WO2021007605A1

WO2021007605A1 - Method for detecting the intensity and/or the extent of body movements of a person

Info

Publication number: WO2021007605A1
Application number: PCT/AT2020/060270
Authority: WO
Inventors: Heinrich Garn; Bernhard Kohn; Markus WASER; Philip TAUPE; Birgit HÖGL; Stefani AMBRA
Original assignee: Ait Austrian Institute Of Technology Gmbh; Medizinische Universität Innsbruck
Priority date: 2019-07-17
Filing date: 2020-07-14
Publication date: 2021-01-21
Also published as: EP3998942A1; AT522841A1

Abstract

The invention relates to a method for detecting the intensity and/or the extent of the body movements of a person (1) while the person sleeps, comprising the following steps: ascertaining a time-dependent activity signal (MA) comprising a number of movement values assigned to successive time points, the movement values characterizing the extent and/or the intensity of body movements, in particular leg movements, of the person (1) at respective time points while the person sleeps, and the activity signal (MA) being ascertained from a height image (H) on the basis of three-dimensional image captures of the person (1) while the person sleeps; optionally applying individual signal processing steps, in particular filtering, to the activity signal (MA); ascertaining a time-dependent movement indicator signal (BI) on the basis of the activity signal (MA) by ascertaining, for each time point, whether the particular movement value of the activity signal (MA) exceeds an adaptive threshold value (TH_Z) ascertained from the activity signal (MA); ascertaining individual movement phases (BP; BP₁;..., BP₃) as continuous time segments in the time-dependent movement indicator signal (BI) in which the movement indicator signal (BI) indicates that the threshold value (TH_Z) is exceeded; and ascertaining or determining at least one feature value for each of the movement phases (BP; BP₁;..., BP₃) and accumulating the feature values thus determined, over the individual movement phases (BP; BP₁;..., BP₃), to form at least one total feature value, the at least one total feature value (GM) being considered to characterize the intensity of the body movements of the person (1) while the person sleeps.

Description

Method for detecting the intensity and / or the extent of the body movements of a person

The invention relates to a method for detecting the intensity and / or the extent of the body movements of a person during their sleep according to the preamble of claim 1 as well as an arrangement according to claim 10 and a data carrier according to claim 19.

Various methods for monitoring sleeping persons, in particular for detecting body movements of persons during their sleep, are known from the prior art. Basically, there is the possibility of identifying pathological conditions of the sleeping person based on the detected body movements and processing them accordingly. A major problem with methods known from the prior art for monitoring sleeping persons is that the complete recording of the behavior of a person during sleep usually leads to large amounts of data, although movements during sleep may only rarely take place. These data are then usually analyzed visually by sleep specialists, which has the disadvantage that movements that are carried out under a blanket, for example, are often difficult to see with the naked eye and a large number of data must be discarded due to the immobility of the sleeping person .

The object of the invention is therefore, on the one hand, to safely and reliably detect the individual movements of a sleeping person, especially when the person is covered by a blanket, and, on the other hand, to reliably detect the intensity and / or extent of the person's body movements during sleep to detect and characterize.

The invention solves this problem with a method of the type mentioned at the beginning with the characterizing features of claim 1.

A method according to the invention comprises the following steps:

- Determination of a time-dependent activity signal comprising a number of movement values assigned to one another in time,

- wherein the movement values are each characteristic of the extent and / or the intensity of body movements, in particular leg movements, of the person at a particular point in time during their sleep, and wherein the activity signal is determined from a height image based on three-dimensional recordings of the person during their sleep,

- if necessary or in particular applying individual signal processing steps, in particular filters, to the activity signal,

Determination of a time-dependent movement indicator signal based on the activity signal by determining for each point in time whether the respective movement value of the activity signal exceeds an adaptive threshold value determined from the activity signal,

- Determination of individual movement phases as coherent time segments in the time-dependent movement indicator signal, in which the movement indicator signal (B1) indicates that the threshold value has been exceeded, and

- Determining or defining at least one feature value for each of the movement phases and accumulation of the feature values thus determined over the individual movement phases to form at least one overall feature value,

wherein the at least one overall feature value is regarded as characteristic of the intensity of the body movements of the person during their sleep.

This procedure advantageously ensures that, on the one hand, relatively weak body movements during sleep are reliably detected and, on the other hand, temporal and / or spatial movement distributions can be determined that are characteristic of the intensity of the body movements of the person during their sleep. Since it is advantageously not necessary with this procedure to attach electrodes to the person's body, no additional cabling is required for monitoring the person, which would impair normal sleep behavior. It is also possible in this way to capture the entire body of the sleeping person.

The extent of the person's body movements during their sleep is to be understood in connection with the invention as how spatially spread the movement is carried out, ranging from small twitching of individual limbs to complex movements of large parts of the body or the entire body, or how the movement is spatially spreads.

In connection with the invention, the intensity of the body movements of the person during their sleep is to be understood as how often and / or how pronounced the movements are carried out by the person during a defined time unit. In order to be able to reliably determine the presence of an RBD disease in a person, it can be provided that, on the basis of the at least one overall characteristic value, a disease indicator value for the presence of an RBD disease is determined by the at least one overall characteristic value with at least one for an RBD Disease characteristic predetermined disease threshold value is compared and that in the event that the respective disease threshold value is exceeded or exceeded, the presence of the RBD disease is determined.

In the case of RBD disease - from engl. REM sleep behavior disorder - it is a REM sleep behavior disorder, i.e. a sleep disorder caused by motor activity ranging from simple twitching to complex movement patterns and / or by vocalization of the sleeping person, caused by a lack of muscle inhibition during sleep and especially in the REM sleep phase. By comparing the total characteristic value with a threshold value for the presence of an RBD disease, it can be reliably determined whether or not a person has it.

In order to determine a particularly reliable activity value for each point in time, it can be provided that, in order to generate the activity signal for each point in time, a) the difference between the current three-dimensional recording and at least one previous three-dimensional recording is determined, and b) based on the determined differences, an accumulated difference value, in particular by forming a maximum or averaging, is determined and used as a movement value of the activity signal at the respective point in time.

In order to reduce the amount of data to be processed and at the same time to increase the processing speed, a selection of pixels of the current three-dimensional recording is made between step a) and step b) and individual pixels are discarded, whereby a region of interest in the three-dimensional recording is selected or specified, it being provided in particular that the image of the person's body parts of interest is located in the region of interest.

In order to reduce the effects of incorrect or irrelevant measurements on the determined overall feature value, it can be provided that before step b) or after Step a) Pixels of the current three-dimensional recording are combined and the image resolution is reduced.

In order to ensure a particularly reliable quantification and detection of the movements of a sleeping person, provision can be made for at least one of the following values to be determined as feature values that are assigned to a movement phase:

- a constant value,

- the duration of the respective movement phase,

- the maximum or mean amplitude of the activity signal within the respective movement phase,

- the sum of the individual movement values of the activity signal within the respective movement phase,

- the change over time in the number of pixels in which the movement value exceeds the specified or another threshold value within the respective movement phase,

- the periodicity and, if applicable, the period duration of the movement phases.

A particularly simple and reliable calculation of the overall feature value can be ensured if the accumulation of the overall feature value takes place by addition or averaging or maximum value formation of the individual feature values, preferably over a predetermined time period.

In order to ensure particularly reliable detection and evaluation of the person's movements during different sleep phases, it can be provided that the predefined time segment is determined by a predefined sleep phase, in particular the REM sleep phase.

A particularly reliable statement as to whether a person suffers from an RBD disease can be made if

- a large number of overall characteristic values are formed,

a predetermined cascade of threshold value comparisons is used or the overall feature values are fed to a classification method, in particular a neural network or a support vector machine, and

- such a disease indicator value for the presence of an RBD disease is determined. A further object of the invention is to provide an arrangement with which a method according to the invention for detecting the intensity of the body movements of a person can be carried out during their sleep.

This object is achieved by the features of claim 10 according to the invention. According to the invention comprises an arrangement for the detection of the intensity and / or the extent of the body movements of a person during their sleep

an image recording unit directed at the person, which is designed to create three-dimensional recordings of the person, in particular continuously and / or at a large number of successive recording times, and to provide measured distance values to the person, and

- A processing unit connected to the image recording unit, which is designed to

- to create a time-dependent activity signal comprising a number of movement values assigned to chronologically successive points in time from a height image based on the three-dimensional recordings of the person during their sleep made by the image recording unit,

- where the movement values are each characteristic of the extent and / or the intensity of body movements, especially leg movements, of the person at a particular point in time during their sleep,

- if necessary or in particular, to apply individual signal processing steps, in particular filters, to the activity signal,

- to determine for each point in time whether the respective movement value of the activity signal exceeds an adaptive threshold value determined from the activity signal and thus to determine a time-dependent movement indicator signal based on the activity signal,

- to determine individual movement phases as coherent time segments in the time-dependent movement indicator signal in which the movement indicator signal indicates that the threshold value has been exceeded, and

- to determine or define at least one feature value for each of the movement phases and to accumulate the feature values thus determined over the individual movement phases to form at least one overall feature value,

wherein the at least one overall feature value is characteristic of the intensity of the body movements of the person during their sleep. This configuration of an arrangement according to the invention advantageously ensures that, on the one hand, the person is completely captured and even the smallest movements can be detected, even if the person is covered by a blanket. Furthermore, this embodiment of the arrangement according to the invention ensures that the intensity of the body movements of the person during his sleep can be reliably quantified on the basis of the three-dimensional recordings made.

A particularly reliable statement as to whether a person suffers from an RBD disease can be guaranteed if the processing unit is designed to determine a disease indicator value for the presence of an RBD disease on the basis of the at least one overall feature value by adding the at least one overall feature value is compared with a predetermined disease threshold value characteristic of an RBD disease and

that the processing unit is designed to determine the presence of the RBD disease in the event that the disease threshold value is exceeded or exceeded.

A particularly reliable activity signal can be generated if the processing unit is designed to generate the activity signal for each point in time

a) to determine pixel-by-pixel the difference between the current three-dimensional recording and at least one previous three-dimensional recording, and b) to determine an accumulated difference value based on the differences determined in this way, in particular by forming a maximum or averaging, and to further increase it as the movement value of the activity signal at the respective time use.

A particularly storage space and computing power saving processing and evaluation of the created three-dimensional recordings for determining an overall feature value can be achieved if the processing unit is designed to between the determination of the difference between the current three-dimensional recording and at least one previous three-dimensional recording and the determination of the accumulated To make a selection of pixels of the current three-dimensional recording and discard individual pixels,

wherein the processing unit is designed to select or specify a region of interest in the three-dimensional image in advance, in particular it is provided that the image of the person's body parts of interest is located in the region of interest.

A particularly effective reduction in the effects of erroneous and irrelevant measurements on the overall feature value can be achieved if the processing unit is designed to process pixels after determining the difference between the current three-dimensional recording and at least one three-dimensional recording that precedes it or before determining the accumulated difference value summarize the current three-dimensional recording and reduce the image resolution.

A particularly reliable overall feature value can be created if the processing unit is designed to determine at least one of the following values as feature values assigned to a movement phase:

- a constant value,

- the duration of the respective movement phase,

- the maximum or average amplitude of the activity signal within the respective movement phase

A particularly simple and reliable creation of the overall feature value can be ensured if the processing unit is designed to accumulate the overall feature value by adding, averaging or maximum value formation of the individual feature values, preferably over a predetermined period of time.

The body movements of a sleeping person during different phases of sleep can be examined particularly advantageously. For this purpose, it can be provided that the processing unit is designed to define a predetermined sleep phase, in particular the REM sleep phase, as a predetermined time segment. For a particularly reliable determination of whether a person suffers from an RBD disease, data processing by means of artificial intelligence can advantageously be used. For this purpose it can be provided that the processing unit is designed to

- to form a large number of overall characteristic values,

- to apply a predetermined cascade of threshold value comparisons or to feed the overall feature values to a classification method, in particular a neural network or a support vector machine, and

- to determine such a disease indicator value for the presence of an RBD disease.

A computer program for carrying out a method according to the invention can preferably be stored on a data carrier.

Further advantages and configurations of the invention emerge from the description and the accompanying drawings.

Particularly advantageous but not restrictive exemplary embodiments of the invention are shown schematically below with reference to the accompanying drawings and are described by way of example with reference to the drawings.

The following shows schematically:

1 shows an arrangement for the acquisition and detection of body movements of a sleeping person and an example for the definition of a height profile,

2 schematically shows a possibility of defining regions of interest,

3 shows an example of a height profile.

4a shows an example of a determined time-dependent activity signal,

FIG. 4b shows the activity signal from FIG. 4a after signal processing,

4c shows the time-dependent movement indicator signal for the activity signal from FIGS. 4a and 4b,

FIG. 5 shows feature values determined for the activity signal from FIG. 4b and the movement indicator signal from FIG. 4c,

6 shows an example for the evaluation of the intensity of the body movements of a person during different sleep phases,

Fig. 7 is a flow chart for determining whether or not a person has RBD disease. In Fig. 1, an arrangement for recording and detecting body movements of a sleeping person 1 is shown from the side. This arrangement can typically be used in sleep laboratories or similar medical monitoring facilities. The person 1 is to be examined, for example, for the presence of certain sleep disorders and is monitored for this purpose while he is sleeping in a bed 10. In order to increase sleeping comfort, the person can also be covered, at least partially, by a duvet.

Arranged above the person 1 is an image recording unit 2 which is designed to take three-dimensional recordings of the person 1. The image recording unit 2 is connected to a processing unit 3 or is in data transmission contact with the processing unit 3. In this way, recordings created by the image recording unit 2 or other data are transmitted to the processing unit 3. Optionally, the processing unit 3 can also be designed to control the image recording unit 2 at predetermined time intervals in order to create three-dimensional recordings of the person 1.

The three-dimensional recordings can be made, for example, with a time-of-flight (TOF) infrared camera which is mounted on the ceiling above the bed 10 of the sleeping person 1. Infrared rays are emitted and their reflection is measured by a sensor in the camera. Due to the known speed of light and the time elapsed between emission and measurement of the infrared rays, a three-dimensional depth image is generated in which each pixel indicates the distance to the image recording unit 2 or to the sensor of the camera. In this way, for example, 30 such three-dimensional recordings or depth images can be created per second.

A method for detecting body movements of a sleeping person, whereby three-dimensional images of the person are continuously created and determined distance values are provided at successive recording times with an image recording unit 2 directed at the person, is described in the Austrian patent application A50049 / 2018. The processing unit 3 is designed to carry out such a method.

According to the method described in A50049 / 2018, movements are detected through pixel-by-pixel changes in three-dimensional depth images over time. For this purpose, preprocessing steps such as resampling of the depth images, interpolation of defective pixels and convolution of several three-dimensional depth images over time are used carried out. The preprocessing and motion detection can be carried out in predefined regions of interest or regions of interest ROI (see FIG. 2). On the basis of this preprocessing data, motion maps or movement maps can be created, in which the movements can be shown, for example, in a false color representation. The movements in a relevant region of interest ROI can be detected quantitatively if, for example, a predefined threshold value is detected within a defined number of pixels.

According to A50049 / 2018, these three-dimensional recordings are usually created in the form of a height profile F1 (see FIG. 3) as part of a recording step a), which each has a sample value for a large number of different rays. This height profile may for example be received by depart for a number of the image pickup unit 2 outgoing rays S that are arranged like a grid and of the image pickup unit 2, each separately distance measurements di, ..., d _n to be added that the distance of the intersection point P of the surface of the person 1 with the beam S to the image recording unit 2 along the predetermined beam S.

In the height profile Fl, a number of at least two points, preferably a plurality of points P lying on rays S arranged in the form of a grid, are generally defined in space, which are on the surface of the person 1 or on the surface of an object located on or next to the person 1, such as the duvet 11 or the bed 10, lie.

The definition of the height profile Fl can therefore be done by the distance measured values d _1; d ₂ ,..., d _{n of} the height profile F1 to be defined as the distances that the points P lying on the rays S have on the surface of the person 1 opposite the image recording unit 2. The distances can be measured in different ways, for example by means of a 3D camera.

Alternatively, as shown in FIG. 1, the individual measured distance values di ', d ₂ ', d _n 'can be specified when creating the height profile as those distances that the determined points P on the surface of the person 1 from another object, in particular from the ceiling 1 1 of the examination room. It is also possible to find an interpolating curve for the individual points that is used as the height profile H. This interpolating curve can also be evaluated at a large number of three-dimensional points, so that for a number of x and y coordinate values arranged in the form of a grid, a z coordinate value is made available, which is also located on the curve.

A separate height profile H, which is shown schematically in FIG. 3, is determined and stored in a data structure for individual recording times. This height profile H has a measured distance value d ₁ for a number of elements or pixels arranged in a grid or image shape _. ..., d _n , where the distance measurement values can be determined as described above.

As shown in FIG. 2, a region of interest ROI can optionally be selected within the height profile, in which, for example, those parts of the person 1 can be found whose movement is to be monitored. Since the movements of the legs of the person 1 are to be monitored in the present case, the region of interest ROI is arranged in the middle section of the height profile H. If, on the other hand, other body regions or other regions within the height profile H are to be monitored, a corresponding different selection of the height profile H can be made.

If a height profile H is determined at each of the recording times, a data structure containing the respective height profile H is available for each of the recording times ti,..., T _p . All data structures created in this way have the same size as one another and have storage positions for the individually determined distance measurement values d _1; ..., d _{n of} the height profile.

The creation of the height profile can be done particularly easily if the individual rays S emanating from the detector unit are arranged in the form of a grid and each of the measured distance values is entered in a matrix data structure that reproduces the grid of the individual rays S or has a structure that corresponds to the structure of the grid of the individual rays S corresponds. If the grid contains 300x300 rays, for example, then the matrix data structure comprises 300x300 entries which, when considering the content of the matrix data structure as an image, can also be referred to as pixels.

At each of the recording times ti,..., T _p , a separate matrix data structure is created in each case in that the recorded at the respective positions Distance measurement values di, d _n are stored in the matrix data structure at the memory positions corresponding to the positions in the grid and kept available.

Those storage positions of the data structure in which measured distance values di, d _{n are} stored, which are located, for example, in the region of interest of the height profile, are also referred to analogously as the region of interest ROM of the data structure.

In this way, it is possible to create individual height profiles of the sleeping person or the surface of the sleeping person 1 or the blanket 11 covering them or the bed 10 located next to them for a large number of times. A data structure is thus available from which the essential movements of the sleeping person 1 can be taken.

In the first processing step following the recording step shown here, time segments Z _1; ..., Z ₃ of changes in the elevation profile H within the first region of interest, in which the measure of the temporal change in the elevation profile H exceeds a predetermined first threshold value. Furthermore, between these time segments Z _1; ..., Z ₃ lying time intervals L _1; L ₂ , ... determined.

The definition and determination of the measure for the temporal change in the height profile H can take place in different ways. In this context, a particularly simple variant provides that for individual times, in particular for all times t _1; ..., t _p , in each case a movement map MM1 is created in which for each measured distance value d _1; ..., d _n or each line of sight S or each entry k (x, y, t) of the matrix data structure, element by element or pixel by pixel, a local change measure value mm (x, y, t) for the change over time of the respective distance measurement value d _1; ..., d _n or the entry k (x, y, t) entered in the respective data structure.

After a first movement card MM1 in this way for a number of times t _1; ..., t _{p was} created, can for each individual point in time t _1; ..., t _p a movement value can be determined which is determined by accumulation over the obtained local change dimension values mm (x, y, t) of the movement map MM1 within the region of interest ROM. When determining an accumulated movement value for a specific point in time, which is entered in the temporal movement function g (t), basically all local change dimension values mm (x, y, t) that are available or formed in different ways can be used. In particular, it is also possible for the individual local change dimension values mm (x, y, t) of the movement map MM1 to be determined pixel by pixel within the region of interest ROM, with individual entries k (x, y, t) or associated with the same pixel or element of the data structure being determined Change measure values that occur within a predetermined time interval around the respective point in time t _1; ..., t _p are to be added weighted. For example, it is possible to determine a local change measure value mm (x, y, t) in a given pixel or entry at the grid position x, y of the region of interest ROM, in which the relevant last recorded distance measurement values or entries k ( x, y, t) are added up weighted, whereby the individual weights can be determined in different ways.

In the simplest case, the change over time can be made, for example, by subtracting the two measured distance values or entries k (x, y, t); k (x, y, t-1) can be determined, which were determined at the same position in immediately successive times. If the type of movement is irrelevant, the amount of the difference between the two entries k (x, y, t); k (x, y, t-1) or measured distance values can be used as a local change measure value mm (x, y, t) for the relevant entry or the relevant relevant pixel at position x, y at time t.

It is also possible to determine the change dimension value mm (x, y, t) a number of entries k (x, y, t), each within a time interval before the respective point in time t in the same pixel at position x, y , recorded, is used and a mean value km ^ x, y, t) or the possibly weighted sum km ^ x, y, t) is determined from these entries k (x, y, t). Likewise, a number of entries k (x, y, t), which were recorded within a time interval after the respective point in time t in the same pixel at position x, y, are used and from these entries k (x, y, t) a mean value km ₂ (x, y, t) or the possibly weighted sum km ₂ (x, y, t) is determined. The change value is then calculated by forming the difference between the two averages or sums created in this way km ^ x, y, t); km ₂ (x, y, t) are formed and used as the change dimension value mm (x, y, t).

A particularly simple way of determining an accumulated total measure for the temporal change in the height profile H at a point in time ti, ..., t _p for the Region ROM of interest can for example be done by summing or adding all local change dimension values mm (x, y, t) that are contained at a point in time or in a movement map MM1 (t). In addition, other procedures for accumulation can also be selected, in particular a function h can be applied to the individual values of the movement map MM1 (t) before the summation.

This function h (x) can be designed differently. In the present case, functions that contain a threshold value comparison and the respective movement value with a predetermined threshold value are particularly recommended

to compare. If the value falls below this threshold value THi, the relevant function h (x) can return a zero value which does not contribute to the accumulation, in particular the value 0. However, if the threshold value is exceeded

is exceeded, the function h (x) can return different values, in particular the function can return a predefined constant value, for example 1, which contributes to the accumulation and does not correspond to the zero value. Using an addition for, a function value for the temporal movement function g (t) is obtained which corresponds to the number of those pixels or entries in which a threshold value was found to be exceeded.

In addition, there are also other options for defining the function h (x), for example, if the threshold value TH-i is exceeded by the argument from the function h (x), the argument, ie in the specific case the respective value of the movement map MM1 are returned so that the function h has the following form, for example: h (x) = {If x <TFI ₁ Then 0, otherwise x}

In addition, the function h (x) can also be defined in such a way that it is activated in the event that the threshold value is exceeded

not the argument x itself, but the extent to which the threshold was exceeded

returned by the argument x or by the respective value of the movement card MM1: h (x) = {If x <TFI ₁ then 0, otherwise x-TPh}

In principle, for the detection of time segments Z _1; Z ₂ , Z ₃ with strong changes or changes that exceed a threshold value are a comparison of the changes over time Movement function g (t) take place with a predetermined threshold value TH _Z. Exceeds the motion function g (t) the threshold value TH _Z concerned, it is the one time section in which the motion function g (t) exceeds the relevant threshold value TH _Z, as a period of time Z _1; Z ₂ , Z ₃ of significant changes in the height profile H or of movements of person 1, or as movement phase BP _1; BP ₂ , BP ₃ , identified and kept available as such.

A movement function g (t) determined in this way is referred to below as an activity signal MA or, after signal processing steps such as filtering, as a filtered activity signal MAp.

A method according to the invention for detecting the intensity of the body movements of a person during their sleep basically follows the following sequence and the processing unit 3 is designed to carry out such a method:

First, as already described above, a time-dependent activity signal MA comprising a number of movement values is created. The movement values are assigned chronologically successive points in time and are each characteristic of the extent and / or the intensity of the body movements of the person 1 at a particular point in time during their sleep. In particular, the leg movements of person 1 can be used to derive the movement values. The activity signal MA is, as described above, determined from a height image H based on three-dimensional recordings of the person during their sleep.

4a shows such a time-dependent activity signal MA, the movement values of which characterize the body movements of the sleeping person at a particular point in time. The activity signal MA can for example be specified for individual pixels or also for predetermined regions of interest ROI.

If necessary, signal processing, for example filtering of the activity signal MA, can be carried out in a next step of a method according to the invention. FIG. 4b shows the activity signal MA from FIG. 4a in a processed, ie filtered, form as a filtered activity signal MAp. As can be seen in FIG. 4b, the filtered activity signal MAp appears smoothed in comparison with the original activity signal MA, since disruptive influences such as, for example, noise are suppressed by the filtering. To generate the activity signal MA, in the exemplary embodiment shown in FIGS. 4a and 4b, the difference between the current three-dimensional recordings of the sleeping person 1 and at least one previous three-dimensional record of the sleeping person 1 is determined pixel by pixel and based on the differences thus determined an accumulated difference value is determined and used as the movement value of the activity signal MA at the respective point in time. This accumulated difference value can be created, for example, by forming a maximum or averaging.

As the next step of a method according to the invention, a time-dependent movement indicator signal B1 is determined based on the activity signal MA or, as in the exemplary embodiment shown in FIG. 4c, based on the filtered activity signal MAp. The movement indicator signal B1 is created by determining for each point in time whether the respective movement value of the activity signal MA or the filtered activity signal MAp exceeds an adaptive threshold value determined from the activity signal MA. Subsequently, individual movement phases BP are determined in the time-dependent movement indicator signal B1 created in this way, which represent coherent time segments within the time-dependent movement indicator signal Bl within which the movement indicator signal B1 indicates that the threshold value has been exceeded. In the exemplary embodiment shown in FIG. 4c, three individual movement phases BP _{1; are recorded} within the time-dependent movement indicator signal B1 _. BP ₂ , BP ₃ identified.

At least one feature value is then established or determined for each of the movement phases, and an overall feature value GM is created by accumulating the individual feature values thus determined over the individual movement phases. This overall characteristic value GM is considered to be characteristic of the intensity of the person's body movements during their sleep. For example, a high overall characteristic value GM can indicate a high number of movements by person 1 or a high strength of the movements.

FIG. 5 shows an exemplary embodiment for feature values which were created on the basis of the movement indicator signal B1 from FIG. 4c. On the individual feature values shown in FIG. 5 for the three sleep phases BP _1; BP ₂ , BP ₃ will be discussed in more detail below. In the case of methods known from the prior art, typically only the movements of the sleeping person are quantified by parameters that are easy to calculate, such as the average number of movements per hour, the duration of movement or the duration of the intervals between movements.

With the procedure according to the invention it is advantageously possible to analyze the temporal movement distribution of the movements of the sleeping person 1 and to analyze the trend of the movement activity during the entire night or the respective sleep phase, for example

- whether the number of movements increases or decreases,

- the variance or stability of movement activity,

- the density of movement, for example whether there are time periods or sleep phases in which the person moves particularly frequently,

the distribution of movement bursts, i.e. short phases with many successive movements,

the change over time in the number of pixels in which the movement value exceeds the predetermined threshold value or a further threshold value within the respective movement phase BP; BP _1; ..., BP exceeds ₃ ,

the periodicity and, if applicable, the period duration of the movement phases BP; BP _1;

..., BP ₃ .

In addition to this analysis of the temporal distribution of movement, a spatial analysis of the movements can optionally also be carried out. For this purpose, as is shown schematically in FIG. 2, a selection of pixels of the current three-dimensional image is made, which correspond, for example, to a previously defined region of interest ROI in the three-dimensional image that is specified or selected in advance. Other pixels that are not located, for example, in this region of interest ROI, can be discarded in order to save memory space and computing power. For example, that region of the three-dimensional image in which the image of the person's body parts of interest is located can be defined as such a region of interest ROI. As shown in FIG. 2, this can relate, for example, to the legs, but also to the head, arms, hands or upper body of the person.

Additionally or alternatively, pixels of the current three-dimensional recording can also be combined and the image resolution can be reduced in such a way that defective or irrelevant pixels have a smaller effect on the differences determined or the accumulated difference value.

If, for example, only the pixels are evaluated that are located within a region of interest ROI of the sleeping person 1, the spatial movement distribution within the region of interest can be analyzed and, for example, feature values for the moving body parts or the movement propagation, ie whether, for example, movements from head downwards in In the direction of the legs or from the center of the body to the limbs.

The determined feature values can be combined as desired and, for example, analyzed with regard to sleep throughout the night or a respective sleep phase, for example the REM sleep phase. Alternatively, other time periods can also be specified. If a sleep phase is established, within which the movements of the sleeping person 1 are to be analyzed, this can be established externally within the temporal movement indicator signal B1, for example by analyzing EEG data of the sleeping person. Such EEG data are also recorded, for example, while person 1 is sleeping and the three-dimensional recordings of person 1 are being made, so that the individual sleep phases that person 1 is going through can be deduced from the EEG data. A schematic representation of such a

Movement indicator signal Bl during various sleep phases is shown in FIG. The dark areas indicate a high overall feature value GM, while the light areas are due to a low overall feature value GM.

Various values come into consideration as feature values, as they are shown schematically in FIG. 5 and are assigned to a movement phase.

For example, it can be a constant value, the duration D of the respective movement phase or the sum S of the individual movement values of the

Activity signal MA or, as in the exemplary embodiment shown in FIG. 4b, the processed activity signal MAp act within the respective movement phase BP.

In FIG. 5, the duration of the movements D _1; D ₂ , D ₃ , the maximum amplitude of the movements AP _1; AP ₂ , AP ₃ , the amplitude sum S _1; S ₂ , S ₃ or Power the movements and the number M _1; M ₂ , M ₃ , the movements per defined time unit or per movement phase BP shown. The detected movements can be characterized quantitatively using these feature values. These feature values can be used individually or in combination to characterize the intensity of the body movements of a person 1 during their sleep. For example, an overall feature value GM can be determined by accumulating all or selected individual feature values, as shown in FIG. 5, for example by adding, averaging or generating maximum values of the individual feature values. This can take place, for example, at a predetermined time period such as a predetermined sleep phase, for example the REM sleep phase.

Sleeping sicknesses are often accompanied or defined by certain atypical movement patterns that a sick person executes during his sleep. Affected patients have increased motor activity, which can range from brief twitches in individual limbs to complex movement sequences. This can reduce the quality of sleep and, in the case of RBD, can lead to injuries to the sleeper or the bed partner. Furthermore, certain sleeping diseases such as REM sleep behavior disorder, also known as RBD disease, are early markers for serious diseases such as Parkinson's or certain types of dementia. The analysis of the movements of sleeping people or their movement patterns is therefore extremely important for the diagnosis.

REM sleep behavior disorder (RBD) is a sleep disorder that is characterized by motor activity ranging from simple twitching to complex movement patterns and vocalization of the sleeper, caused by the lack of muscle inhibition (atony) during sleep, and particularly during REM sleep, as described in American Academy of Sleep Medicine: The International Classification of Sleep Disorders, Third Edition (ICSD-3), 2014. Those affected often injure themselves or their bed partners through defensive or aggressive movements. Studies in recent years have fueled the suspicion that RBD is a precursor to various a-synocleinopathies such as Parkinson's disease, multiple system atrophy or Lewy body dementia, as in Högl, Stefani, Videnovic. Idiopathic REM Sleep Behavior Disorder and Neurodegeneration— an Update. Nature Reviews Neurology 14, 40-55, 2018.

In the case of a method according to the invention, it is advantageously optionally possible that a disease indicator value is based on at least one overall feature value GM is determined for the presence of an RBD disease. For this purpose, for example, a disease threshold value T which is characteristic of the RBD disease is specified by a doctor, for example, and the overall feature value GM is compared with the specified threshold value T. Depending on whether the disease threshold value T is above or below, the presence of an RBD disease is determined.

Optionally, a large number of overall feature values GM _1; ..., GM _{4 are} formed and a predetermined cascade of threshold value comparisons are used and a disease indicator value for the presence of an RBD disease is determined in this way.

Alternatively, the overall feature value GM or, if applicable, the overall feature values GM _1; ..., GM _{4 can} also be fed to a classification method, in particular a neural network or a support vector machine, and in this way a disease indicator value for the presence of an RBD disease can be determined.

7 schematically shows a flowchart for a cascade of threshold value comparisons for four total feature values GM _1; ..., GM ₄ to determine whether a person 1 suffers from an RBD disease. This flowchart shows a rule-based decision tree in which the four overall feature values GM _1; ..., GM ₄ with a predetermined cascade of disease threshold values T _1; ..., T _{4 are} compared and such a disease indicator value for the presence of an RBD disease is determined. In the exemplary embodiment, the disease indicator value is shown in FIG. 7 in the form of a yes statement indicated by “RBD” or a no statement indicated by “No RBD” for the presence of an RBD disease.

Claims

1. A method for detecting the intensity and / or the extent of the body movements of a person (1) during their sleep, comprising the following steps

- Determination of a time-dependent activity signal (MA) comprising a number of movement values assigned to successive points in time,

- wherein the movement values are each characteristic of the extent and / or the intensity of body movements, in particular leg movements, of the person (1) at a respective point in time during their sleep, and

- The activity signal (MA) being determined from a height image (H) based on three-dimensional recordings of the person (1) during their sleep,

- If necessary, applying individual signal processing steps, in particular filtering, to the activity signal (MA),

- Determining a time-dependent movement indicator signal (Bl) based on the activity signal (MA) by determining for each point in time whether the respective movement value of the activity signal (MA) exceeds an adaptive threshold value (TH _Z ) determined from the activity signal (MA),

- Determination of individual movement phases (BP; BP _1; ..., BP ₃ ) as coherent time segments in the time-dependent movement indicator signal (Bl) in which the movement indicator signal (Bl) indicates that the threshold value (TH _Z ) has been exceeded, and

- Determining or defining at least one feature value for each of the movement phases (BP; BP _1; ..., BP ₃ ) and accumulation of the thus determined feature values over the individual movement phases (BP; BP _1; ..., BP ₃ ) at least an overall characteristic value,

wherein the at least one overall characteristic value (GM) is regarded as characteristic of the intensity of the body movements of the person (1) during their sleep.

2. The method according to claim 1, characterized in that on the basis of the at least one overall feature value (GM) a disease indicator value for the presence of an RBD disease is determined by the at least one overall feature value (GM) with at least one predetermined characteristic of an RBD disease Disease threshold value (T; T 1, ..., T ₄ ) is compared and that in the event that the respective disease threshold value (T; T _1; ..., T ₄ ) is above or exceeded, the presence of the RBD disease is detected.

3. The method according to claim 1 or 2, characterized in that for creating the activity signal (MA) for each point in time

a) the difference between the current three-dimensional recording and at least one previous three-dimensional recording is determined pixel by pixel, and b) based on the differences determined in this way, an accumulated difference value is determined, in particular by maximum formation or averaging, and is determined as the movement value of the activity signal (MA) at the respective point in time is used further.

4. The method according to claim 3, characterized in that between step a) and step b) a selection of pixels of the current three-dimensional image is made and individual pixels are discarded, a region of interest (ROI) being selected or specified in the three-dimensional image beforehand is, in particular it is provided that in the region of interest (ROI) is the image of the body parts of interest of the person (1).

5. The method according to claim 3 or 4, characterized in that before step b) or after step a) pixels of the current three-dimensional recording are combined and the image resolution is reduced.

6. The method according to any one of the preceding claims, characterized in that at least one of the following values is determined as feature values which are assigned to a movement phase (BP; BP _1; ..., BP ₃ ):

- a constant value,

- the duration of the respective movement phase (D; D _1; ..., D ₃ ),

- the maximum or mean amplitude (AP; AP _1; ..., AP ₃ ) of the activity signal (MA, MAp) within the respective movement phase (BP; BP _1; ..., BP ₃ ),

- the sum (S; S _1; ..., S ₃ ) of the individual movement values of the activity signal (MA, MAp) within the respective movement phase (BP; BP _1; ..., BP ₃ ),

- the change over time in the number of pixels in which the movement value exceeds the specified or another threshold value within the respective movement phase (BP; BP _1; ..., BP ₃ ),

- the periodicity and, if applicable, the period duration of the movement phases (BP; BP _1;

BP ₃ ).

7. The method according to any one of the preceding claims, characterized in that the accumulation of the overall feature value (GM) by addition or averaging or maximum value formation of the individual feature values takes place, preferably over a predetermined period of time.

8. The method according to claim 7, characterized in that the predetermined period of time is determined by a predetermined sleep phase, in particular the REM sleep phase.

9. The method according to any one of the preceding claims, characterized in that

- that a large number of overall characteristic values (GM _1; ..., GM ₄ ) are formed,

- that a predetermined cascade of threshold value comparisons is used or that the overall feature values (GM _1; ..., GM ₄ ) are fed to a classification method, in particular a neural network or a support vector machine, and

- That such a disease indicator value for the presence of an RBD disease is determined.

10. Comprehensive arrangement for detecting the intensity and / or the extent of the body movements of a person (1) during their sleep

- An image recording unit (2) directed at the person (1) which is designed to create three-dimensional recordings of the person (1), in particular continuously and / or at a large number of successive recording times, and to provide measured distance values to the person, and

- A processing unit (3) connected to the image recording unit (2) and designed to

- to create a time-dependent activity signal (MA) comprising a number of movement values assigned to one another in time from a height image (H) based on the three-dimensional recordings made by the image recording unit (2) of the person (1) during their sleep,

- where the movement values are each characteristic of the extent and / or intensity of body movements, in particular leg movements, of the person (1) at a particular point in time during their sleep,

- if necessary, apply individual signal processing steps, in particular filters, to the activity signal (MA),

- to determine for each point in time whether the respective movement value of the activity signal (MA) exceeds an adaptive threshold value (TH _Z ) determined from the activity signal (MA) and thus a time-dependent one To determine the movement indicator signal (Bl) based on the activity signal (MA),

- to determine individual movement phases (BP; BP _1; BP ₃ ) as coherent time segments in the time-dependent movement indicator signal (Bl) in which the movement indicator signal (Bl) indicates that the threshold value (TH _Z ) has been exceeded, and

- in each case at least one feature value for each of the motion phases (BP; ..., BP _{_3;} BP ₁₎ determine ON or set and the feature values determined over the individual movement phases (BP; BP _1, ..., BP ₃₎ to at least to accumulate an overall trait value (GM),

wherein the at least one overall feature value (GM) is characteristic of the intensity of the body movements of the person (1) during their sleep.

1 1. Arrangement according to claim 10, characterized in that the processing unit (3) is designed to determine a disease indicator value for the presence of an RBD disease on the basis of the at least one overall feature value (GM) by the at least one overall feature value (GM) is compared with a predetermined disease threshold value (T; T _1; ..., T ₄ ) characteristic of an RBD disease and

that the processing unit (3) is designed to determine the presence of the RBD disease in the event that the disease threshold value (T; T _1; ..., T ₄ ) is exceeded or exceeded.

12. The arrangement according to claim 10 or 1 1, characterized in that the processing unit (3) is designed to generate the activity signal for each point in time

a) to determine pixel-by-pixel the difference between the current three-dimensional recording and at least one previous three-dimensional recording, and b) based on the differences determined in this way, to determine an accumulated difference value, in particular by forming a maximum or averaging, and as a movement value of the activity signal (MA) to the respective Time to continue to use.

13. The arrangement according to claim 12, characterized in that the processing unit (3) is designed to between the determination of the difference between the current three-dimensional recording and at least one temporally preceding three-dimensional recording and the determination of the accumulated To make a selection of pixels of the current three-dimensional recording and discard individual pixels,

wherein the processing unit (3) is designed to select or predetermine a region of interest (ROI) in the three-dimensional image in advance, it being provided in particular that the image of body parts of the person (1) of interest is located in the region of interest (ROI).

14. Arrangement according to claim 12 or 13, characterized in that the processing unit (3) is designed to after determining the difference between the current three-dimensional recording and at least one temporally preceding three-dimensional recording or before determining the accumulated difference value pixels of the current three-dimensional Summarize the recording and reduce the image resolution.

15. Arrangement according to one of claims 10 to 14, characterized in that the processing unit (3) is designed to determine at least one of the following values as feature values assigned to a movement phase:

- a constant value,

- the duration of the respective movement phase (D; D _1; ..., D ₃ ),

BP ₃ ).

16. Arrangement according to one of claims 10 to 15, characterized in that the processing unit (3) is designed to accumulate the overall feature value (GM) by adding or averaging or maximum value formation of the individual feature values, preferably over a predetermined period of time.

17. The arrangement according to claim 16, characterized in that the processing unit (3) is designed to define a predetermined sleep phase, in particular the REM sleep phase, as a predetermined time segment.

18. Arrangement according to one of claims 10 to 17, characterized in that the

Processing unit (3) is designed to

- to form a large number of overall characteristic values (GM ^ ..., GM ₄ ),

to apply a predetermined cascade of threshold value comparisons or to feed the overall feature values (GM ^ ..., GM ₄ ) to a classification method, in particular a neural network or a support vector machine, and

19. Data carrier on which a computer program for carrying out a method for detecting the intensity of the body movements of a person (1) during their sleep according to one of claims 1 to 9 is stored.