WO2009083032A1

WO2009083032A1 - Method to detect physical blocking episodes on an individual activity

Info

Publication number: WO2009083032A1
Application number: PCT/EP2007/064613
Authority: WO
Inventors: Haritz Zabaleta Recondo; Thierry Keller; Eric Fimbel
Original assignee: Fundacion Fatronik
Priority date: 2007-12-28
Filing date: 2007-12-28
Publication date: 2009-07-09

Abstract

The present invention refers to a method to detect physical blocking episodes on an individual activity. Said method comprises the step of acquiring a movement signal of a, at least one, body part of an individual. Movement signals can be static and dynamic acceleration and rotational speeds of said, at least one, body part of an individual. With said values, a jerk valuefor the static and dynamic acceleration measuredcan be calculated. Considering that a movement signal is stable when the determined jerk remains below anestablished threshold value for a predetermined period of time, it is determined whether de movementsignal is stable or not. If the movement signal is stable,it is determinedthat the body is quiet,and the stable movement signals are analysed, to obtain at least one user-condition parameter. If the movement signal is not stable, a physical blocking detection is performed.

Description

METHOD TO DETECT PHYSICAL BLOCKING EPISODES ON AN

INDIVIDUAL ACTIVITY

D E S C R I P T I O N

OBJECT OF THE INVENTION

The present invention relates to a method designed to detect physical blocking episodes on an individual activity. The method is based on the data acquired by a number of sensors that monitor continuously the individual while he is moving. When the individual remains in a stable position, that is, when the measured jerk is less than to a pre-established threshold for a period of time, the system parameters are readjusted.

BACKGROUND OF THE INVENTION

The freezing of gait is a common and disabling symptom of Parkinson's disease and responds poorly to medical treatment. This episodic phenomenon is often incapacitating for the affected person physical and psychologically because of injuries that could occur. The embarrassment and discomfort which people may feel as a result of their freezing episodes may discourage them to have social contacts or participate in public events, exclude them from activities that they had previously enjoyed. Freezing of gait is likely to disturb balance and thereby represents a common cause of falls in people with Parkinson's disease. Falls are initially absent early in the course of the disease, but then become increasingly prevalent as balance becomes progressively impaired, and eventually disappear again when patients become progressively immobilized in late-stage Parkinson's disease (Bastiaan R. Bloem et al, 2004, "Falls and Freezing of Gait in Parkinson's Disease: A Review of Two Interconnected, Episodic Phenomena.").

It is impossible to forecast whether a person will suffer from freezing or not, although it has been observed that it is slightly more common in people whose initial Parkinson's symptoms include difficulties with gait and problems with balance. It is also far more likely to occur in people who have had Parkinson's for some time and who have been on a treatment with L- dihidroxi-fenilalanina, commonly known as levodopa, for a number of years.

However, as freezing can occur in people who are not being treated with levodopa, the condition cannot be simply described as a side effect of the medication. It has been suggested that other mechanisms in the brain besides dopamine might be involved in freezing. It is not clear what the cause of the freezing is, but it is know that it become worse if a person gets anxious, in crowded places, during entry of doorways, elevators and narrow spaces such as corridors or sudden change of the walking surface.

Besides freezing events subjects with idiopathic Parkinson's Disease suffer also from changes in movement capabilities caused by dopaminergic medications. This medication is periodically administrated and causes a cycle in movement capabilities. Right after administering levodopa ("on" phase of the medication cycle) the motor control is improved, however dyskinesias (involuntary movements) may occur as side effect. At the end of the medication cycle ("off phase of the medication cycle) the movements become slower and patient can no longer walk, is rigid, and has tremor.

Falling may also occur during these states. This phenomenon is known as on-off movement fluctuations and affects also to the movement parameters in idle movements (quiet body).

Concerning the freezing phenomenon several approaches have been made to overcome this problem. First and most simple are the ones, which the patient can do himself, such as weight shifting methods, sound and mental rhythm, or the installation of simple alerting means, like floor strips in areas were freezing is more likely to appear. However, some of them will not work in other unfamiliar situations. A basic approach to this problem, for example US 6,330,888, is a beam that is displayed on the ground, providing a virtual target to step over to help overcome freezing episode. However it has to be turned on and off manually or with a weight detector, whereas during blocking episodes the user is not capable of controlling the movements, so it is difficult to implement such an activating methodology. Other devices, disclosed in international patent applications WO

01/087411 and WO 2003/039662, focus on cue generation devices, acoustic, visual and electrical, according to the walking rhythm of the user, but again, their cue is generated continuously while the device is turned on only.

For an effective unblock of the freezing, the cue has to be generated only while the blocking episode appears, and leave the control of the body movement to the motor control system during the time that the movement is working properly.

There are several patents that claim the use of motion sensors such as accelerometers and gyroscopes attached to the body to monitor the body movement. Accelerometry as a technique is being used to analyze body movements since 1970s (J. R. W. Morris, "Accelerometry-A Technique for the Measurement of Human Body Movements", J. Biomechanics, 1973, vol. 6, Pergamon Press, Great Britain, pp. 729-736.). MEMS technology accelerometer appeared in 1990s when e.g. Analog Devices started to manufacture the MEMS ADXL50 accelerometer, introduced in 1991 and in volume production by 1993, as the first commercially available accelerometer.

A body movement data acquisition system with possibility of off-line post processing of data is, for example, described in the doctoral thesis entitled "Ambulatory monitoring of motor functions in patients with

Parkinson's Disease using kinematic sensors" Salahan A., 2006. Such systems are not suitable to overcome movement blocking which require a real time detection of blocking as performed with herein described invention.

In order to have an accurate blocking detection the system needs to have user-condition parameters detected and updated frequently. These parameters can be determined during periods of quiet body, that is, the subject remains quiet in a certain position no matter in witch position lying, sitting or upright standing

Residual body limb movements e. g. body sway produces movement data that can serve for characterizing the user condition automatically. The used condition parameters can be used to automatically tune the blocking detection module at discrete times.

BRIEF DESCRIPTION OF THE INVENTION

The invention relates to a method to detect physical blocking episodes, also known or referred as freezing, on an individual activity as, for example, walking.

The present invention discloses the steps needed to carry out how to detect a physical blocking episodes and whether the body is quiet or in motion.

If the body is quiet a user-condition identification is done. The user- condition parameters are determined and stored in a memory. These parameters are used to recalibrate the blocking detection unit. Else, if the body is in motion the system distinguishes between voluntary movement and movement blocking or freezing.

Detection of physical blocking is performed continuously and requires to carry out, firstly, an acquisition of movement signals of at least one body part of an individual. Said body parts can be selected from, for example, the trunk, the thigh, the shank or the foot. The movement signals comprise of acceleration and rotational speeds. Sensors may be attached to more than one body part. Once the static and dynamic acceleration is acquired, jerk values are determined for the accelerations measured. As it has been previously disclosed, jerk is the rate of change of the acceleration.

The movement signals are used to determine if the body is quiet. To detect the difference between quiet body and body movement the derivative of acceleration, jerk, can be used as measure (Fimbel, E. et a/, 2003, "Event identification in movement recordings by means of qualitative patterns"). During quiet body the movement signals remain more or less stable, referring in this situation to the movement signals as stable signals. The jerk of the movement signal remains below a predetermined threshold value, whereas for walking or other movements the jerk presents episodes with values higher than the predetermined threshold. These higher values can be named as jerk peaks. The predetermined threshold value is dependent on the filter characteristics of the filtering module. For example, when the movement signal is sampled with 1000Hz and low pass filtered with a third order butterworth filter at 10Hz, a jerk threshold value of 0.2 g/s is a sufficient value for distinction of quiet body or movement. If there is no jerk peaks for longer than a determined period of time, e.g. 4 s, the signals are considered stable and therefore the body is thought to be quiet. Moreover, if the system is equipped with more than one accelerometer, one in each lower limb segment for example, the system is able to distinguish between different types of quiet body, e.g. between quiet lying, quiet sitting and quiet standing by comparing the accelerations measured with the values of the following tables.

State Lying

Body part Vertical acceleration Horizontal acceleration

Thigh O g -i g

Shank O g -i g

State Sitting

Body part Vertical acceleration Horizontal acceleration

Thigh O g -i g

Shank i g O g

State Quiet standing

Body part Vertical acceleration Horizontal acceleration

Whenever the movennent signals indicate the body is quiet the user- condition identification unit analyses the stable movennent signals and stores the user-condition parameters in memory. The knowledge about the actual user condition is necessary to adjust the jerk thresholds to the activity states of the user caused for example by medications. The user-condition changes steadily during medication cycle, typically a cycle of 6 hours, and therefore, the user-condition identification is only required from time to time, and therefore the analysis only needs to be performed, for example, 30 seconds after the previous parameter identification. In case this time has not been elapsed the signal is considered not stable.

If the signal is not stable, that is, the body is not quiet, blocking detection is performed.

On a first aspect of the invention, the physical blocking detection can be performed considering the step duration. An increase in the number of steps for a certain time period, in other words, a continuing decrease of the step duration, can be considered as a blocking episode. In this case, a timestamp can be stored and an output signal can be generated.

On a second aspect of the invention, the physical blocking detection can be performed considering a dominant frequency component method. The movement signals can be transformed to the frequency domain, using, for example the short time fourier transform, STFT, obtaining a dominant frequency per time frame. The movement signals are first divided into overlaping or non-operlapping time frames of frame length of a duration of less than 10 seconds. Then each frame is windowed and transformed into the frequency domain. The dominant frequency is the frequency with the highest spectral power of the transformed movement signal. If the dominant frequency on each time frame considered has a continuing change towards higher frequencies, e.g. in more than 2 Hz in a given number of time frames, motor blocking or freezing is detected. The increase to higher frequencies, or threshold, that must occur to detect the motor blocking is as well dependent on the user-condition, which is determined during the user-condition identification. This implies that the same frequency increase may imply freezing in one situation and may not imply freezing in another situation. This different outcome of the method reflects the different behaviour of the body of an individual depending on the effect along a time period of the drugs dosed to a patient. If the threshold, defined according to the user-condition parameters, is reached or passed a timestamp can be stored and an output signal can be generated.

On a third aspect of the invention, the physical blocking detection can be performed considering a spectral centroid of a signal. The spectral centroid of a signal is the midpoint of its spectral density function, i.e. the frequency that divides the distribution into two equal parts. In this case, the signal is also divided into overlapping or non-overlapping time frames and the detection method works similar as the dominant frequency variation. If the spectral centroid has a continuing change towards a higher frequencies, e.g. in more than 2 Hz, in a given number of time frames, motor blocking is detected. A frequency threshold can be defined in order not to produce unnecessary output signals. The threshold values are dependant on the user-condition, and are determined during the user-condition identification. If the threshold is reached or passed a timestamp can be stored and an output signal can be generated. Again, the threshold depends on the user-condition parameters. As in the previous case, the same change in the spectral centroid may or may not imply freezing depending on the status of the patient, that is, his user-condition parameters.

On a forth aspect of the invention, the physical blocking detection can be performed considering a spectral power analysis method. The power spectrum approximately up to the dominant frequency corresponds to the power of the oscillations that happen during voluntary movements like cyclic movement of walking and the power above the dominant frequency corresponds to oscillations that happen in non voluntary movements. This method again divides the movement signal into overlapping or not- overlapping time frames and compares both spectrum regions above and below the dominant frequency of each time frame. If the non voluntary movements become dominant, motor blocking is detected. The threshold values are dependant on the user-condition, and are determined during the user-condition identification

Whenever motor blocking is detected the system will record all the relevant data into the memory to study the physical blocking and the output signal may include a cue generation in order to overcome the physical blocking.

The disclosed method provides the means to improve the known physical blocking detectors. The method of the invention works continuously on the individual, providing a cue only when it is necessary, without the need of any action from the individual or user. At the same time, when the individual body is quiet, the method recalculates the user-condition parameters. This adjustment is strictly necessary to be able to detect the physical blocking episodes.

BRIEF DESCRIPTION OF THE DRAWINGS

To complement the description being made and for the purpose of helping to better understand the features of the invention according to a preferred practical embodiment thereof, a set of drawings is attached as an integral part of said description, showing the following with an illustrative and non-limiting character:

Figure 1 shows a representation of the method claimed. Figure 2 shows a general block diagram of a device that performs the claimed method. Figure 3 shows a body of an individual with three possible locations for the sensors. PREFERRED EMBODIMENT OF THE INVENTION

In view of the discussed figures, a possible embodiment of a method to detect physical blocking episodes in an individual activity according to the invention is disclosed.

The following invention to overcome movement blocking is based on measuring static and dynamic accelerations and rotational speeds of specific body parts. The data from accelerometers and gyroscopes provide direct information about the body motion which are used by the processor to detect the blocking of movement.

As shown in Fig 1 , the sensor module, comprising a plurality of sensors (1 ), is in charge of sensing the body movements. MEMS accelerometers such as the ADXL 330 of Analog Devices can measure the static acceleration of gravity in tilt-sensing applications, as well as dynamic accelerations resulting from motion, shock, or vibration. Its size, 4 mm * 4 mm x 1.45 mm, and power consumption, around 1 mW, makes it easily wearable and mountable on the body. Gyroscopes like ADXRS of Analog Devices are angular rate sensors that generate an angular rate proportional voltage signal. The size, 7.5 mm x 2.5 mm x 2 mm, and power consumption, 3OmW, also make them easily portable. Each kinematic sensor module comprises a combination of these two types of sensors (1 ). The acquired data are transmitted to the central unit (2).

The central unit (2) is a wearable computer that acquires and processes and generates outputs, which are sent to a display, memory (6) and/or the stimulation module (9).

The signal processing module (3) has two blocks or modules: the signal digitalization and filtering unit (4) and the signal conversion unit (5). The signal digitalization and filtering unit (4) acquires and filters the sensor data. The signal is first sampled and converted into a digital signal and then is filtered to remove random noise of the signal. The resulting data represents noise filtered digital data. The signal digitalization and filtering unit (4) provides the filtered digital data to the signal conversion unit (5). The block diagram of the signal conversion module is designed to convert the sensor data into acceleration and to provide to the blocking detection unit (8) and the user-condition identification unit (7) the needed converted and processed values. Therefore, the signal is offset corrected and gain compensated according to the data that the manufacturers provide in technical documentation. As result signal is converted into acceleration values in case of an accererometer as input signal generator and rotational speed in case of a gyroscop as input signal generator. Finally the data is converted into position, velocity, acceleration and jerk data.

The signal processing module (3) provides this information to the user- condition identification unit (7) and the blocking detection unit (8).

The user-condition identification unit (7) performs the movement signal stability test to see if the body is quiet. Therefore, the derivative of the jerk calculated by the signal conversion module (5) is used. If the jerk of the movement signal remains below a threshold value, there are no jerk peaks, for longer than a determined period of time, i.e. 4 s, the signals are considered stable and the body is quiet. If it is so, the user-condition identification unit (7) analyses the stable movement signals obtained during the period without jerk events and stores the user-condition parameters in memory (7).

The blocking detection (8) is continuously running whenever the user- condition parameters are not being updated. This unit uses the user- condition parameters in memory (6) and digitalized, noise filtered and converted movement data, to analyse them in the temporal and frequency domain, to determine the speed and regularity of the movements of the body. During physical blockings or freezing episodes the movements of the body tend to be faster and rougher than during voluntary motor activity. This behaviour could be explained in terms of frequencies.

Any change of the speed of the movement towards higher frequencies can be interpreted as possible physical blockings. If a physical blocking is detected, the signals that have been interpreted as physical blocking and the time that it happened is stored in memory (6) in order to have historical data of the physical blocking episodes of the user. An output signal is generated as well for the stimulation module (9).

The approaches of blocking detection can be divided into two broad categories: approaches that use frequency domain, like Dominant Frequency Component Analysis, Spectral Power Analysis, Wavelets, and approaches that use time frequency domain, for example, detection of periods. Several approaches can be used in the frequency domain. The linear time frequency representations such as using Short Time Fourier Transform, STFT, can be used to determine the regularity of the movement. The STFT consists in pre-windowing the acquired signal around a particular time, calculating its Fourier Transform, FT, and repeating that for each time instant. This way the frequency components of the signal are known during time and the spectral density or power spectrum can be easily calculated. The spectral density changes significantly during blocking episodes compared to voluntary movement, and therefore the blocking episode can be accurately detected. Two principal methods the dominant frequency component and signal power spectrum analysis will be explained.

In the dominant frequency component method the signal is first divided into overlapping or non-overlaping time frames and each time frame transferred into frequency domain. The dominant frequency is the frequency with the highest power and represents the frequency in which the movements are being done. When the movements have a continuing change towards a higher frequency, in other words, the dominant frequency component increases, e.g. for more than 2 Hz, freezing is detected. This threshold value is obtained from memory (6) and determined by the user-condition identification unit (7) each time that the user-condition parameters are updated.

In the spectral centroid method the signal is first divided into overlapping or non-overlaping time frames and each time frame transferred into frequency domain. The spectral centroid of a signal is the midpoint of its spectral density function. When the movements have a continuing change towards a higher frequency, in other words, the spectral centroid increases, e.g. for more than 2 Hz, freezing is detected. This threshold value is obtained from memory (6) and determined by the user-condition identification unit (7) each time that a the user-condition parameters are updated.

The power spectrum approximately up to the dominant frequency corresponds to the power of the oscillations which happen during voluntary movements and the power above the dominant frequency corresponds to oscillations which happen in non voluntary movements. The voluntary movements are smoother than the movements during freezing episodes which means that during freezing episodes, the spectrum power of higher frequencies increases during freezing episodes. This unit examines the changes of the power in each of the regions to detect physical blocking and freezing. The threshold values of the increase of the spectrum power of high frequencies are obtained from memory (6) and determined by the user- condition identification unit (7) each time that a the user-condition parameters are updated. Concerning the time domain approach, another possible way to detect blocking of movement can be measuring the step duration. Festination is a tendency to speed up the walking by taking very short steps in parallel with a loss of normal amplitude of the movements. This increases the number of steps in a certain time period and therefore measuring the step duration and step-to-step time can be used to detect freezing of gait.

In any case when freezing of gait is detected, the signals that have been interpreted as physical blocking and the time that it happened is stored in memory (6) in order to have historical data of the physical blocking episodes of the user and an output signal is generated for the stimulation module (9).

External cues that provide stimuli benefit motor activity in Parkinson's disease patients and can be used to overcome physical blocking episodes (Suteerawattananon M et a/., "Effects of visual and auditory cues on gait in individuals with Parkinson's disease". J Neurol Sci. 2004; 219(1-2):63-69). External cues can be applied in the form of visual, auditory, tactile or any other nature, to trigger movements or to provide rhythmic or spatial support to improve the quality of movements. The output generated by the blocking detection unit is used by the stimulation module (9) to generate a cue of stimulus for this purpose.

In view of this description and set of drawings, a person skilled in the art will understand that the embodiments of the invention that have been described can be combined in many ways within the object of the invention. The invention has been described according to several preferred embodiments thereof, but it will be evident for a person skilled in the art that many variations can be introduced in said preferred embodiments without exceeding the scope of the claimed invention.

Claims

C L A I M S

1.- Method to detect physical blocking episodes on an individual activity characterised in that said method comprises the steps, i- acquiring a movement signal of a, at least one, body part of an individual, said movement signal comprises a static and dynamic acceleration or rotational speeds or both of said, at least one, body part of an individual, ii- determining a jerk value for the static and dynamic acceleration measured, iii- determining whether the movement signal is stable or not, the movement signal is stable when the determined jerk remains below a established threshold value for a predetermined period of time, iv- if the movement signal is stable, v- determining that the body is quiet, vi- analysing the stable movement signals, to obtain at least one user-condition parameters, vii- if the movement signal is not stable, perform a physical blocking detection.

2.- Method to detect physical blocking episodes according to claim 1 , characterised in that the physical blocking detection comprises the steps, viii- measuring and processing the mean time duration of a stride during more than two consecutive strides, ix- if the mean time duration of the stride decreases continuously , store a timestamp and generate an output signal.

3.- Method to detect physical blocking episodes according to claims 1 , characterised in that the physical blocking detection comprises the steps, x- dividing the movement signal of up to the five last strides into overlapping time frames and transforming each time frame into the frequency domain.

4.- Method to detect physical blocking episodes according to claims 1 , characterised in that the physical blocking detection comprises the steps, xi- dividing the movement signal of up to the five last strides into non-overlapping time frames and transforming each time frame into the frequency domain.

5.- Method to detect physical blocking episodes according to claims 3 or 4, characterised in that the physical blocking detection further comprises the step, xii- obtaining a dominant frequency of said transformed movement signal for each of the time frames.

6.- Method to detect physical blocking episodes according to claims 5, characterised in that the physical blocking detection further comprises the step, xiii- if the dominant frequency obtained in step xii has a continuing change towards higher frequencies, store a timestamp and generate an output signal.

7.- Method to detect physical blocking episodes according to claims 1 , characterised in that the physical blocking detection comprises the steps, xiv- dividing the movement signal of up to the five last strides into overlapping time frames and transforming each time frame into the frequency domain.

8.- Method to detect physical blocking episodes according to claims 1 , characterised in that the physical blocking detection comprises the steps,

XV- dividing the movement signal of up to the five last strides into non-overlapping time frames and transforming each time frame into the frequency domain.

9.- Method to detect physical blocking episodes according to claims 7 or 8, characterised in that the physical blocking detection further comprises the step, xvi- obtaining a spectral centroid of said transformed movement signal for each of the time frames.

10.- Method to detect physical blocking episodes according to claims 10, characterised in that the physical blocking detection further comprises the step, xvii- if the spectral centroid obtained in step xvi has a continuing change towards higher frequencies, store a timestamp and generate an output signal.

11.- Method to detect physical blocking episodes according to claim 1 , characterised in that the physical blocking detection comprises the steps, xviii- dividing the movement signal of up to the five last strides into overlapping time frames and transforming each time frame into the frequency domain.

12.- Method to detect physical blocking episodes according to claim 1 , characterised in that the physical blocking detection comprises the steps, xix- dividing the movement signal of up to the five last strides into non-overlapping time frames and transforming each time frame into the frequency domain.

13.- Method to detect physical blocking episodes according to claims 11 or 12, characterised in that the physical blocking detection further comprises the step,

XX- obtaining a dominant frequency of said transformed movement signal for each of the time frames, xxi- obtaining a spectral density below and a spectral density above the dominant frequency for each of the time frames.

14.- Method to detect physical blocking episodes according to claim 13, characterised in that the physical blocking detection further comprises the step, xxii- if the spectral density obtained in step xxi above the dominant frequency obtained in step xviii becomes prevalent over the spectral density obtained in step xx below the dominant frequency obtained in step xviii store a timestamp and generate an output signal.

15.- Method to detect physical blocking episodes according to any of claims 1-14, characterised in that the, at least one, body part is selected from the trunk, the hip, the thigh, the knee, the shank the calf and the foot.

16.- Method to detect physical blocking episodes according to any of claims 1 -15, characterised in that the output signal triggers a stimulus cue generation.