WO2002083221A2 - Method of controlling the desired pressure of a device used for carrying out cpap therapy and device for carrying out cpap therapy - Google Patents

Abstract

The invention relates to a method of controlling the desired pressure of a device used for carrying out CPAP therapy and to a device for carrying out CPAP therapy. The inventive method comprises the following steps: repeatedly measuring a respiratory air flow during operation of the device for carrying out CPAP therapy. From the measured temporal progression of the respiratory air flow at a constant desired pressure at least one criterion such as for example a reverse correlation of respiratory cycles, the change of inhalation volumes or a bend feature during inhalation is calculated and the desired pressure is increased or reduced on the basis of this criterion.

Description

 Method for controlling the target pressure of a device for performing the CPAP therapy and a device for performing the CPAP therapy

This invention relates to a method for controlling the target pressure of a device for performing the CPAP therapy according to the preambles of claims 1, 8 or 14 and a device for performing the CPAP therapy for performing these methods.

Devices for carrying out CPAP (continuous positive airway pressure) therapy are known. The CPAP therapy is in Chest. Volume No. 110, pages 1077-1088, October 1996 and Sleep, Volume No. 19, pages 184-188. A CPAP device uses a compressor, preferably a humidifier, a hose and a nasal mask, to apply a positive pressure of up to about 30 mbar in the patient's airways. This overpressure is intended to ensure that the upper respiratory tract remains fully open throughout the night and thus no obstructive breathing disorders (apneas) occur (DE 19849 571 A1).

1 shows the CPAP device 1 and a patient 19. The CPAP device in turn comprises a compressor 4, a ventilation hose 9, a ventilation mask 18, a pressure sensor 11 and a flow sensor 16. The compressor contains a turbine 8 to generate excess pressure. In the CPAP device shown, the pressure sensor 11, which is located in the compressor housing, is connected to the respiratory mask 18 via a pressure measuring tube 10 in order to measure the actual pressure in the respiratory mask. However, the pressure sensor can also be located in the respiratory mask and can be connected to the compressor housing via electrical lines. One or more small holes 2 are made in or near the mask, so that an air flow from the compressor to the holes arises on average over time. This prevents the accumulation of CO ₂ in the ventilation tube 19 and enables the patient to breathe. The speed of the turbine 8 is regulated so that the actual pressure corresponds to a predetermined target pressure. The setpoint pressure is conventionally preset under the supervision of a doctor and is referred to as the titration pressure. The flow sensor can e.g. B. be a sensor with heating wire 17, which delivers its measurement signal via measurement line 15 to a controller in the compressor housing. With another design of the CPAP- A device can be provided with a constriction in the breathing tube for the respiratory flow measurement, the differential pressure being measured via the constriction. The pressure sensors can be arranged directly in the ventilation hose or connected to it via further pressure measurement hoses. The controller can also take over pressure control.

It was found that the patients felt the overpressure generated by the CPAP device as an unpleasant resistance against which they had to exhale. Control procedures for CPAP devices have therefore been developed that lower the set pressure as much as possible. Such a control is known from WO00 / 24446. This control is based on an algorithm in which at least three pressure values are set in succession during an “AutoSef” operation. If the tidal volume is independent of the set pressures, the pressures were too high. If the tidal volume increases with the set pressures, the pressures were too low. According to this document, the optimal pressure results from the point of intersection of the straight line through points in an area in which the tidal volume increases linearly with the pressure and a parallel to the pressure axis through points where the tidal volume is independent of the pressure.

In order to reduce the uncomfortable overpressure, BiPAP devices and multilevel devices were also developed. These devices have the property of helping the patient to breathe by lowering the pressure when exhaling and increasing the pressure when inhaling. So these devices work with at least two pressure levels.

WO 94/23780 describes a method for controlling the pressure of a CPAP device. If there are no breathing disorders during sleep, the pressure is gradually reduced. If sleep disorders such as apneas, hypopneas or snoring occur, the pressure is increased. Obstructive and central apneas are distinguished at a frequency of 5 Hz due to deliberately created pressure fluctuations.

US 5,335,654 describes a similar process. It determines the respiratory flow limitation that occurs in hypopneas based on the shape of the respiratory flow during inspiration. If the course of the respiratory flow shows a plateau or a flat area or if it deviates from a sinus shape, this is interpreted as hypopnea.

EP 0 934 723 A1 contains a similar disclosure.

DE 691 32 030 T2 also describes a control method for printing a pressure system for respiratory tracts. The pressure is raised by a valve during inhalation and reduced during exhalation. The valve is controlled so that the pressure is kept constant during inhalation and exhalation. If the valve position changes only slowly during an inhalation process, this is interpreted as the end of the inhalation process. Inaudible vibrations or changes in pressure can be evaluated to determine whether the patient's breathing is regular, irregular, or apneic. In addition, the duration of inhalation and exhalation as well as the flow velocities can be determined. This information can be stored in a memory. Finally, admittance can be calculated from respiratory flow divided by pressure. The time course of the admittance can be compared with stored admittance schemes. The number of the most appropriate admittance scheme can be used as a "pointer" for a table that contains the action to be performed, such as a pressure increase.

EP 0 612 257 B1 also describes an auto-CPAP system that detects apneas, hypopneas and unstable breathing in order to adjust the pressure. A large change in respiratory parameters is defined as a coefficient of variation value of 0.3 or more for 4 or more of the defined features such as: B. Inspiration time, total duration of a breath, mean inhalation flow, peak inhalation flow and roundness in 5 or 10 breaths. Hypoventilation is defined as five consecutive breaths with an average inspiratory flow of less than 40% of the predicted supine mean inspiratory flow when awake. Apnea is defined by the fact that there is no change in the breathing phase for 10 seconds. Flatness is defined as the relative deviation of the observed airflow from the mean airflow. to

Calculating the roundness, the squared differences between that in

Amplitude and time normalized inspiratory breath flow profile and one

Sine half-wave summed up. WO 99/24099 also describes a control method for an auto-CPAP device that takes into account apneas, hypopneas, reduced respiratory flow and snoring. The snoring signal is obtained by bandpass filtering the respiratory flow signal with a lower cutoff frequency of 30 Hz and an upper cutoff frequency of 100 Hz. A flow limitation index is calculated as an RMS deviation over the middle range of the (normalized inhaled air flow-1).

According to US Pat. No. 5,740,795, the respiratory flow signal is fed to a band-limited differentiator. If the output signal of the differentiator exceeds an inhalation threshold or falls below an exhalation threshold, an exhalation detection signal or an inhalation detection signal is determined.

WO 00/27457 relates to error diagnosis in CPAP devices. If the pressure remains below 2 mbar for more than 0.3 seconds at an engine speed of more than 12,000 revolutions per minute, an error is detected.

According to EP 0 661 071 B1, the wearing of the face mask of a CPAP device is determined on the basis of the height of the air flow.

EP 0 934 723 A1 also relates to the control of a CPAP device based on the detection of apneas and partial occlusion of the upper respiratory tract.

Fuzzy logic is also known in the prior art. According to conventional logic, logical variables can only assume the states 0 or 1 - also called "false" or "true". In fuzzy logic, fuzzy variables can have any value between 0 and 1, including 0 and 1. Fuzzy logic is mainly used for controls that should take into account the experience of experts.

According to fuzzy logic, fuzzy variables indicate membership in a set. In the case of a fuzzy control, the quantity corresponds to a specific operating state of the device to be controlled. With the aid of fuzzy logic, it is possible to design a controller taking into account a limited number of typical operating states. Fuzzy logic provides a formalism for interpolation between the states considered. The object of the invention is to specify methods for controlling the target pressure of a CPAP device and a CPAP device for carrying out the methods, which allow a CPAP target pressure that is optimal for a patient to be set on the basis of the recorded respiratory flow profile of a patient.

This object is achieved by the subject matter of the independent claims.

Preferred developments of the invention are the subject of the dependent claims.

Preferred embodiments of the invention are explained below with reference to the accompanying drawings. Show:

1 shows a device for performing CPAP therapy,

2 above 50 seconds of the respiratory flow curve of a patient in the sleep stage NREM 2, below the estimated first derivative of the respiratory flow curve after the time and in the middle the automatically detected transitions between inspiration and expiration, marked by vertical lines,

3 above shows a respiratory flow curve of a patient in the sleep stage NREM 2, in the middle the last breathing cycle from the upper respiratory flow curve, below the correlation between the section above and the breathing cycle in the middle,

4 above shows the respiratory flow curve of a patient in the sleep stage NREM 2, below the mean deviation of the correlation maxima from one,

5 above the respiratory flow curve of a patient in the sleep stage REM, below the mean deviation of the correlation maxima from one,

Fig. 6 above a patient's respiratory flow curve, in the middle the associated CPAP pressure curve, below the variance of the actual CPAP pressure per breath, and

7 shows a respiratory flow curve with the automatically detected transitions between inspiration and expiration, which are marked by vertical lines, the inspiration volume IV and the expiration time EL. The control method according to the invention first calculates features from a measured respiratory flow curve and a measured actual pressure curve of a CPAP device, which are described under section 1 "Features". Special combinations of the features are combined to form detectors, which are discussed in Section 2 "Detectors". Flags are set in the detectors when they detect an event. The control process then changes the target pressure based on the event flags of the detectors, which is explained in section 3 "Control Process". The control method has three different states, namely a normal state, a sensitive state and a leak state, between which it is possible to switch back and forth. Some detectors operate in a sensitive state with parameters that deviate from the normal state. The control process changes to the sensitive state when the control process lowers pressure in the normal state. By selecting the parameters for the sensitive state, the control process reacts faster if the actual CPAP pressure is too low. If there is a leak, the system changes to the leak state.

1 characteristics

1.1 Expiration time

The basis for the analysis of the respiratory flow is the robust detection of individual breaths. This is explained with reference to FIG. 2. 2 shows 50 seconds above the respiratory flow curve of a patient in the sleep stage NREM 2 (NREM: non rapid eye movement). The various sleep stages are described in "A manual of standardized terminology, techniques and scoring systems for sleep stages of human subjects" by Rechtschaffen, A., Kales, A. (eds.), NIH publ NO. 204, U.S. Government Printing Office, Washington D. O, 1968. In the upper graphs in FIGS. 2 to 6 and in the graph in FIG. 7, a high flow (above) indicates inspiration and a low flow (below) expiration. During the transition from inspiration to expiration, a distinct flank can be seen in the respiratory flow, which is used to detect individual breaths.

The first and second derivatives of the respiratory flow curve are estimated to detect the flanks. The first derivative is shown in Fig. 2 below. Owing to

Noise in the respiratory flow curve is not just derived from the respiratory flow curve, but also low pass filtered. The derivation and low-pass filtering takes place in one filter step by suitable selection of the coefficients of a digital filter.

The local maxima of the first derivative correspond to the maximum slope of the respiratory flow during the transition between inspiration and expiration. From the end of the inspiration, the beginning of the inspiration is sought by looking for the first local minimum in the estimated derivative.

The middle curve in FIG. 2 shows the automatically detected points in time, which are marked by vertical lines.

The expiration time results from the time difference between a minimum of the estimated derivative and the previous maximum of the estimated derivative. An expiration time is also entered with the reference symbol EL in FIG. 7.

1.2 backward correlation

An example of a 50-second section of a respiratory flow curve of a patient in the sleep stage NREM 2 can be seen in FIG. 3 above. The most recent selected breath is shown in the middle. The cross-correlation function between the data series above and the individual breathing pattern is shown in the lower curve. The correlation curve has values between one and minus one, the correlation becoming one if the two breaths match each other and equaling minus one if the curves are negatively correlated, i.e. when a peak in the breathing pattern exactly matches a valley in the data piece under consideration.

In another embodiment, instead of the correlation, another dependency measure such as e.g. B. mutual information can be used.

The correlation curve shows whether breathing is regular and whether there are no breaths. If the successive breathing patterns are similar, then the correlation curve has a periodic course with local maxima close to one and local minima close to minus one. The backward correlation is the mean value over a certain number of local maxima of the correlation curve before the current point in time. The backward correlation is a measure of how well the most recent breathing cycle matches the previous breathing cycles. The backward correlation is between 0 and 1.

The upper area of FIG. 4 shows the respiratory flow curve of a patient in stage NREM 2. The lower area of FIG. 4 shows the difference between one and the local maxima of a correlation curve, which was calculated from the respiratory flow curve shown in FIG. 4 above. The higher the value of the lower curve in FIG. 4, the more the current breathing cycle deviates from the corresponding past breathing cycle.

FIG. 5 corresponds to FIG. 4, but the respiratory flow curve now comes from the REM sleep stage (REM: rapid eye movement). The comparison of the mean correlation maxima from FIGS. 4 and 5 shows that the mean difference between the correlation maxima and one in the REM sleep stage is significantly larger.

1.3 average inspiration volume

The inspiration volume is hatched in FIG. 7 as area IV in a respiratory flow curve. The short vertical lines in FIG. 7 mark the beginning of inspiration, the long vertical lines end of inspiration. The inspiration volume can be standardized to the length of the inspiration.

1.4 mean curvature of the respiratory flow during inspiration

As a characteristic of the respiratory flow limitation, the mean curvature of the respiratory flow during inspiration is calculated. For this purpose, the first derivative of the respiratory flow during inspiration is estimated, i.e. additionally subjected to low-pass filtering. A straight line is then fitted to the estimated first derivative. The slope of this fitted straight line gives the mean curvature of the inspiration. 1.5 Zero crossings in the alternating component of the actual CPAP pressure

Snoring is caused by vibration of the walls of the upper respiratory tract. It has been shown that the number of zero crossings in the alternating component of the actual CPAP pressure is a reliable characteristic for snoring.

As described above, a typical CPAP device 1 has a pressure control loop in which the turbine speed is controlled so that the pressure in the breathing mask 18 largely corresponds to a predetermined target pressure. Some CPAP devices have faster pressure regulation, which is able to compensate for the pressure differences caused by the patient's breathing. With other CPAP devices, the actual pressure fluctuates with the patient's breathing. However, the pressure control of a typical CPAP device is not so fast that it would be able to control snoring noises.

This is illustrated in FIG. 6. A section of a respiratory flow signal is shown in FIG. 6 above, the associated CPAP actual pressure is shown below. The curve below shows the variance of the actual CPAP pressure per breath. The variance increases significantly when the actual CPAP pressure is changed by the patient snoring in the range of 30 to 40 seconds. In one embodiment, the variance is therefore used to detect snoring. In this embodiment, the actual pressure signal can be high-pass filtered before the variance is calculated in order to eliminate pressure fluctuations caused by breathing. In the preferred embodiment, however, the number of zero crossings in the alternating component of the actual CPAP pressure is used. It is the more robust feature when using CPAP devices with slower pressure control.

The zero crossings are only counted during the inspiration phase so that the control reacts to inspiratory snoring. The snoring feature is used indirectly in the control, ie if snoring is present, the respiratory flow limitation and hypopneas are taken into account more by adding a snoring bonus to the respective features.

Table 1: Characteristic types

2 detectors

Special combinations of the features become detectors. Flags are set in the detectors when they detect an event. The control method then changes the pressure on the basis of the event flags of the detectors.

2.1 Respiratory arrest detector

A long pause in breathing occurs if more than max_no_breath_time data points are analyzed without detecting a breath. If the number of long breath stops = max_number_of_no_breath_times, the automatic pressure control stops.

Table 2: Parameters for the respiratory arrest detector in the normal state and in the sensitive state (see section 3 control procedure)

The max_no_breath_time value corresponds to a time of 2 minutes at the sampling rate of 100 Hz used here.

2.2 Apnea detector

The expiration time is used for apnea detection. The apnea detector does the following: 1. Detection of respiratory arrests: there is a respiratory arrest if the expiration time is greater than exp_schwelle_in_sec.

2. Search for successive breath stops. An apnea event occurs if:

(a) either with 2 consecutive breath stops one of the breath stops lasts longer than long_apnoe_time_in_sec.

(b) or if there are more than 3 consecutive breath stops.

Breathing stops are consecutive if the duration of the intervening hyperventilation block and breathing period is <search_schwelle_in_sec.

3. Apneas can change the start of the detection of stable breathing. After n_reset_apnoen Apnoen, the detection of stable breathing begins again with the normal detector.

Table 3: Parameters for apnea detection in the normal state and in the sensitive state

2.3 Hypopnea detector

The non-normalized mean inspiration volume, the backward correlation and the snoring feature are used for hypopnea detection.

1. In the last n_insp_vol_breaths breaths, the change in the inspiration volumes of the first half of the n_insp_vol_breaths breaths compared to the second half of the n_insp_vol_breaths breaths is calculated. 2. If snoring is additionally detected in the first half of the n_insp_vol_breaths breaths, then the snoring_volume_bonus is added to the change in the inspiration volume.

3. If the change in the inspiration volume is> inspiration_volume_change_schwelle, hypopnea is present. In the n_insp_vol_breaths breaths there is stable hypopnea if the volume change is> = normal_breath_inspiration_volume_change_schwelle.

4. A hypopnea event occurs if minjumps hypopneas are found in the last n_breaths breaths.

5. A stable hypopnea event occurs when minjumps stable hypopneas are found in n_breaths breaths and the backward correlation in the n_breaths breaths is> = hypopnoe_detection_normal_correlation_schwelle.

6. Hypopneas can change the start of stable breathing detection. After n_hypopnoen_reset_normal Hypopnoen, the detection of stable breathing begins again with the normal detector.

Table 4: Parameters for the hypopnea detector in normal condition

Table 5: Parameters for the hypopnea detector in the sensitive state

2.4 Respiratory flow limitation detector

The snoring characteristic, the mean curvature and the backward correlation are used for the detection of respiratory flow limitation.

1.4.1 Weak respiratory flow limitation with stable breathing - the curvature short detector

1. Within the last curvatureShort_schwelle breaths it is checked whether the snoring characteristic is above the schnarch_min_schwelle threshold. If this is the case, the curvatureShort_snoring_bonus is added to the curvature characteristic.

2. Within the last curvatureShort_schwelle breaths are counted how often both the curvature feature exceeds the curvatureShort_mean_schwelle threshold and the backward correlation exceeds the curvatureShort_correlation_schwelle threshold.

3. If the proportion of breaths of the above criterion is met> = curvatureShort_min_above_schwelle, then there is a respiratory flow limitation event.

Table 6 Parameters for the detection of weak respiratory flow limitation with stable breathing in the normal state

Table 7: Parameters for the detection of weak respiratory flow limitation with stable breathing in a sensitive state

2.4.2 Limitation of respiratory flow with stable and unstable breathing - The Curvature-Long detector

1. Within the last curvatureLong_schwelle breaths it is checked whether the snoring characteristic is above the schnarch_min_schwelle threshold. If this is the case, the curvatureLong_snoring_bonus is added to the curvature characteristic.

2. Within the last curvatureLong_schwelle breaths are counted how often the curvature characteristic exceeds the threshold of curvatureLong_mean_schwelle. In the normal state, the threshold can be set equal to curvatureLong_medium_mean_schwelle or equal to curvatureLong_high_mean_schwelle. This will be discussed later in the

Section 3.1.2 explained in more detail. 3. If the proportion of breaths that meets the above criterion> = curvatureLong_min_above_schwelle, then there is a respiratory flow limitation event.

Table 8: Parameters for the detection of respiratory flow limitation with stable and unstable breathing in the normal state

Table 9: Parameters for the detection of respiratory flow limitation with stable and unstable breathing in a sensitive state

2.4.3 the respiratory flow limitation increase feature - the curvature after pressure down detector

The Wilcoxon rank sum test is used to detect the increase in the respiratory flow limitation feature. The Curvature After Pressure Down Detector is only used in the sensitive state. The Wilcoxon - Rank Sum Test is in the process of being hardened: Statistics, textbook and handbook of applied statistics, Oldenburg-Verlag, Munich 1999 and in D.R. Cox, C.V. Hinkley: Theoretical Statistics, Chapman & Hall 1974.

1. The rank of the curvature features is determined for the last n_curv_breaths breaths.

2. The n_curv_breaths ranks are divided into two parts, the second part being the ranks from n_curv_breaths - n_big_curv_breaths + 1 to n_curv_breaths contains. For the second part, the sum of the ranks is calculated and this sum normalized to the maximum possible sum of the ranks.

3. If this normalized value is above 0.99, the respiratory flow limitation increases and if the curvature value with the highest rank in the second part is above the curvatureDown_min_above_scheshold threshold, then there is an event for the increase in the respiratory flow limitation feature.

Table 10: Parameters for the detection of the respiratory flow limitation increase characteristic in the sensitive state

2.5 Normal detector (stable breathing)

The correlation feature is used to identify stable breathing. The normal detector does the following:

1. Averaging the correlation characteristic over all breaths within the last normal_schwelleJn_sec seconds. The parameter normal_schwelleJn_sec is adjusted via the control (see section 3).

2. There is a stable breathing event if the calculated mean correlation, ie the backward correlation> = normal correlation threshold.

Table 11: Parameters for the normal state and the sensitive state

The same parameters are used for the normal detector both in the normal state and in the sensitive state. 2.6 Leak detector

For example, a leak can already result from the mask sliding relative to the patient's face.

1. For the leak detection, the mean value of the respiratory flow and the actual CPAP pressure is calculated in a data window of the width leakagejean low ime.

2. If the calculated mean flow becomes greater than the threshold vjeakage or if the difference between the mean actual CPAP pressure and the specified CPAP target pressure lies above the threshold pressure_diff, then there is a leak event.

3. If the calculated mean flow is less than the threshold vjeakage and the difference between the mean actual CPAP pressure and the specified target CPAP pressure is less than pressure_diff, the leak is over.

4. The data window for leak detection is advanced with a step size of step_size.

Table 12: Parameters for the detection of a leak state The parameters are independent of the state of the control.

3 tax procedures

The control method has three different states, namely a normal state, a sensitive state and a leak state, between which it is possible to switch back and forth. Some detectors operate in a sensitive state with parameters that deviate from the normal state. In the sensitive state of the control, there is a change if the control is in the normal state lowers the pressure. By selecting the parameters for the sensitive state, the control reacts faster if the CPAP pressure is too low. When a leak is detected, the control switches to the leak state.

The controller changes the pressure within the limits of lower_pressureJimit and upper_pressurejimit. The printing recommendation is determined by titration_pressure. The controller uses two different pressure steps for increasing the pressure: big_pressure_step and small_pressure_step. Pressure_down_step is used to lower the pressure.

Table 13: Pressure values for the control process

3.1 normal condition

The following detectors are active in the normal state:

Respiratory arrest detector

Apnea detector

Hypopnea detector

Curvature-Long detector

Curvature-short detector

Normal detector

Leak Detector

The events of the detectors are used for the pressure change. 3.1.1 Pressure change

Respiratory failure detector: If no breath is detected for longer than max_no_breathJime, the pressure is increased by big_pressure_step.

Apnea event: If an apnea event is determined, the pressure is increased by big_pressure_step if the current pressure is less than or equal to titration_pressure plus small_pressure_step.

Hypopnea event: If a hypopnea event is detected, the pressure is increased by small_pressure_step.

Respiratory flow limitation: If the curvature long detector or the curvature short detector detects a respiratory flow limitation, the pressure is increased by small_pressure_step. The curvature after pressure down detector is not active in the normal state.

Normal event: If the apnea or hypopnea detector does not reset the start of normal breathing, the pressure is reduced by pressure_down_step when breathing is stable.

An apnea event only leads to an increase in pressure if the current pressure is <= titration_pressure + small_pressure_step. The reason for this is that there are two types of apneas. So-called obstructive apneas are caused by occlusion of the upper respiratory tract. You can be treated with a higher CPAP pressure. So-called central apneas are based on an instruction from the brain to temporarily stop breathing. They cannot be remedied by CPAP therapy. For this reason, the parameter titration_pressure is chosen so high that experience has shown that obstructive apneas no longer occur.

If the pressure is increased, the parameter normal_schwelleJn_sec to detect stable breathing is multiplied by the factor normal_scale_schwelle if the pressure is <= titration_pressure + small_pressure_step. 3.1.2 Adjusting the curvatureLong_mean_schwelle

If

a) the curvature long detector increases the pressure maxNumberOfCurvatureEvents in a row and

b) the pressure is> titration_pressure and <= titration_pressure + big_pressure_step,

then curvatureLong_mean_schwelle = curvatureLong_medium_mean_schwelle

set.

If

a) the curvature long detector increases the pressure maxNumberOfCurvatureEvents in a row and

b) the pressure is> titration_pressure + big_pressure_step

then curvatureLong_mean_schwelle = curvatureLong_high_mean_schwelle is set.

This is to prevent the pressure from being constantly increased if, despite the pressure increase, the respiratory flow limitation does not disappear.

3.1.3 Switching to another state of the control procedure

The system switches from the normal state to the sensitive state when the pressure drops and to the leak state when a leak occurs.

3.1.4 Initialization of the normal state

When changing from the sensitive state to the normal state, only the parameter normal_schwelleJn_sec is multiplied by the factor normal_scale_schwelle.

If a change is made to the normal state after a pressure increase or from the leak state, all detectors used in the normal state are switched to the Start time of the normal state is set and the parameter normal_schwelleJn_sec multiplied by the factor normal_scale_schwelle.

Table 14: Parameters

3.2 Sensitive condition

The following detectors are used in the sensitive state:

Respiratory arrest detector Apnea detector Hypopnea detector Curvature long detector • Curvature short detector

Curvature-after-Pressure-down detector

Normal detector

Leak Detector

In addition to the normal state, the curvature after pressure down detector is active in the sensitive state.

The events of the detectors are used for the pressure change.

3.2.1 Pressure change

Respiratory failure event: If no breath is detected for longer than max_no_breathJime, the pressure is increased by big_pressure_step.

Apnea event: If an apnea event is determined, the pressure is increased by big_pressure_step if the current pressure is <= titration_pressure + small_pressure_step. Hypopnea event: If a hypopnea event is detected, the pressure is increased by small_pressure_step.

Respiratory flow limitation: If the curvature long detector, the curvature short detector or the curvature after pressure down detector detects a respiratory flow limitation, the pressure is increased by small_pressure_step.

Normal event: If the apnea or hypopnea detector does not reset the start of normal breathing, the pressure is reduced by pressure_down_step when breathing is stable.

3.2.2 Switching to another state

It will switch back to the sensitive state if

a) the pressure is reduced, or

b) the pressure is increased and if the pressure is less than the pressure before the last pressure drop and if the number of breaths since the last pressure drop is <n_breathsJor_restart_APDC_at_pressure_up.

It will go back to normal when

a) the pressure is not reduced and the pressure is increased for the previous reasons, or

b) the number of breaths since the last pressure drop is> max_events_after_pressure_down.

3.2.3 Initialization of the sensitive state

If the CPAP pressure is increased during the sensitive state and not switched to the normal state, then the parameter normal_schwelle_in_sec for the normal detector is multiplied by normal_scale_schwelle.

All detectors are set to the start time of the sensitive control. 3.3 Leak condition

1. If a leak is detected during the normal state or the sensitive state, the system switches to the leak state. During the leak state, only the leak detector is active, the remaining detectors continue to work, but cannot change the pressure.

2. When the leak has ended, the system switches back to the normal state. If the leak duration is longer than leakage_reset_pressure ime, the CPAP setpoint pressure is decreased by leakage_pressure_down_step and the system switches to the normal state.

3. When the leak is over and the system has returned to normal, the leak detector waits for leakage_restartJlowJime seconds until it starts again with the leak detection.

Table 15: Parameters for the leak state

4 fuzzy logic

In a further embodiment, the events determined by the detectors are treated as fuzzy variables. The advantage of this embodiment is that the control works more continuously. The transition from "no event" to "event occurred" preferably takes place, that is to say the region in which the fuzzy variable increases from 0 to 1, so that the corresponding fuzzy variable reaches the value 0.5 at the limit value specified above. Taking into account the gradual transition from 0 to 1 of fuzzy variables, one can therefore formulate that, for example, a normal event is detected or recognized the more clearly the backward correlation exceeds a threshold value (see section 2.5). The width of the selected transition area and the course of the transition function are of minor importance for the quality of the tax procedure. For example, the normal fuzzy variable can take the value zero if the backward correlation is less than 0.82, increase linearly from 0 to 1 if the backward correlation falls in the range between 0.82 and 0.9 and be 1 if the backward correlation exceeds 0.9. However, other functions such as a suitably scaled arctan function or a probability integral Φ (x) can also be used to design the transition area:

When using fuzzy variables, the pressure change is determined from the sum of the fuzzy variables supplied by the individual detectors and preferably weighted with coefficients. The coefficients take into account the fact that, for example, the pressure is increased rapidly when a breathing arrest is detected, while the target pressure of the CPAP device is increased more slowly when the breathing flow is limited. As a result, for example, the coefficient for the apnea fuzzy variable will be greater than that for the respiratory flow limitation fuzzy variable.

The normal state and the sensitive state of the control process can also be regarded as fuzzy variables. In such an embodiment, the respiratory flow limitation feature calculated in section 2.4.3 multiplied by the fuzzy variable for the sensitive control is incorporated into the pressure control.

In a further embodiment, only some of the detectors can deliver fuzzy variables as results. The use of fuzzy variables is particularly advantageous in the case of the hypopnea detector, the curvature long detector, the curvature short detector and the curvature after pressure down detector.

The use of fuzzy variables does not seem to make much sense in the leak state, since the leak state is not an orderly operation of the CPAP device, but an exceptional state. In a further preferred embodiment, intermediate results that occur in the detection of events in the detectors are also implemented as fuzzy variables. This applies in particular to the decision whether one of two breath stops lasts longer than long_apnoeJimeJn_sec, whether three breath stops are consecutive (section 2.2), whether the snoring feature is fulfilled, whether both the curvature feature and the correlation feature exceed a corresponding threshold (section 2.4.1 ) and whether the curvature characteristic exceeds CurvatureLong_mean_schwelle (section 2.4.2).

The target pressure control method according to the invention described above can also be used in BiPAP devices and in multilevel devices. The setpoint pressure determined according to the control method can be used as the higher pressure in BiPAP devices or the highest pressure in multilevel devices. In another embodiment, the pressure determined according to a control method according to the invention specifies the time average of the pressures generated by a BiPAP or multilevel device.

The invention was previously explained in more detail using preferred embodiments. However, it will be apparent to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. Therefore, the scope of protection is determined by the following claims and their equivalents.

LIST OF REFERENCE NUMBERS

1 CPAP device 15 electrical measuring line 2 hole 16 flow sensor

4 compressor 17 heating wire

8 turbine 18 respiratory mask

9 Respiratory tube 15 19 sleeping

10 Pressure measuring tube IV inspiration volume 11 Pressure sensor EL expiration time

Claims

claims: 1. A method of controlling the target pressure of a device to perform CPAP therapy using: repeated measurement of a breath of air during operation of the device, Determining at least one criterion from the measured time profile of the breathing air flow, on the basis of which the target pressure is increased or decreased, marked by Calculating a backward correlation of breaths during a predetermined time before the current breath, and the more likely the backward correlation exceeds a predetermined threshold value, the more likely it is to lower the target pressure. 2. The method according to claim 1, characterized in that the more clearly the backward correlation exceeds a predetermined threshold value, the more a normal breathing event is recognized. 3. The method according to claim 2, characterized in that the target pressure is lowered by a first predetermined pressure difference when a normal breathing event is detected. 4. The method according to any one of the above claims, characterized in that individual breaths are distinguished by analyzing the falling or rising edges of the temporal course of the breathing air flow. 5. The method according to claim 4, characterized in that from the analysis of the falling or rising flanks expiration times for several successive breaths are determined and the more an apnea event is recognized, the clearer the expiration time of two successive breaths is a first expiration threshold time exceed or the more clearly the expiration times of three successive breaths exceed a second expiration threshold time. 6. The method according to claim 5, characterized in that the target pressure is increased by a second predetermined pressure difference when an apnea event is detected and the target pressure before the increase is below a titration pressure. 7. The method according to any one of claims 4 to 6, characterized in that the target pressure is increased by a third predetermined pressure difference if no breath is detected within a predetermined time of preferably two minutes, but the target pressure is no longer changed when the target pressure a predetermined number of times has been increased based on this criterion. 8. A method for controlling the target pressure of a device for performing the CPAP therapy, in particular according to one of the above claims, with: repeated measurement of a breath of air during operation of the device, Determining at least one criterion from the measured time profile of the breathing air flow, on the basis of which a target pressure is increased or decreased, marked by In a first predetermined number of breaths, calculate the change in the inspiration volumes of the first half of these breaths compared to the second half, and the more likely the change exceeds a predetermined change threshold, the more likely it is to raise the target pressure. 9. The method according to claim 8, characterized in that if snoring is detected in the first half of the breaths, the change in the inspiration volumes is increased by a first snoring bonus before it is compared with the change threshold value. 10. The method according to claim 8 or 9, characterized in that the more hypopnea is present, the more clearly the change in the inspiration volumes lies above the change threshold. 11. The method according to claim 10, characterized in that the more a hypopnea event is determined, the more clearly the number of 10 hypopneas exceeds a predetermined number of hypopneas in a second predetermined number of breaths. 12. The method according to claim 11, characterized in that the target pressure is increased by a fourth predetermined pressure difference when a hypopnea event is determined. 13. The method according to claim 11 or 12, insofar as it relates indirectly or directly to claim 1, characterized in that when a hypopnea event occurs, the predetermined time during which the backward correlation is calculated starts again from the hypopnea event starts running. 20 14. A method for controlling the target pressure of a device for performing the CPAP therapy in particular according to one of the above claims with: repeated measurement of a breathing air flow during operation of the device at several points in time, Determining a curvature characteristic from the measured temporal course of the respiratory air flow, on the basis of which a desired pressure is increased or decreased, marked by Estimating the first time derivative of the respiratory flow during inspiration at several points in time using the measuring points of the respiratory flow, Fitting a straight line to the estimated derivative values, the slope of the straight line representing the curvature feature, and the greater the curvature characteristic, the more likely it is to raise the target pressure. 15. The method according to claim 14, insofar as it relates to claim 1, characterized in that within a third predetermined number of breaths the proportion of the breaths is determined during which both the curvature feature a first Curvature feature threshold value and also a backward correlation exceeds a first correlation threshold value, and that the more a breath flow limitation event is detected, the more clearly the proportion of the breaths exceeds a first proportion threshold value. 16. The method according to claim 14 or 15, characterized in that the curvature feature is increased by a second snoring bonus before it is compared with the curvature feature threshold value if snoring is detected during the third predetermined number of breaths. 17. The method according to any one of claims 14 to 16, characterized in that within a fourth predetermined number of breaths a curvature feature is determined for each breath, the fraction of breaths being determined during which the curvature feature exceeds a second curvature feature threshold value and that all the more a respiratory flow limitation event occurs, the more clearly the proportion of the breaths exceeds a second proportion threshold. 18. The method according to claim 17, characterized in that the curvature feature is increased by a third snoring bonus before it is compared with the second curvature feature threshold value if snoring is detected during the fourth predetermined number of breaths. 19. The method according to any one of claims 14 to 18, characterized in that 5 for a fifth predetermined number of breaths the more likely Respiratory flow limitation event is detected, the more significant is the increase in a curvature feature. 20. The method according to claim 19, characterized in that, the significant increase in the curvature feature with the Wilcoxon 10 rank sum test is determined, whereby the ranks are divided into two groups, whereby the sum of the ranks is calculated for the higher group and normalized to the maximum possible sum, and the more a normalized value is above a normal threshold value, the more a respiratory flow limitation event is detected and the clearer the curvature value 15 with the highest rank in the second group is above a third curvature feature threshold. 21. The method according to any one of claims 14 to 20, characterized in that the target pressure is increased by a fifth predetermined pressure difference when a respiratory flow limitation event is detected. 22. The method according to one of the above claims, characterized in that snoring is detected the more, the more the number of zero crossings in the alternating component of the measured CPAP actual pressure exceeds a predetermined threshold value. 23. The method according to any one of the above claims, characterized in that the method has a first and a second state, a part of the predetermined threshold values having different values in both states. 24. The method according to any one of claims 2, 3, 5, 6, 10, 11, 12, 13, 15 and 16 to 23, insofar as claims 16 to 22 relate to claim 15, characterized in that the pressure change is calculated as the sum of fuzzy variables weighted with coefficients, the fuzzy variables indicating to what extent the normal breathing event, the apnea event, the hypopnea, the hypopnea event and / or the respiratory flow limitation breathing event are present. 25. Device for carrying out the CPAP therapy with a command memory and a central processing unit which processes commands stored in the command memory, so that the device for carrying out the CPAP therapy carries out a method according to one of claims 1 to 24.