD rotation or orientation changes. However, when a single axis is used, the best axis selected can vary between devices, implant location, patients, and time, as the axis that results in the most reliable spike detection due to a patient fall can vary with these and other factors.

[0064] In other examples, two-axis or all three-axis signals can be sampled to detect an amplitude of a body acceleration signal at block 262 that is equal to or greater than an amplitude threshold. A combination of x-axis, y-axis, and / or z-axis components can be selected as the body acceleration signal. The combination of axis signal components can be configured with hardware rather than software or firmware processing functions for determining the combination of axis signal components to conserve power 98. The amplitude of the body acceleration signal can be determined as the amplitude of a single-axis sample point, the sum of the absolute amplitudes of two-axis or all three-axis signals at a simultaneous sample point, or the sum of the squares of the amplitudes of two-axis or all three-axis signals at a simultaneous sample point.

[0065] At block 262, control circuit 80 can detect a body acceleration signal spike based on a threshold amplitude and / or threshold slew rate or other spike morphology criteria. As previously described, a positive slew rate threshold can need to be met by a positive slew rate and then a negative slew rate threshold can need to be met within a time limit after the positive slew rate to detect a body acceleration signal spike. At block 262, the absolute value of the body acceleration signal can be compared to an amplitude threshold, for example, by rectifying the ADC signal output of a bandpass filtered accelerometer signal and passing the rectified signal to a comparator. If the body acceleration signal does not meet the amplitude threshold and / or one or more slew rate thresholds during a fall window, a fall is not detected. Control circuit 80 continues to sample the body posture signal to detect a change in posture that meets the fall detection criteria. In some examples, both a threshold change in body posture and a body acceleration signal spike need to be detected within a predetermined time interval of each other, referred to herein as a "fall window" or "fall time window," in order to detect a fall.

[0066] When a body acceleration signal spike is detected at block 262 within a fall window that includes a detected body posture change, at block 264, IMD 10 can detect a patient fall. In some examples, prior to fall detection at block 264, at block 263, control circuit 80 can verify that other fall detection criteria are met. For example, control circuit 80 can analyze the body acceleration signal to verify that the detected spike is not one of a plurality of spikes occurring at a regular cadence associated with a body activity. Control circuit 80 can compare the amplitude and / or slew rate within the entire fall window or an extended body activity analysis window that includes the fall window for detection of a plurality of spikes that can indicate a repetitive activity being performed by the patient. When a plurality of body acceleration signal spikes occur within the fall window or an extended time period that includes the fall window, control circuit 80 can not detect a fall (e.g., inhibit fall detection based on direction change and spike detection within the fall window). In some examples, control circuit 80 can determine whether a plurality of spikes above an amplitude threshold occur at regular intervals, which indicates a repetitive body activity involving body posture changes and impacts. When a plurality of spikes occur at a cadence, for example, within a longer time window than the fall window, control circuit 80 can determine that the body acceleration amplitude criteria are not met at block 265 and inhibit fall detection at block 265. However, when at least one or more body acceleration signal spikes occur within a fall window that meets spike detection criteria, a fall can be detected at block 264 when other patient body posture change criteria are also met. During some falls, such as a fall down stairs, a plurality of spikes can occur within the fall window and be accompanied by a threshold body posture change.

[0067] It should be recognized that other sensor signals can be used to corroborate the fall detection at block 263, such as blood pressure, cardiac electrical signals, patient physical activity (e.g., remaining low after a fall), or other signals generated by sensing circuit 86 and monitored by IMD 10. Such signals can be monitored to detect other patient conditions or can depend on a medical condition that makes the patient prone to falling and provide diagnostic or prognostic data to a clinician for selecting and managing therapy for the patient. In some examples, after detecting the body acceleration signal spike and the body posture change, control circuit 80 determines a patient physical activity metric at block 263 to verify that the patient physical activity is low or below a threshold. When other fall detection criteria are not met, such as if the body activity is high after the spike, the fall detection based on the body posture change and the detected acceleration signal spike can be suppressed at block 265. In some examples, when the fall detection is suppressed at block 265 (no fall is detected), control circuit 80 can generate and store data related to the suppression of the fall detection. For example, the event of the patient activity signal or other corroboration signal that caused the fall detection to be suppressed (i.e., even though the directional change and spike were detected within the fall window) can be stored in memory 82 along with the signals and / or data related to the detected acceleration signal spike and body posture change.

[0068] When all of the fall detection criteria are met, control circuit 80 detects a fall at block 264. At block 266, control circuit 80 generates fall data for storage in memory 82 and / or immediate or future transmission via telemetry circuit 88. In some examples, generating the fall data at block 266 includes generating a notification that can be transmitted to external device 40 and / or personal device 50 to notify a caregiver, clinician, or first responder of the detected fall. For example, the notification can be transmitted to personal device 50, and personal device 50 can be configured to send a text message, a phone call, or other form of messaging to a caregiver, first responder, or medical center.

[0069] The fall data generated at block 266 can include a time and date stamp and other sensor signal data acquired by sensing circuit 86 and / or any therapy delivered by therapy delivery circuit 84. For example, when IMD 10 is configured to receive cardiac electrical signals, the cardiac electrical signal events before and / or containing the detected fall can be stored with the fall data. In some examples, control circuit 80 continues to monitor the body posture signal after the fall detection to determine a recovery time between the fall detection and a subsequent body posture change (e.g., sitting up or standing up after the fall is detected). The recovery time can be included in the fall data generated at block 266. As Figure 8As described, a recovery time can be determined for prioritizing storage of the fall data in memory 82 and / or generating notifications that are transmitted to external device 40 or personal device 50.

[0070] IMD 10 can be configured to receive a fact signal at block 268 from another device, e.g., from external device 40 or personal device 50. In some examples, the fact signal can be a signal transmitted from the external device in response to a patient or caregiver initiating transmission after a detected fall. The fact signal can be transmitted in response to a prompt received from IMD 10 with a fall detection notification, requesting verification of the fall. In other examples, a list of detected falls with date and time and other fall data can be generated by external device 40 or personal device 50 based on fall data received from IMD 10. The patient or caregiver can select a fall detection corresponding to an actual fall to confirm a true fall detection, and can transmit a confirmation signal data back to IMD 10. Fall detections that are not confirmed by the fact signal are marked by IMD 10 as false fall detections.

[0071] When a detected fall at block 270 is identified as a false fall detection, control circuit 80 can adjust one or more fall detection control parameters at block 274. Prior to adjusting the one or more fall detection control parameters, control circuit 80 can determine whether a threshold number or frequency of false fall detections has been reached. As examples, the amplitude threshold, the slew rate threshold or other spike detection threshold, the direction change threshold, the body posture signal average time interval, the sampling rate, the filter cutoff frequency, the start and / or end time of the drop window relative to a detected acceleration signal spike or relative to a detected body posture change, or any combination thereof can be adjusted at block 274. In some examples, the adjustment at block 274 can include updating a reference vector corresponding to a known upright posture of the patient. In one example, a fall detection that is identified as a false fall detection based on a fact signal received at block 268 can cause control circuit 80 to increase the direction change threshold and / or to increase the amplitude threshold, the slew rate threshold, the frequency threshold, or other threshold used to detect acceleration signal spikes.

[0072] The fall detection control parameters can be adjusted at block 274 to increase the specificity of the fall detection by improving the discrimination between posture and acceleration signals corresponding to a fall determined from the posture and acceleration signals when the patient has not fallen and can be engaged in other activities, such as exercise or sports that can involve sudden posture changes and / or high acceleration signal amplitudes. As described below in connection with FIG. 5, the fall detection control parameters can be adjusted to increase the specificity of the fall detection by increasing the direction change threshold and / or the amplitude threshold, the slew rate threshold, the frequency threshold, or other threshold used to detect acceleration signal spikes. Figure 9The described adjustment of the direction change threshold and / or acceleration magnitude threshold for fall detection can include analyzing the body posture vector direction changes and / or body acceleration signals acquired during daily life activities and / or increased activity corresponding to exercise or other patient activity, respectively, to establish a range or trend of acceleration signals and body posture changes that occur during activities that the patient typically engages in that do not include a fall.

[0073] In examples where the other fall criteria at block 263 is to be satisfied based on a corroborating signal, such as patient body activity or other sensed signal, the adjustment of the fall detection control parameters at block 274 can include adjustment of the other corroborating fall detection criteria. For example, if the patient activity signal is used to verify a fall detection based on body posture changes and acceleration signal spikes, the patient activity signal can cause the fall detection to be suppressed due to post-fall activity being greater than a low activity threshold. However, if the suppressed fall detection is subsequently confirmed to be a true fall based on the fact signal received at block 268, the post-fall patient body activity threshold used to suppress the detected fall can be adjusted.

[0074] In other examples, when data and signals related to a suppressed fall detection are stored in memory 82, the morphology of the patient body activity signal that caused the fall detection to be suppressed can be analyzed such that patient body activity morphology features can be used to match post-fall patient activity signal morphology for use in confirming future fall detections. Such adjustment of the post-fall body activity signal morphology criteria can be helpful in confirming falls related to, for example, seizures, where the morphology of the post-fall body activity signal can correspond to seizure activity and should not cause the fall detection to be suppressed.

[0075] When the fact signal does not include an indication of an erroneous fall detection at block 270, the fall detection control parameters can then remain unadjusted, and control circuit 80 can return to monitoring the body posture signals and body acceleration signals for fall detection. While the fact signal is shown as being received after a fall detection, it is contemplated that the fact signal can be received after multiple falls have been detected or when no falls have been detected. The fact signal can be received from external device 40 based on patient data stored in an electronic medical record during a patient visit, hospitalization, or other time that can or can not coincide with a fall detection. In some cases, the fact signal received at block 268 can include an indication of a fall that was not detected by IMD 10.

[0076] When the fact signal indicates that no fall has been detected (the "Yes" branch of box 272), the control circuit 80 can adjust the fall detection control parameters at box 274. Any of the example parameters listed above can be adjusted at box 274 in response to the signal indicating no fall has been detected; however, in this case, the adjustment is made to make the fall detection parameters more sensitive to fall detection or based on the confirmation signal, which is less likely to result in the suppression of fall detection. For example, the orientation change threshold and / or acceleration amplitude and / or conversion rate threshold can be decreased, the average time interval used to determine body posture sampling points can be increased or decreased, the fall detection window can be widened, the filter cutoff frequency can be increased or decreased, or the sampling rate can be adjusted to increase the sensitivity of detecting falls in a given patient.

[0077] The adjustment of the fall detection control parameters at box 274 can be performed in response to a single undetected fall or a single erroneous fall detection. However, in some examples, the fall detection control parameters can be adjusted in response to the number or rate of erroneous fall detections reached or the number or rate of missed fall detections.

[0078] In some cases, the adjustment of the control parameters at block 274 in response to the threshold number of erroneous fall detections may depend on whether any missed fall detections have been reported, for example, to allow for increased specificity without reducing sensitivity. When the fact signal indicates that a missed fall detection has been recorded, the number or rate of erroneous fall detections that triggers the adjustment of the fall detection control parameters can be adjusted higher to avoid increasing the rate of missed fall detections. After any adjustments are made at block 274 as needed, the control circuit 80 can return to block 254 to monitor body posture and body acceleration signals according to the adjusted fall detection control parameters.

[0079] Figure 5 This is a flowchart 300 illustrating a method performed by an IMD system to establish a reference vector corresponding to an upright body posture, based on some examples. At box 302, the IMD 10 can transmit data to external devices, such as... Figure 1 The personal device 50 shown, as an example, could be a patient's mobile smartphone transmitting wireless communication signals to inquire or ping. The IMD 10 and the patient's personal device 50 can be configured to, for example, communicate via... Pairing for communication. At box 304, IMD 10 may transmit a ping to personal device 50 and wait to receive a reply. The transmitted ping may correspond to a request to establish communication and may include a request for a reply confirming the upright task or activity. In other examples, a ping may be transmitted to establish communication, and when a reply is received in response to the ping, IMD 10 may transmit a second signal requesting the personal device 50 to confirm the upright task or activity.

[0080] Meanwhile, IMD 10 can sample the body posture signal at block 303 and store the sampled signal data in memory 82. In some examples, IMD 10 determines a body posture vector by averaging the sampled body posture signal over an average time interval as described above and storing a body posture vector for each average time interval.

[0081] At block 320, patient's personal device 50 can receive the transmitted ping from IMD 10. In response to receiving the ping (and the associated request to confirm an upright task or activity), personal device 50 can be configured to generate a notification for display to the patient, which can include an audible alert, instructing the patient to perform a task that requires being in an upright position. The patient can be instructed to stand or sit in an upright position for 20 seconds (or other selected time interval) to enable IMD 10 to acquire a body posture signal during the upright position. The patient can respond to the notification by confirming that the upright position has been assumed using a touchscreen or other user interface of personal device 50. In some cases, personal device 50 can display a timer or other prompt to guide the patient in completing the requested task. Personal device 50 confirms the task at block 324 based on patient interaction with personal device 50 and transmits a reply confirming that the patient assumed the upright position at block 326.

[0082] When personal device 50 does not receive confirmation from the patient that the task has been performed within a predetermined time limit, e.g., within ten minutes or less, or five minutes or less, personal device 50 can cancel the notification and wait for the next ping from IMD 10. IMD 10 can be configured to listen for the reply at block 304 for a limited listening time, e.g., ten minutes or less or five minutes or less, in order to reduce current drain on power source 98. When the task is confirmed, personal device 50 can repeatedly transmit the reply at 326 during a predetermined listening period, so that telemetry circuit 88 can go to sleep and wake up to repeatedly listen during the listening period until the reply is received at block 304 or the listening period expires at block 306. If the listening period expires without receiving the reply, IMD 10 can wait for a day, a few hours, an hour, or other predetermined period of time and again transmit a ping at block 302 to cooperate with personal device 50 to reattempt confirmation of the upright position.

[0083] In response to receiving a reply at box 304, IMD 10 establishes a reference vector at box 308. The reference vector may be a recently stored body pose vector, the next body pose vector determined after receiving a reply, or a combination of two or more body pose vectors sampled within the reply's time window. In some examples, upon receiving a reply, IMD 10 begins an averaging window within which the body pose signal is averaged to determine the reference vector corresponding to an upright position or posture. Once the reference vector is established or updated at box 308, telemetry circuitry 88 may cease transmitting a ping at box 302 until a predetermined update period has elapsed, such as a week or other predetermined time period.

[0084] In other examples, Figure 5 The process may involve a third device, such as device 52 configured to communicate with the patient's personal device 50. Figure 1 (As shown). In response to receiving an inquiry ping at box 320, personal device 50 generates a patient notification at box 322 instructing the patient to use device 52 to perform a task. When the patient performs the instructed task, personal device 50 can interact with another device, such as... Figure 1 The device 52 shown is paired with a BLUETOOTH-enabled device. In this illustrative example, device 52 could be a weighing scale. The patient's personal device 50 can generate a notification instructing the patient that it is time to weigh themselves. Upon stepping onto the weighing scale, the scale and personal device 50 can pair and establish a link 54 (e.g., ...). Figure 1 As shown in the diagram, this allows personal device 50 to receive the patient's weight from the scale and confirm that a task requiring the patient to be in a standing upright position has been performed (box 324). In response to confirming the upright posture task based on communication with device 52, personal device 52 can transmit a response back to IMD 10 at box 326. IMD 10 receives the response at box 304 and establishes a reference vector at box 308 using body posture signals sampled during the time period consistent with the confirmed upright posture task. For example, personal device 52 can transmit a timestamp in its response to IMD 10 of when the patient performed the upright task, thereby allowing control circuitry 80 to identify the sampled body posture vector stored in memory with the corresponding timestamp.

[0085] In another example, IMD 10 can ping personal device 50 at a predetermined time of day when the patient is expected to perform an upright task or activity, such as daily exercise on a treadmill, walking, or sitting in a car. Personal device 50 can be paired with device 52 associated with daily activities, such as an activity tracker or car accessory. Audio system. The personal device 50 can transmit a reply back to the IMD 10 at block 326 to confirm that the patient is engaged in a daily activity, such as walking or sitting in a car. The IMD 10 establishes a reference vector at block 308 by acquiring a body posture signal in response to the reply.

[0086] In some examples, receiving a reply from the personal device 50 indicating that the patient is riding in a car or driving a car can cause the IMD 10 to temporarily disable the fall detection based on the personal device 50 being paired with the car audio system, as the patient is less likely to experience a fall, and disabling the fall detection can conserve power source 88. In this case, the control circuit 80 can disable the fall detection and schedule a time to resume monitoring for falls at block 310. The duration of time for which the fall detection is disabled can be fixed or programmable, and can be customized for a given patient, for example based on data extracted from the personal device 50 relating to average travel or commute times. In some examples, the fall detection is disabled until a signal is received by the IMD 10 from the personal device 50 indicating that it is no longer paired with the car audio system or that the patient is detected to be moving in the car.

[0087] Figure 6 is performed by the IMD system of Figure 1 A flowchart 350 of a method for establishing a reference vector corresponding to an upright posture performed by the IMD system of

[0088] When the IMD telemetry circuit 88 receives a reply from the personal device 50 at block 356, the telemetry circuit 88 can then establish a communication link with the personal device 50 and transmit a log request to the personal device 50 at block 360. The personal device 50 of the patient can acquire a log of patient activity corresponding to an upright position along with time and date information. The personal device 50 can transmit the log data to the IMD 10 in response to the request.

[0089] For example, a patient can use a weighing scale to transfer their weight to personal device 50. Personal device 50 stores the weight with a time and date stamp. In some cases, when a patient is taking diuretics, has heart failure, or for other medical reasons, the patient may follow a prescribed weighing schedule, such as daily weighing, so that the periodically acquired weight data is stored in personal device 50. In another example, personal device 50 may include a mileage tracking application that tracks the time the patient spends in a car. Time and date data corresponding to when the patient is in the car (most likely sitting in an upright position) can be stored in a log in personal device 50. In yet another example, personal device 50 may record activity or receive activity tracking from another device, such as a wrist-worn activity tracker. For example, personal device 50 may record the number of steps the patient takes throughout the day (during walking). The activity log may include the date and timestamp obtained when it is detected that the patient is walking (or jogging) and in an upright position. Various logs of patient activity associated with an upright position can be obtained directly by personal device 50 or retrieved from device 52 and stored by personal device 50.

[0090] IMD 10 receives log data at box 362 via telemetry circuitry 88. Control circuitry 80 searches stored body posture data at box 364 to match one or more timestamps and datestamps in the received log data with the timestamps and datestamps of stored body posture vectors. One or more body posture vectors obtained from timestamps and datestamps in the log data can be used by control circuitry 80 to establish a reference vector at box 366. When more than one body posture vector is identified by control circuitry 80 as having timestamps and datestamps matching the time and datetamps of an upright activity log, multiple body posture vectors can be averaged or combined by control circuitry 80 to establish a reference vector. In other examples, a single body posture vector having timestamps and datestamps matching the recorded activity time and datetamps can be stored as a reference vector in memory 82.

[0091] When two stamps are within a predetermined time interval between each other, such as 10 seconds or less, or 5 seconds or less, the time and date stamps of the body posture vector can be determined to match the recorded activity time and date stamps. In other examples, additional criteria can be applied to the selection of one or more stored body posture vectors with time and date stamps that match the recorded activity data. For example, criteria related to time of day, patient physical activity measurements, heart rate, or other standards can be applied to accept the body posture vector as a valid vector for establishing a reference vector.

[0092] Figure 7 It is based on another example for use by Figure 1Flowchart 400 shows the method for establishing reference vectors in a medical device system. Figure 5 and Figure 6 The process involves communication between IMD 10 and an external device, initiated by IMD 10 transmitting a ping via telemetry circuit 88. Current consumption of power supply 98 can be saved by avoiding powering IMD telemetry circuit 88 in listener mode to receive pings from the external device initiating the communication session. However, it is conceivable that an external device, such as external device 40 or personal device 50, could initiate communication with IMD 10 by transmitting a ping to IMD telemetry circuit 88 to establish a communication link.

[0093] exist Figure 7 At box 402, an external device can detect or receive evidence or information about body position. For example, personal device 50 can detect walking or pairing with a device associated with an upright position. In other examples, the external device can accumulate data logs with timestamps and dates, such as mileage or patient weight. The external device can transmit a ping at box 404 to request communication with IMD 10. Meanwhile, at box 420, IMD control circuitry 80 can control IMD telemetry circuitry 88 to wake up at a first relatively low rate, such as once per hour, and remain in listening mode for a predetermined period of time.

[0094] If the IMD telemetry circuit 88 has not yet received a transmitted ping (box 422), the control circuit 80 can determine at box 424 whether it is time to increase the listening rate. At box 426, for example, the control circuit 80 can increase the frequency of waking the telemetry circuit 88 to listen for pings at predetermined times of day and / or when a threshold level of physical activity is detected based on patient physical activity signals. If it is time to increase the listening rate, the control circuit 80 wakes the telemetry circuit 88 more frequently, for example, once per minute or more frequently at predetermined time intervals, such as 10 to 15 minutes, to listen for pings. If no ping is received, the control circuit 80 can switch back to waking the telemetry circuit 88 at a lower rate to listen for pings.

[0095] At block 428, the control circuit 80 can determine whether an update interval for the reference vector has expired, such as three to five days, a week, two weeks, or other selected minimum update interval. When no pings have been received for establishing or updating the reference vector over an extended period of time (YES branch of block 428), the control circuit 80 can then adjust the predetermined time of day to switch to the high rate of waking the telemetry circuit 88 at block 430. For example, according to habit, the patient is expected to weigh themselves, ride in a car, or walk at certain times of the day when the high wake rate is scheduled to occur at block 426. If the telemetry circuit 88 has not received a ping to cause the control circuit 80 to update the reference vector prior to expiration of the update interval at block 428 (since the last update), the predetermined time of day for starting the increased wake rate at block 426 can be adjusted at block 430, such as 10 to 15 minutes earlier or later, an hour earlier or later, or other time interval is shifted, to attempt to increase the likelihood of receiving the expected ping from the external device to confirm that the patient is engaged in activities involving an upright position.

[0096] Once the ping is received at block 422, the control circuit 80 can sample the body posture signal at block 434 based on the ping being an indication that the patient is currently engaged in an upright activity. The control circuit 80 can establish or update the reference vector at block 436 based on the currently sampled body posture signal in response to receiving the ping. In some examples, the body posture vector can be sampled at multiple consecutive points in time at block 434 to validate a stable body posture vector prior to establishing a new reference vector. If the control circuit 80 is unable to validate a stable body posture vector, such as based on the consecutively sampled body posture vectors being within a threshold difference of each other, the previously established reference vector can remain stored in the memory 82 as the current reference vector at block 436.

[0097] The control circuit 80 can identify the currently sampled body posture vector as "stable" by comparing the currently sampled body posture vector to the previously stored body posture vector and / or the previously established reference vector to verify that the currently sampled body posture signal is within a difference threshold of the previously stored body posture vector and / or the previously stored reference vector (which indicates that the body posture vector is stable). If the currently sampled body posture vector is greater than the difference threshold from the previously stored body posture, the currently sampled body posture vector can or can not be stored in the memory 82. In some examples, the currently sampled body posture vector is stored for use in verifying that subsequent body posture vectors are stable to update the reference vector at a later point in time. However, the previously established reference vector can continue to be used as the updated reference vector at block 436 until the sampled body posture vector is verified as stable, e.g., within the difference threshold of the previous reference vector or the previously sampled body posture vector. Once the control circuit 80 determines that the currently sampled body posture vector is stable relative to the previous body posture vector, the reference vector can be updated based on the currently sampled body posture vector at block 436.

[0098] In some examples, the telemetry circuit 88 can transmit a reply back to the sending device at block 432 in response to receiving the ping at block 422 to establish two-way communication. The external device 40 or the personal device 50 can receive the transmitted reply at block 406 and transmit a signal indicating that the patient is engaged in an upright activity or transmit a cumulative data log including a date and time stamp associated with a task or activity involving an upright posture at block 408. Rather than using the currently sampled body posture signal, the control circuit 80 can search the body posture vectors previously stored in the memory 82 at block 434 to match the time and date stamp to establish or update the reference vector at block 436.

[0099] In addition to or instead of waiting for a ping from the IMD 10 as described in connection with Figure 5 and Figure 6 The personal device 50 can be configured to repeatedly transmit a ping to the IMD 10 at block 404 when upright activity is identified (e.g., based on pairing with another device 52) or when an updated log corresponding to upright activity is available, instead of or in addition to waiting for a ping from the IMD 10 as described in connection with

[0100] Figure 8 is a flowchart 500 of a method that can be performed by the IMD 10 for determining a recovery time after a fall detection. At block 502, the control circuit 80 detects a fall, e.g., as described in connection withFigure 4 As described. At box 504, control circuitry 80 continues to monitor body posture signals. Control circuitry 80 can determine a body posture vector at the end of each n-second averaging interval. The n-second averaging interval can be the same averaging interval used to monitor body posture signals before fall detection or a shorter interval, such as 0.25 seconds, 0.5 seconds, or 1 second, to enable greater temporal resolution in determining the time used for recovery after a fall.

[0101] At box 506, control circuit 80 determines a metric for the angular change between the body pose vector detected during fall detection and the body pose vector detected after the current fall. The body pose vector detected during fall detection satisfies... Figure 4 The body pose vector at box 258 is used to determine the orientation change threshold, thus triggering fall detection. If the orientation change metric between the current post-fall body pose vector and the fall-detected body pose vector does not meet the recovery threshold, as determined at box 508, then no recovery from the fall is detected. A posture change that does not meet the recovery criteria indicates that the patient may still be lying on the ground and has not yet recovered from the fall.

[0102] In other examples, control circuitry 80 may additionally or alternatively determine, at block 506, a measure of directional change related to the angle between the reference vector and the body posture vector detected after the fall. Recovery can be detected at block 514 when the directional change between the body posture vector detected after the fall and the reference vector is less than a specified change threshold. A directional change measure between the reference vector and the body posture vector detected after the fall that is less than the change threshold can indicate that the patient has recovered and returned to the body posture before the fall. When the directional change measure between the body posture vector detected after the fall and the reference vector is still greater than the change threshold, the patient may still be on the ground or rolling on the ground after the fall. Therefore, at block 508, control circuitry 80 may require that the directional change measure between the body posture vector detected after the fall and the body posture vector detected after the fall is greater than the recovery threshold and / or that the directional change measure between the reference vector and the body posture vector detected after the fall is less than the recovery threshold.

[0103] When the recovery criteria are not met at block 508, the control circuit 80 can compare the time since the fall detection at block 510 to an alert threshold. For example, the control circuit 80 can start a timer upon detecting a fall and compare the time since the fall detection to an alert threshold time interval. If the fall was severe, the patient can be injured or unconscious and in need of assistance. The control circuit 80 can determine whether recovery has not been detected and whether the alert threshold time interval has expired at block 510, an alert can be generated at block 512. The control circuit 80 can control the telemetry circuit 88 to transmit the alert to an external device (e.g., the external device 40 or the personal device 50) to trigger an alarm and / or transmit a message to a first responder, call center, clinician, or other caregiver by text, phone, email, or other means.

[0104] If the alert threshold time interval has not expired or an alert has been generated, the control circuit 80 continues to sample the body posture signals at block 504 to detect recovery. When the change in direction between the body posture vector after the current fall and the body posture vector of the fall detection meets the recovery criteria, as determined at block 508, a recovery from the fall is detected at block 514. The control circuit 80 determines the recovery time as the time interval from the fall detection to the recovery detection at block 516 (which can be based on a number of average intervals of the body posture signals). The control circuit 80 can use the recovery time to prioritize the fall detection relative to other fall detections for purposes of data storage and reporting at block 518. In some cases, the recovery time can be so short (e.g., one second or less) that the fall detection is canceled, suppressed, or flagged as a false detection even when the patient body posture and signal spike detection criteria for detecting a fall are met. In other examples, the control circuit 80 can identify the fall as a low-priority fall in response to a very short recovery time. The recovery time determined at block 516 can be compared to one or more fall recovery time thresholds in order to prioritize the fall detection at block 518. The fall recovery time thresholds can be previously determined fall recovery times such that a fall detection with a longer recovery time than another fall detection is given a higher priority at block 518. The control circuit 80 can generally prioritize fall detections based on the recovery time, with longer recovery times corresponding to more severe, higher-priority fall detections for purposes of data storage, reporting, or other fall detection response purposes.

[0105] At block 520, the control circuit 80 performs a fall detection response in accordance with the priority determined at block 518. For example, if a previously detected fall has a recovery time that is shorter than the currently determined recovery time, the fall detection data for the previous fall can be overridden by the fall detection data acquired for the currently detected fall. If the recovery time exceeds a severity time threshold, which can be less than the alarm threshold time interval applied at block 510, the control circuit 80 can control the telemetry circuit 88 to transmit a notification to an external device to notify a clinician and / or that a medical examination to ensure for examination of injuries and / or patient follow-up or clinical review of the fall data can be warranted to manage treatment of a medical condition, adjustment of prescription medication, etc. For example, the fall data that can be determined by the control circuit 80 and stored in the memory 82 for transmission and review can include trends in the fall data, such as a trend in the frequency of fall detections and / or a trend in the severity of falls based on a trend in the fall recovery duration, so that a clinician can identify whether the patient is more frequently falling and / or suffering more severe falls.

[0106] When the recovery time is less than previously stored fall detection data, and the memory allocated for storing a complete set of fall detection data is full, the response at block 520 can include storing a limited amount of fall detection data, such as a detection with a time and date stamp, without or with limited additional signal data or other information. If the recovery time is relatively fast, e.g., within a few seconds or less than a short recovery time threshold such as 10 seconds or less, the telemetry circuit 88 can not transmit a notification or alarm at the time of the fall detection.

[0107] Figure 9 is a flowchart of a method performed by the IMD 10 for setting or adjusting fall detection control parameters in accordance with some examples. The IMD 10 can perform the method of 600 at the initial implantation of the IMD 10 or when the fall detection is first turned on to establish fall detection criteria for detecting patient falls. The IMD 10 can perform the method of 600 to adjust the fall detection criteria after receiving a fact signal indicating that a false fall detection or a missed fall detection has occurred, e.g., at block 274 of FIG. 2. Figure 4

[0108] ​At block 602, control circuit 80 monitors the accelerometer signals to determine a measure of patient physical activity. As described above, a bandpass filtered accelerometer signal, which can be a combination of one, two, or all three axis component signals, can be passed to control circuit 80 as a patient physical activity signal. Control circuit 80 can analyze the patient physical activity signal to determine a measure of patient physical activity, for example by determining an activity count related to the level of patient physical activity at predetermined time intervals, for example every two seconds. Control circuit 80 can detect patient physical activity at block 602 that is greater than a predetermined threshold, for example an activity measure that is greater than an ADL threshold. An activity measure that is greater than an ADL threshold can represent patient activity that can occur with changes in posture and relatively higher peaks in the body acceleration signal that, when spike detection and body posture changes are satisfied during normal patient activity, can result in false fall detections. The behavior of the accelerometer signal during normal patient physical activity can be used as an indication of what the lower bound for the amplitude threshold and the slew rate threshold or frequency threshold applied to the body acceleration signal should be. In some patients, a fall can occur relatively slowly with multiple relatively lower amplitude peaks in the body acceleration signal, for example when a patient stumbles and attempts to re-stabilize or steady themselves during a fall, but ultimately falls. In other cases, a fall can be sudden and subject to one large impact, for example if a patient experiences syncope and loses muscle control. Thus, characterizing the individual patient physical activity accelerometer signal during detected patient activity and using prior fall detection data, if available, can enable IMD 10 to set fall detection criteria to improve the sensitivity and / or specificity of fall detection.

[0109] At block 604, control circuit 80 can determine the peak amplitude of the body acceleration signal (or one or more of the x, y, and z signal components) during the detected patient physical activity. In some examples, control circuit 80 can determine the morphology or waveform shape of the body acceleration signal obtained from one or more axes during the detected physical activity. Control circuit 80 can process the body acceleration signal to detect the amplitude, slew rate, frequency, cadence of signal peaks, or other characterizing features of the body acceleration signal morphology during activity.

[0110] Control circuit 80 can additionally or alternatively determine body posture signal data at block 606 during the detected patient physical activity. Control circuit can determine characteristics of the body posture signal during the detected patient physical activity, such as the signal x, y, and / or z component amplitudes, the resultant body posture vector amplitude, and / or the maximum body posture vector direction change relative to a reference vector. Characterizing the body posture signal data during the detected physical activity can guide control circuit 80 in the selection of body posture direction change thresholds and other fall detection criteria.

[0111] At block 608, the control circuit 80 can set the fall detection criteria based on the analysis of the body posture signal data and the body acceleration signal data determined during the detected patient physical activity. The control circuit 80 can set a direction change threshold based on the maximum directional change of the body posture signal detected during the patient activity, which can be over a time interval corresponding to a fall detection window or a specified number of body posture signal averaging windows. The control circuit 80 can also set other body posture criteria, such as a threshold directional change that is not a change determined from a specified starting x-component, y-component, or z-component of the body posture vector associated with normal patient activity or a specified ending x-component, y-component, or z-component of the body posture vector associated with body postures that can be presented during normal patient activity. The body acceleration signal spike detection criteria can be set to avoid detection of acceleration signal spikes that occur during normal patient physical activity. The fall detection criteria to be applied to the body acceleration signal for spike detection can be set based on the analysis of the body acceleration signal during the detected patient activity to exclude detection of spikes having a morphology matching spikes during normal patient activity. The fall detection criteria can be set to avoid false fall detections during normal physical activity. For example, the control circuit 80 can set a magnitude threshold applied to the body acceleration signal to be greater than a maximum magnitude of the body acceleration signal during normal patient physical activity. If real fall detection data is available, the signal spike magnitudes and body posture change metrics acquired during real fall detections can be compared to the maximum body acceleration signal magnitude and maximum body posture change during normal physical activity, such that the acceleration signal magnitude threshold and the posture vector direction change threshold can be set to be lower than the lowest respective values corresponding to real fall detections and higher than the maximum respective values during normal physical activity of the patient.

[0112] When there is confounding data representing a true fall detection or a missed fall detection with body acceleration signal amplitudes and / or body posture signal characteristics that overlap with corresponding characteristics determined during normal physical activity, the morphology of the body acceleration signal can be used to distinguish between a fall and a patient's physical activity. For example, fall detection criteria can be set in relation to the slope, cadence, frequency power spectrum, overall morphology, or other features of the acceleration signal waveform of the body acceleration signal to distinguish between patient physical activity and patient falls based on the morphology of the body acceleration signal. The morphology of the body acceleration signal can be acquired during normal patient physical activity, which can include ADLs or more strenuous activities and stored for comparison with potential signal spikes during fall detection monitoring. A fall detection spike can be detected based on a morphology that does not match the body acceleration signal morphology acquired during normal physical activity. Detecting a non-matching morphology can be based on wavelet analysis, slope, area, differences between sample points determined for time alignment, frequency domain analysis, or other methods.

[0113] Figure 10 is a flowchart 700 of a method for detecting a body posture change that can be performed by the IMD 10 according to some examples. The method of the flowchart 700 need not establish a reference vector corresponding to an upright position. Figure 10 The reference vector in the method of is a sampled body posture vector at a time point earlier than the currently sampled body posture vector. The relative change in body posture vector direction between the previously sampled body posture vector and the current body posture vector is determined according to a direction change metric and compared to a threshold. Figure 10 The method of can be used to detect a body posture change corresponding to a fall that does not necessarily originate from an upright position, e.g., a fall from a bed or from a reclined or bent position.

[0114] At block 702, the control circuit 80 samples the body posture signal at a selected sampling rate to obtain each of the x, y, or z components of the body posture signal, which can include averaging the sampled signal components over an averaging window, e.g., 0.25 seconds, 0.5 seconds, 1 second, 2 seconds, or other averaging window. In one example, each x, y, and z amplitude component of the sampled body posture vector can be determined by averaging the sample points of each of the x, y, and z axis signals of the accelerometer 12 over an averaging window of 0.5 seconds at a sampling rate of 32 Hz, such that the x(i), y(i), and z(i) amplitude components of the body posture vector are provided once every 0.5 seconds. It will be appreciated that other sampling rates and averaging windows can be selected.

[0115] In conjunction with Figure 10In the described example, three-dimensional body posture vectors are monitored such that each of the x-axis, y-axis, and z-axis component signals are analyzed as component signals of a three-dimensional resultant body posture vector. However, it should be recognized that the method can be implemented using one-axis or two-axis signals instead of all three-axis signals for detecting body posture changes in one or two-axis directions. Figure 10

[0116] At block 704, the control circuit 80 determines the magnitude difference between the reference vector components and the corresponding body posture vector components sampled within the n average window after the reference vector. For example, the reference vector can be referred to as the ith body posture vector having x(i), y(i), and z(i) components. The subsequently sampled body posture vectors can be identified as the ith + n body posture vectors, which are obtained within the n average window after the reference vector. Each of the ith + n body posture vectors includes x(i + n), y(i + n), and z(i + n) component magnitudes.

[0117] At block 704, the control circuit 80 determines the magnitude difference between the reference vector components and the corresponding body posture vector components sampled within the n average window after the reference vector. For example, the reference vector can be referred to as the ith body posture vector having x(i), y(i), and z(i) components. The subsequently sampled body posture vectors can be identified as the ith + n body posture vectors, which are obtained within the n average window after the reference vector. Each of the ith + n body posture vectors includes x(i + n), y(i + n), and z(i + n) component magnitudes.

[0118] At block 706, a weighted sum of the vector component differences is determined for each of the n body posture vectors after the reference vector corresponding to n different time points after the reference vector. In one example, the magnitude of the reference vector component can be used as a weighting factor that is multiplied by each of the n difference values determined for that vector component. The product of each reference vector component magnitude and the n vector component differences determined for each of the x-axis, y-axis, and z-axis are summed together to determine a measure of the directional change of the body posture vector at each of the n time points. For example, the measure of directional change at a time point corresponding to n = 1, e.g., one average interval after the reference vector, can be calculated as: ​

[0119] metric(l) = X(ref) * Δx(l) + Y(ref) * Δy(l) + Z(ref) * Δz(l)

[0120] where X(ref), Y(ref), and Z(ref) each represent the magnitude of the respective x-component, y-component, and z-component of the reference vector, and Δx(l) is the magnitude difference x(i) - x(i+1), Δy(l) is the magnitude difference y(i) - y(i+1), and Δz(l) is the magnitude difference z(i) - z(i+1).

[0121] The direction change metric can be determined for each of the n time points using the reference vector component magnitudes and each of the x(i+n), y(i+n), and z(i+n) vector components of the subsequent n body posture vectors. For example, a direction change metric, metric(n), can be determined for each of the reference vector and each of the four subsequent body posture vectors obtained at 0.5 second intervals. In addition to the metric(l) determined above, the following three direction change metrics can be determined based on the same reference vector obtained at time point i:

[0122] metric(2) = X(ref) * Δx(2) + Y(ref) * Δy(2) + Z(ref) * Δz(2)

[0123] metric(3) = X(ref) * Δx(3) + Y(ref) * Δy(3) + Z(ref) * Δz(3)

[0124] metric(4) = X(ref) * Δx(4) + Y(ref) * Δy(4) + Z(ref) * Δz(4)

[0125] At block 708, the control circuit 80 can determine whether any of the n direction change metrics is greater than a change threshold. When at least one of the direction change metrics determined over the n average intervals is greater than the change threshold, a body posture change corresponding to a patient fall is detected at block 710. The body posture change detection can be performed at block 710 when at least one of the n direction change metrics is greater than the change threshold. Figure 4 The sequential direction change metrics can be determined at block 256 of FIG. 2, each based on a weighted sum of body posture vector component differences, and the body posture change detection can be made at block 258 based on at least one of the direction change metrics exceeding a change threshold. For example, the change threshold can be set such that at least one vector component difference needs to change in an amount of at least 20% to 80% of the reference vector.

[0126] When a body acceleration signal spike is detected within the fall window of the detected body posture change, a fall can be detected as described above in connection with Figure 4 In some examples, the control circuit 80 can monitor the body acceleration signal to detect a body acceleration signal spike and then analyze the body posture signal to use theFigure 10 This technique is used to detect changes in body posture within a fall window of a detected signal spike. For example, a reference vector can be derived from the body posture signal at a predetermined time interval before the detected body acceleration signal spike, and each of the subsequent n body posture vectors can be obtained within each of n averaging windows after the reference vector time point, such that the n body posture vectors span the spike detection. For example, the reference vector can be determined within an averaging interval of two to six seconds before the body acceleration signal spike detection, and the subsequent n body posture vectors can include two to six body posture vectors sampled after the body acceleration signal spike detection.

[0127] This technique, which defines the measure of orientation change as the sum of the differences in weighted vector components, is independent of any changes in the orientation of the IMD 10 relative to the patient's body that may occur over time. The measure of orientation change determined at box 706 may be related to the angle between the reference vector and the vector at the (i+n)th time point, but is independent of the change in the accelerometer axis relative to the direction of gravity due to the rotation of the IMD 10 within the patient's body. Figure 10 The technology can provide greater sensitivity for detecting falls involving a series of body posture changes that may occur during rolling, tripping, or tumbling. While this technology for detecting changes in body posture has been described for the purpose of detecting falls, it should be recognized that using… Figure 10 The technology for detecting changes in body posture can perform post-fall detection to detect changes in a patient's posture after a fall, such as rolling on the ground to confirm a fall. When the accelerometer axes are oriented such that the x-axis is vertical when the patient is in an upright position, the greatest changes in the amplitude of the body posture vector components during rolling are likely to occur on the y- and z-axis. Therefore, in some examples, Figure 10 The method enables fall detection using the y-axis and z-axis (or radial axes relative to the vertical axis and corresponding to the upright body position) to detect rolling on the ground and distinguish rolling from body activities indicating fall recovery based on posture changes.

[0128] In some examples, it can be enabled at night or when the patient is expected to be asleep. Figure 10 Methods are used to detect falls from a bed. In some examples, the amplitude of the radial axis component, which is close to 1, or the y-axis or z-axis signal components, which may be relatively close to zero when the patient is upright, can be used to enable [the detection of falls]. Figure 10 The trigger for this method. For example, a high amplitude (close to 1) in the y-axis or z-axis component might indicate that the patient is lying down, bending over, or leaning forward. If this high amplitude persists for a specified duration, such as one minute, it can be enabled. Figure 10The method of FIG. 1 can be used to detect a fall from a bed. Different orientation change metrics, e.g., as described above in connection with FIGS. 2-4, can be determined and compared to corresponding orientation change thresholds during the day or at times when the patient is expected to be awake and active to detect body position changes from an upright position that satisfy the fall detection criteria. Figure 4

[0129] The control circuit 80 can determine the orientation change metric according to one method during a portion of a 24-hour period and determine a different orientation change metric according to a different method during a second portion of the 24-hour period. In this way, the control circuit 80 can detect body position changes from an upright to a non-upright position during a first portion of the 24-hour period when the patient is expected to be at risk of falling from an upright position, and detect body position changes from a non-upright to a non-upright position during a second portion of the 24-hour period when the patient is expected to be in a lying position and at risk of falling from the lying position, e.g., rolling out of bed. The control circuit 80 can switch between determining the first and second orientation change metrics one or more times per day based on the time of day or based on detecting a high (or low) magnitude of the accelerometer axis signal component corresponding to the horizontal (or upright) orientation.

[0130] It should be understood that, in accordance with examples, certain actions or events of any of the methods described herein can be performed in a different order, omitted, combined, or entirely skipped (e.g., not all described actions or events are necessary for practicing the method). Moreover, in certain embodiments, actions or events can be performed concurrently, e.g., through multi-threaded processing, interrupt processing, or multiple processors as opposed to sequentially. Additionally, for the sake of clarity, while certain aspects of the present disclosure are described as being performed by a single module or unit for one or more examples, it is contemplated that the techniques of the present disclosure can be performed by a combination of units or modules associated with, for example, a medical device.

[0131] In one or more examples, the functions described can be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions can be stored as one or more instructions or code on a computer-readable medium and executed by a hardware-based processing unit. Computer-readable media can include computer-readable storage media, which corresponds to a tangible medium such as data storage media (e.g., RAM, ROM, EEPROM, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer).

[0132] ​Instructions can be executed by one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, application specific integrated circuits (ASICs), field programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. Accordingly, as used herein the term "processor" can refer to any of the foregoing structure or any other structure suitable for implementation of the techniques described herein. Also, the techniques could be fully implemented in one or more circuits or logic elements.

[0133] Thus, a medical device has been presented in the foregoing description with reference to specific examples. It is to be understood that various aspects of the disclosure disclosed herein can be combined in different combinations than the specific combinations presented in the accompanying drawings. It is to be understood that various modifications can be made without departing from the scope of the disclosure and the following claims.

Claims

1. A medical device comprising: An accelerometer circuit, the accelerometer circuit being configured to generate an accelerometer signal and generate a body posture signal and a body acceleration signal from the accelerometer signal; A control circuit configured to receive the body posture signal and the body acceleration signal, and: The body posture signal and the body acceleration signal are determined to meet the fall detection criteria by the following: A first body posture vector is determined based on the body posture signal at the first time point; At a second time point later than the first time point of the first body posture vector, a second body posture vector is determined based on the body posture signal; Determine a measure of directional change related to the angle between the first body pose vector and the second body pose vector; The directional change metric is determined to be greater than a threshold directional change. Detect acceleration signal spikes from the body acceleration signal; as well as Determine that the acceleration signal spike and the second time point are within each other's descent time window; A patient fall is detected in response to the body posture signal and the body acceleration signal meeting the fall detection criteria. The third body pose vector is determined at a third time point after the second time point; Determine a measure of directional change between at least one of the first body pose vector and the second body pose vector and the third body pose vector; The directional change metric is determined to meet the fall recovery criteria; In response to the orientation change metric satisfying the fall recovery criterion, the fall recovery time is determined as the time interval from the second time point to the third time point; as well as In response to detecting a fall in the patient, determine at least one of a trend in the frequency of fall detection and a trend in the duration of fall recovery; as well as Telemetry circuit, the telemetry circuit being configured as follows: The trend determined by transmission; A first fact signal is received from another device, indicating that the detected patient fall was an erroneous patient fall. In response to the telemetry circuit receiving the first fact signal, the control circuit is further configured to adjust at least one fall detection control parameter for controlling at least one of the following: generating the body posture signal from the accelerometer signal; generating the body acceleration signal from the accelerometer signal; and determining that the fall detection criteria are met.

2. The medical device of claim 1, wherein the control circuit is configured to determine the second body posture vector at a predetermined time interval later than the first body posture vector without verifying that the first body posture vector corresponds to an upright patient posture.

3. The medical device according to claim 1 or 2, wherein: The telemetry circuit is also configured to receive body posture confirmation signals from a personal device; and The control circuit is configured to establish the first body posture vector from the body posture signal in response to receiving the body posture confirmation signal.

4. The medical device according to any one of claims 1 to 3, wherein: The control circuit is also configured to determine multiple body posture vectors over time based on the body posture signal; The medical device also includes a memory configured to store each of the plurality of body pose vectors with corresponding time and date stamps: The telemetry circuitry is also configured to receive a data log transmitted from another device, the data log including a time and date stamp associated with a reference body posture; and The control circuit is configured to establish the first body pose vector by: At least one body pose vector is identified from the stored plurality of body pose vectors, the at least one body pose vector having a time and date stamp, the time and date stamp being matched with the time and date stamp of the data log associated with the reference body pose, and The first body pose vector is established based on the at least one body pose vector having a time and date stamp that matches the time and date stamp of the data log.

5. The medical device according to claim 1, further comprising: Memory used to store fall detection data; The control circuit is further configured to: It is determined that the fall recovery time is greater than the threshold recovery time; as well as In response to the fall recovery time being greater than the threshold recovery time, a priority fall detection response is executed, the priority fall detection response including at least one of generating a fall detection alarm and storing fall detection data in the memory.

6. The medical device according to any one of claims 1 to 5, wherein the control circuit is further configured to: Based on the body posture signal, a plurality of body posture vectors are determined, each of the plurality of body posture vectors including at least one body posture signal component determined at a predetermined time interval; Based on the plurality of body posture vectors, at least one orientation change metric is determined to be greater than a body posture change threshold; each of the orientation change metrics is determined by determining the sum of weighted body posture signal component differences between two of the plurality of body posture vectors; and In response to at least one of the determined directional change measures being greater than a body posture change threshold, one of the patient's fall and fall recovery is detected.

7. The medical device according to any one of claims 1 to 6, wherein: The telemetry circuit is configured to receive a second fact signal indicating a missed fall detection; and The control circuit is configured to adjust the fall detection control parameters in response to the second fact signal.

8. The medical device according to any one of claims 1 to 7, wherein: The telemetry circuit is configured to receive signals from a personal device, the signals corresponding to a pairing between the personal device and the equipment. as well as The control circuit is configured to disable fall detection in response to a signal from the personal device.

9. The medical device according to any one of claims 1 to 8, wherein: The accelerometer circuit is configured to generate signals of the patient's physical activity. The control circuit is configured to adjust the fall detection control parameters by: The patient's physical activity measurement is determined based on the patient's physical activity signals; The patient's physical activity measure was determined to be greater than a threshold activity level; Determine at least one of the following: a change in direction in the body posture vector originating from the body posture signal and a feature of the body acceleration signal corresponding to the patient's body activity metric; and The fall detection control parameters are set based on one of the features of the directional change in the body posture vector and the body acceleration signal corresponding to the patient's body activity metric.

10. The medical device according to any one of claims 1 to 9, wherein the control circuit is configured to determine that the body posture signal meets the fall detection criterion by: During the first part of the 24-hour cycle, a first directional change measure is determined based on the body posture signal; During the second part of the 24-hour cycle, the process switches from determining the first directional change measure to determining a second directional change measure based on the body posture signal, wherein the second directional change measure is determined differently from the first directional change measure and corresponds to a change from a first non-upright position to a second non-upright position. as well as In response to one of a first directional change metric that satisfies a first directional change threshold and a second directional change metric that satisfies a second directional change threshold different from the first directional change threshold, the body posture signal is determined to satisfy the fall detection criterion.

11. The medical device according to any one of claims 1 to 10, wherein: The accelerometer circuit is also configured to generate patient body activity signals from the accelerometer signals; The control circuit is also configured to: In response to the body posture signal and the body acceleration signal satisfying the fall detection criteria, the morphology of the patient's body activity signal is determined; and The morphological preservation method for patient fall detection based on the patient's physical activity signals.

Images