System equipped with sensing unit for detecting sleep state of subject

Abstract

To provide a device, system, and method for detecting a sleep disorder that may occur in a subject.SOLUTION: A sensing unit includes: an accelerometer which is adapted to measure motions of the head and / or mandible of a subject; and a gyroscope which is adapted to measure the motions of the mandible of the subject, and transmits data of the rotational motion measured from the gyroscope and data of the acceleration measured from an accelerometer to a data analysis unit. The data analysis unit derives multiple measured rotation values and acceleration values from the sampled data of the rotational motion and acceleration, collates the measured rotation values and acceleration values with one or more of N mandibular motion classes indicative of a sleep event, generates diagnostic information for diagnosing a sleep disorder related to the sleep event on the basis of one or more of the collated N mandibular motion classes indicative of the sleep event, and creates a report containing the diagnostic information for diagnosing a sleep disorder related to the detected sleep event.SELECTED DRAWING: None

Description

[Technical field]

[0001] The present invention relates to an apparatus, system and method for detecting possible sleep disorders in a subject. [Background technology]

[0002] The most common method for evaluating sleep disorders, and more specifically sleep-disordered breathing, is laboratory polysomnography (PSG). The test requires an overnight stay in a specialized facility under the supervision of trained technicians. However, the method is expensive and time-consuming and cannot keep up with demand. In a PSG test, multiple physiological signals (EEG, EMG, ECG, thermistor, pressure, video, etc.) are recorded using various types of sensors. A medical professional then reviews the data obtained from these sensors.

[0003] Other systems have been considered in the art. US 2017 / 0265801 relates to a bruxism detection system for detecting bruxism and chewing. The system includes an accelerometer attached to the jaw. It senses and records the change in acceleration at the beginning and end of jaw clenching. Data from the accelerometer is processed and compared to an accelerometer threshold to distinguish bruxism-related movements from other movements of the head.

[0004] US 2017 / 0035350 also relates to a system for detecting bruxism. The system includes two accelerometers attached to the masseter muscles. One accelerometer is attached to the skin of the left masseter muscle and the second accelerometer is attached to the skin of the right masseter muscle. Bruxism is detected when the recorded data of the two accelerometers are substantially equal.

[0005] US 2007 / 273366 relates to a sleep disorder detector system. The system includes a distance measuring device by detecting emitted magnetic fields. The device can be attached to a support placed on the head to measure mouth movements. Data from the device is processed to detect sleep-related respiratory disorders such as snoring.

[0006] A problem with these known systems is that the head movements and the mandible movements of the subject wearing the sensing unit are considered separately from each other. Similar problems apply to the head and mandible positions calculated from the movements measured by the accelerometer. However, the data from the accelerometer has limitations and can be affected by the movements of other body parts such as the chest and trachea during breathing. Thus, the association between these various movements and positions is not sufficiently taken into account for the analysis of sleep disorders, which can negatively affect the diagnosis based on the measured data stream.

[0007] In fact, mandibular movement can be induced by either respiratory or non-respiratory movements. Thus, head movements when a person is asleep can cause mandibular movement. The mandibular can be considered both as a mechanical association with tracheal traction or as an effect of control by the brain. Thus, mandibular movement is either passively induced by the respiratory movement of tracheal traction or directly controlled by the brain. The traction is the traction force exerted by the chest on the human head. Since the respiratory muscles are controlled by the brain and the chest moves with breathing, the traction force is at the human respiratory frequency. Thus, if the head moves at the respiratory frequency, the mandible, which is attached to the head, will follow the movement imposed on the head at the respiratory frequency. This is a passive movement that follows the movement of the head. Even if the head does not move, and it does not move most of the time, the mandibular movement is likewise directly and actively controlled by the brain. The brain controls the mandibular movement by stimulating the mandibular muscles. Therefore, it would be useful to be able to distinguish between mandibular movements controlled by the brain and those controlled by tracheal traction attachment. Systems are needed that can more accurately interpret signals from the brain and more accurately identify sleep disorders. Summary of the Invention

[0008] It is an object of the present invention to provide a system of sensing units and a data processing device for temporally correlating measurements of the movement and position of the subject's head and mandible in an analysis of the measurement data.

[0009] In particular, the present invention relates to a system including a sensing unit (equivalently, in combination) and a processing unit for processing the data related to possible sleep disorders of a subject, wherein the sensing unit includes a gyroscope adapted to measure the movement of the mandible of the subject. Surprisingly, the inventors have found that the gyroscope can be used to capture the rotation of the mandible, thereby evaluating the activity of the brainstem that controls the movement of the mandible during sleep.

[0010] In some embodiments, the present invention relates to a system including a sensing unit and a data processing device, e.g. a processing unit, related to a sleep disorder that may occur in a subject, where the sensing unit includes an accelerometer adapted to measure the movement of the head and / or mandible of the subject, and a gyroscope adapted to measure the movement of the mandible of the subject. The sensing unit is adapted to generate measurement signals based on measurements made. The processing unit includes first and second inputs for receiving first and second time streams of measurement signals from the accelerometer and gyroscope, respectively.

[0011] Thus, provided herein is a system for characterizing sleep disorders in a subject having a head and a mandible, comprising a gyroscope and a data analysis unit connected by a data link. In a particular embodiment, the system is characterized in that it comprises: - a gyroscope configured to measure the rotational movement of the subject's mandible; - a data analysis unit and a data link, the data link being configured for transmitting the rotational movement data measured from the gyroscope to the data analysis unit; Wherein the data analysis unit includes a memory unit configured to store N mandibular movement classes, where N is an integer greater than 1, and at least one of the N mandibular movement classes indicates an onset of a sleep disorder; - where, for each jth (1< j < N) the mandibular movement classes include a j-th set of rotation values, each j-th set of rotation values ​​indicative of at least one velocity, velocity change, frequency, and / or amplitude of mandibular rotation associated with the j-th class; - where the data analysis unit includes a sampling element configured to sample data of the measured rotational movement during a sampling period, thereby obtaining sampled rotational movement data; - where the data analysis unit is configured to derive a plurality of measured rotation values ​​from the sampled rotational movement data; and wherein said data analysis unit is further configured for matching said measured rotation values ​​with said N mandibular movement classes.

[0012] In some embodiments, the system comprises an accelerometer adapted to measure acceleration, the acceleration being indicative of movement and / or position of the subject's head and / or mandible. - the data link is further configured for transmitting measured acceleration data from the accelerometer to the data analysis unit; - where, for each jth (1 < j < N) a mandible movement class includes a j-th set of acceleration values, each j-th set of acceleration values ​​indicative of at least one movement of the mandible or head associated with the j-th class; - wherein the sampling element is configured to sample measured acceleration data during a sampling period, thereby obtaining sampled acceleration data; - wherein the data analysis unit is configured to derive a plurality of measured acceleration values ​​from the sampled acceleration data; and wherein said data analysis unit is further configured for matching said measured acceleration values ​​with said N mandibular movement classes.

[0013] In some embodiments, the system further comprises a magnetometer adapted to measure magnetic field data, variations in which are indicative of movement of the subject's head and / or mandible. - the data link is further configured for transmitting measured magnetic field data from the accelerometer to the data analysis unit; - where, for each jth (1 < j < N) a mandibular movement class includes a j-th set of magnetic field data values, each j-th set of magnetic field data values ​​indicative of at least one velocity or velocity change of mandibular or head movement associated with the j-th class; - where the data analysis unit includes a sampling element configured to sample measured magnetic field data during a sampling period, thereby obtaining sampled magnetic field data; - where the data analysis unit is configured to derive a plurality of measured magnetic field values ​​from the sampled magnetic field data; and wherein said data analysis unit is further configured for matching said measured magnetic field values ​​with said N mandibular movement classes.

[0014] In some embodiments the gyroscope and optionally the accelerometer and / or magnetometer or parts thereof are comprised in a sensing unit, which is attachable to the subject's mandible.

[0015] In some embodiments one or more of said N mandibular movement classes are characterized by a predetermined frequency range.

[0016] In some embodiments, the analysis unit is configured for identifying head movements of the subject based on the gyroscope, accelerometer, and / or magnetometer data.

[0017] In some embodiments, at least one of the N mandibular movement classes indicates that the subject is awake and a plurality of the N mandibular movement classes indicates that the subject is asleep.

[0018] In some embodiments, at least one of the N mandibular movement classes indicates that the subject is in an N1 sleep state and at least one of the N mandibular movement classes indicates that the subject is in a REM sleep state, and optionally, at least one of the N mandibular movement classes indicates that the subject is in an N2 sleep state and / or at least one of the N mandibular movement classes indicates that the subject is in an N3 sleep state.

[0019] In some embodiments, one or more of the N mandibular movement classes are indicative of obstructive apnea, obstructive hypopnea, respiratory effort related arousals, central apnea, and / or central hypopnea.

[0020] In some embodiments, one of the N mandibular movement classes is indicative of bruxism, wherein the measured rotational movement data exhibits mandibular movement amplitudes of at least 1 mm, or, if the movement is phasic, mandibular movement amplitudes of greater than 1 mm at an established frequency in the range of 0.5-5 Hz during at least three respiratory cycles, or sustained and tetanic for at least 2 seconds.

[0021] Further provided is a method for assisting in characterizing sleep disorders in a subject having a mandible, comprising the steps of: - receiving, at a data analysis unit, rotational movement data from a gyroscope placed on the subject's mandible via a data link; - storing, in a memory unit included in the data analysis unit, N mandibular movement classes, where N is an integer greater than 1, and at least one of the N mandibular movement classes is indicative of an onset of a sleep disorder; - where, for each jth (1 < j <N) the mandibular movement classes include a j-th set of rotation values, each j-th set of rotation values ​​indicating at least one velocity, velocity change, frequency, or amplitude of mandibular rotation associated with the j-th class; - sampling data of the rotational movement during a sampling period using a sampling element included in the data analysis unit, thereby obtaining sampled data of the rotational movement; - deriving a plurality of measured rotation values ​​from the sampled rotational movement data using the data analysis unit; and - matching said measured rotation values ​​with said data analysis unit to said N mandibular movement classes.

[0022] In some embodiments, the method further comprises the steps of: - measuring acceleration using an accelerometer, said acceleration being indicative of movement and / or position of the subject's head and / or mandible; - transmitting measured acceleration data from the accelerometer to the data analysis unit using the data link; - where, for each jth (1 < j < N) a mandible movement class includes a j-th set of acceleration values, each j-th set of acceleration values ​​indicative of at least one movement of the mandible or head associated with the j-th class; - sampling the measured acceleration data during a sampling period using a sampling element, thereby obtaining sampled acceleration data; - deriving a plurality of measured acceleration values ​​from the sampled acceleration data using the data analysis unit; and - matching the measured acceleration values ​​with the N mandibular movement classes using the data analysis unit.

[0023] In some embodiments, the method further comprises the steps of: - measuring magnetic field data using a magnetometer, variations in the magnetic field data being indicative of movement of the subject's head and / or mandible; - transmitting the measured magnetic field data from the magnetometer to the data analysis unit using the data link; - Here, the jth (1 < j < N) a mandibular movement class includes a j-th set of magnetic field data values; each j-th set of magnetic field data values ​​indicative of at least one velocity or velocity change of mandibular or head movement associated with the j-th class; - sampling the measured magnetic field data during a sampling period using a sampling element included in the data analysis unit, thereby obtaining sampled magnetic field data; - deriving a plurality of measured magnetic field values ​​from the sampled magnetic field data using the data analysis unit; and - matching said measured magnetic field values ​​with said data analysis unit to said N mandibular movement classes.

[0024] In some embodiments, the method further comprises using the analysis unit to identify head movements of the subject based on the gyroscope, accelerometer, and / or magnetometer data.

[0025] In some embodiments, at least one of said N mandibular movement classes is indicative of bruxism, wherein said measured rotational movement data exhibits mandibular movement amplitudes of at least 1 mm, or, if said movement is phasic, mandibular movement amplitudes of greater than 1 mm at an established frequency in the range of 0.5-5 Hz during at least three respiratory cycles, or sustained and tetanic for at least 2 seconds. [Brief description of the drawings]

[0026] The invention will now be described in more detail with the aid of figures illustrating the system and its operation. The system can be described as a system of sensing units and a device or unit for processing the sensed data. The figures are as follows:

[0027] [Figure 1] A system according to the present invention.

[0028] [Figure 2A-2B] Two streams while a person's head changes position while lying in bed.

[0029] [Figure 3A-3B] Stream captured by sensing unit during teeth grinding.

[0030] [Figure 4] Loop gain.

[0031] [Diagram 5] Identifying microarousals after preprocessing.

[0032] [Figure 6] The measured signal after applying bandpass filtering.

[0033] [Figure 7] A signal indicating micro-arousal.

[0034] [Figure 8] 1 shows an example of first and second measurement signal streams during obstructive apnea.

[0035] [Figure 9] 1 shows an example of first and second measurement signal streams during obstructive hypopnea.

[0036] [Figure 10] 1 shows an example of first and second measurement signal streams in a mixed apnea.

[0037] [Figure 11] 1 shows an example of first and second measurement signal streams during central apnea.

[0038] [Figure 12] 13 shows an example of first and second measurement signal streams during central hypopnea.

[0039] [Figure 13] 1 is an example of the first and third measurement signal streams in respiratory effort related arousal (RERA).

[0040] [Figure 14] Spectrogram of frequency distribution of mandibular movement.

[0041] [Figure 15] An example of the procedure for feature extraction, data processing, and data description.

[0042] [Figure 16-17] Analysis of mandibular movement data acquired using a magnetic sensor.

[0043] [Figure 18] An example of a method for automatic sleep stage detection from mandibular movement data acquired using a gyroscope and an accelerometer, which is described in detail in Example 18.

[0044] In Figure 1 the following numbering is used: 1: sensing unit; 2: accelerometer; 3: gyroscope; 4: magnetometer; 5: oximeter; 6: thermometer; 7: audio sensor; 8: electromyography unit; 9: photoplethysmography; 10: data processing unit; 11-1: first input; 11-2: second input; 11-3: third input; 11-4: fourth input; 12: identification unit; 13: analysis unit. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

[0045] Before describing the systems and processes of the present invention, it is to be understood that it is not limited to the particular systems and methods or combinations described, as such systems and methods or combinations may, of course, vary, and it is also to be understood that the terminology used herein is not intended to be limiting, since the scope of the terms will be limited only by the appended claims.

[0046] As used herein, the singular forms "a," "an," and "the" include both singular and plural references unless the context clearly dictates otherwise.

[0047] The terms "comprising", "including" and "consisting of" as used herein are synonymous with "including", "including", "containing" or "containing" and are inclusive or open-ended and do not exclude additional, unrecited members, elements or method steps. The terms "comprising", "comprising" and "consisting of" as used herein will be understood to include the terms "consisting of" and "consisting of".

[0048] The recitation of numerical ranges by endpoints includes all values ​​and fractions subsumed within the respective ranges, as well as the recited endpoints.

[0049] The terms "about" or "approximately" as used herein when referring to a measurable value such as a parameter, amount, duration, etc., are meant to encompass such variations from the specified value as appropriate for practice in the disclosed aspects and embodiments, and within a range of + / - 10% or less, preferably + / - 5% or less, more preferably + / - 1% or less, and even more preferably + / - 0.1% or less of the specified value. It is to be understood that the value to which the modifier "about" or "approximately" refers is itself specifically and preferably disclosed.

[0050] The terms "one or more" or "at least one," such as one or more, or at least one member of a group, are themselves clear and, by way of further illustration, said terms include reference to, among other things, any one of the members, or any two or more of the members, for example, any three or more, four or more, five or more, six or more, seven or more, etc., of the members, up to all of the members.

[0051] All references cited herein are incorporated herein in their entirety by reference. In particular, the teachings of all references specifically mentioned herein are incorporated herein by reference. Unless otherwise defined, all terms used herein, including technical and scientific terms, have the meanings that are commonly understood by those skilled in the art. With further guidance, as described herein, definitions of terms are included to better understand the present teachings.

[0052] In the following sections, various aspects are defined in more detail. Each aspect so defined may be combined with one or more other aspects, unless expressly stated to the contrary. In particular, any feature indicated as being preferred, particular or advantageous may be combined with any other feature indicated as being preferred, particular or advantageous.

[0053] The present invention relates to the measurement and evaluation of mandibular movement in subjects during sleep. The mandible or mandible is located under the upper jaw and forms the lower jaw. It is the only mobile bone of the human skull (except for the ossicles of the middle ear). During movement, the mandible rotates around the temporomandibular joint, where it connects to the skull in front of the ear (temporal bone). During mandibular movement, the length-tension relationship of muscle fibers anchored to the mandible changes. This can lead to stiffening of the upper airway in subjects at risk of instability during sleep. This movement is activated under agonist / antagonist muscles to raise and lower the mandible, thereby opening and closing the mouth, respectively. The agonist / antagonist muscles are innervated by motor nerves originating from the nucleus of the trigeminal nerve in the brainstem (pontine center) and are supported by the motor branches of said nerves.

[0054] Provided herein is a system for characterizing sleep disorders in a subject having a head and a mandible. The system includes a gyroscope. The gyroscope is configured to measure the rotational movement of the subject's mandible. This is an activity for which the gyroscope is particularly suited, as observed by the inventors. The gyroscope can be used to evaluate the activity of the brainstem that stimulates mandibular movement during sleep in a way that keeps the upper airway (pharynx) open and prevents sleep apnea. The mandibular movable bone rotates like a lever to stretch pharyngeal muscle fibers that are attached to the arch of the mandible directly or indirectly (through the second movable bone, the hyoid bone), including the tongue.

[0055] The gyroscope movement represents, to some extent, a central drive. This means that the trigeminal nuclei in the pons act to finely displace the mandible under the influence of higher centers involved in the sleep mechanism (stages of sleep) with respect to the respiratory center also located in the brainstem. As a result, the provision of a gyroscope in the sensing unit can be used to evaluate various sleep-related activities by examining the rotational mandible displacement, which may include breathing, sleep stages, or other events (e.g., movement or movement onset). Also, in addition to metrics derived directly or indirectly from the measurements, values ​​measured by a gyroscope positioned to measure the rotational movement of the subject's mandible (such as the speed and amplitude of the mandible gyroscope signal) can be used to obtain an assessment of the central drive originating from the trigeminal nuclei.

[0056] The inventors have found that other sensing units are not suitable for the measurement and evaluation of mandibular movements provided herein. For example, inertial sensors such as accelerometers are not suitable for measuring rotational displacement of the mandible because they are limited in measuring linear acceleration. Accelerometer measurements can be affected by body and head movements, such as chest and trachea during breathing, making it difficult to distinguish the origin of the data, adding unnecessary noise and complexity to the system. As a result, existing systems for analyzing sleep disorders do not adequately take into account the association between possible body and head movements. This negatively impacts diagnosis based on the measured data stream. The inventors have found that mandibular rotation carries the necessary information for accurate evaluation, and furthermore, such movements can be accurately recorded by gyroscopes.

[0057] The system further comprises a data analysis unit and a data link, which provides a communication path between the gyroscope and the data analysis unit, and preferably the data link is a wireless data link, although a data link using wired communication is certainly possible, e.g. to improve the subject's comfort.

[0058] Rotational movement data is transmitted from the gyroscope to the data analysis unit via the data link, which is of a general purpose nature and includes provisions for wireless or wired transfer of data.

[0059] The data analysis unit comprises a memory unit, e.g., a data storage device such as a hard drive, solid state drive, memory card, etc. The memory unit is configured to store a number (N) of mandibular movement specific patterns (classes), where N is an integer greater than 1. At least one of the N mandibular movement classes is indicative of the onset of a sleep disorder. Preferably, the N mandibular movement classes include a plurality of movement classes indicative of various mandibular movements. Each jth (1 < j <N) The mandibular movement classes are comprised of a j-th set of rotation values, each j-th set of rotation values ​​indicating at least one velocity, velocity change, frequency, and / or amplitude of mandibular rotation associated with the j-th class.

[0060] The rotational movement data measured or recorded by the gyroscope is linked to the mandibular movement class as follows:

[0061] The data analysis unit includes a sampling element configured to sample data of the measured rotational movement during a sampling period. Thus, sampled data of the rotational movement is obtained. Thus, information contained in the signal recorded by the gyroscope can be extracted for further analysis. It should be understood that, although the provision of specialized hardware is certainly possible, in some embodiments, the data analysis unit can be included in a general-purpose computing device such as a personal computer or a smartphone.

[0062] The data analysis unit is configured to derive a number of measured rotation values ​​from the sampled rotation movement data and to match the measured rotation values ​​with N mandibular movement classes. Preferably, deriving the measured rotation values ​​from the sampled rotation movement data includes one or more of the following procedures: discretization, time averaging, fast Fourier transformation, etc. Also, the matching can be fully or partially automated by providing a machine learning model. Thereby, the data analysis unit is configured to learn some statistical and / or physical metrics to capture the characteristics of the signal in frequency and time domain and to identify patterns of the rotation signal for specific events such as sleep stages or respiratory effort. Thus, providing a machine learning model can provide an automatic interpretation of the relevant information and / or matching of the characteristic data with sleep disorder onset. Thus, the test on mandibular movement during sleep provides information on the respiratory control state according to changes in the airflow permeability or resistance of the upper airway, whether or not related to a sequence of changes in head position. Analysis of the nature of mandibular movements using the system according to the invention can also detect the occurrence of repetitive non-respiratory movements during sleep, such as teeth grinding or chewing, or other features such as orofacial movement disorders. Swallowing and sucking movements in infants can also be clearly identified. Swallowing movements can also be detected in adults. This allows differentiation between arousals and micro-arousals.

[0063] In some embodiments, one or more mandibular movement classes exhibit additional major mandibular movement (IMM), which is related to microarousals or respiratory disorder induced arousals, and thus can be effectively inferred from said measurements and analysis.

[0064] In some embodiments, the process of matching the measured rotation values ​​to the N mandibular movement classes utilizes artificial intelligence methods, such as random forests.

[0065] In some embodiments, the system further comprises an accelerometer. The accelerometer is adapted to measure acceleration (including acceleration fluctuations) indicative of the movement and / or position of the subject's head and / or mandible. The inventors have found that an accelerometer is particularly suitable for measuring head movement and position. Adding an accelerometer to the system allows further evaluation of the mandible's behavior during sleep. In particular, the inventors have found that acceleration can be measured to account for unexpected changes in gyroscope movement, amplitude, and / or velocity. Thus, accelerometer measurements can be used to supplement the gyroscope measurements.

[0066] The measured or recorded acceleration data is transmitted by the accelerometer via a data link to a data analysis unit. < j < N) mandibular movement classes are composed of j-th set of acceleration values. Each j-th set of acceleration values ​​or metrics indicates at least one mandibular movement or head movement associated with the j-th class. The sampling element is configured to sample the measured acceleration data during a sampling period. After sampling, the measured acceleration data is called sampled acceleration data. The data analysis unit is configured to derive a plurality of measured acceleration values ​​from the sampled acceleration data, for example by discretization and optionally time averaging. Information contained in the signal recorded by the accelerometer may be extracted for further analysis. The data analysis unit is further configured to match the measured acceleration values ​​with N mandibular movement classes. It is understood that the matching process includes automatically determining the mandibular movement class closest to the measured acceleration value. The matching can be fully or partially automated with a machine learning model. This allows automatic interpretation of the matching of relevant information and / or characteristic data with sleep disorder onset.

[0067] The inventors have found that the accelerometer is particularly sensitive to head movements. The gyroscope and accelerometer also allow for efficient discrimination of head movements from mandibular movements, thereby improving the detection of sleep disorder onset. As a result, the provision of a gyroscope and accelerometer in a single system can improve the sensitivity and accuracy of the system, and can also be used to evaluate new information that could not be interpreted from measurements by the gyroscope or accelerometer alone. For example, changes in head position stimulated by central activation can affect the rotational movement of the mandible, which can be misinterpreted as a change in the degree of opening and closing of the mouth. Thus, the combination of gyroscope and accelerometer allows for discerning head movements from jaw movements. Considering the superior and unexpected functionality provided by the combination, the presence of a gyroscope cannot be considered as a substitute for other sensing devices, such as a second accelerometer.

[0068] In some embodiments, the system further comprises a magnetometer adapted to measure magnetic field data. Variations in the magnetic field data indicate the direction and / or position of the subject's head and / or mandible movement. Adding a magnetometer to the system allows further evaluation of the mandible's behavior during sleep. As can be understood, providing a magnetometer to the system serves to evaluate the orientation of the sensing unit in a manner similar to a compass. Thus, the magnetometer is not intended to function as a unit for measuring distance as contemplated in the art, although it is understood that its main function is not limited to the scope of the system.

[0069] The data link is further configured to transmit measured or recorded magnetic field data from the magnetometer to the data analysis unit. < j <N) mandibular movement classes include a j-th set of magnetic field data values. Each j-th set of magnetic field data values ​​indicates at least one speed or speed change of mandibular movement or head movement associated with the j-th class. The data analysis unit includes a sampling element configured to sample the measured magnetic field data during a sampling period. Thus, sampled magnetic field data is obtained. The data analysis unit is configured to derive a plurality of measured magnetic field values ​​from the sampled magnetic field data. The data analysis unit is further configured to match the measured magnetic field values ​​with the N mandibular movement classes.

[0070] In a particular form, the magnetometer may include two portions: one portion is attached to the patient's forehead and another portion is attached to the patient's mandible, which the inventors have found to be a particularly effective configuration for detecting mandible movement.

[0071] In some embodiments, signals originating from the magnetometer, gyroscope, accelerometer and / or further sensors are transferred over a single physical medium, for example time division multiplexed and / or on carrier waves of different frequencies.

[0072] In some embodiments, the gyroscope and / or accelerometer and / or magnetometer, or parts thereof, are included in a sensing unit. The sensing unit can be attached to the mandible of the subject. This is an embodiment with a very compact form factor, which allows for easy application and improves patient comfort. It is understood that the provision of the accelerometer and / or magnetometer does not replace the function of the gyroscope, but allows for new interpretations that are only possible by the combination of a gyroscope with one or more additional sensing devices, such as an accelerometer and / or magnetometer. Preferably, first the interpretation of the acquired signals is related to the data from the gyroscope, and in a second step is supplemented with data from the accelerometer and / or magnetometer. For example, first the data from the gyroscope can be used to analyze the angular velocity of the mandible, and a comprehensive cycle can be arrived at in a cycle analysis. The data from the accelerometer can then be used to provide the context in which the cycles are produced, for example the origin of activation (cortical and subcortical), the end type (dynamics of respiratory disturbances), the type of muscular chewing activity (more or less tonic or phasic). The combination of data also allows new assessments to be made that are not possible with data from a single sensing unit alone. For example, a precise description of the type of onset opens the possibility of making predictions about the onset or recurrence of sleep disorders or about changes in breathing (for example peripheral capillary oxygen saturation SpO2).

[0073] Preferably, the size of the sensing unit is at most 5 cm in length, 2 cm in thickness and 1 cm in height, which reduces interference with the subject's normal sleep.

[0074] In some embodiments, one or more of the N mandibular motion classes are associated with a predetermined frequency range. In other words, in these embodiments, one or more of the N mandibular motion classes include mandibular motions occurring in a predetermined frequency range. Preferably, at least two of the N mandibular motion classes are associated with a predetermined frequency range including a predetermined A frequency range and a predetermined B frequency range, and the predetermined A frequency range and the predetermined B frequency range do not overlap.

[0075] In some embodiments, the at least one predetermined frequency range comprises a frequency range of 0.15 Hz to 0.60 Hz, 0.25 Hz to 0.50 Hz, or 0.30 Hz to 0.40 Hz, which is a frequency range of a signal indicative of respiration of the subject.

[0076] In some embodiments, the system further comprises one or more auxiliary components selected from the list comprising an oximeter, and / or a thermometer, and / or an audio sensor, and / or an electromyography unit, and / or a pulse photoplethysmograph. Preferably, these auxiliary components are operatively connected to the analysis unit via a data link.

[0077] In some embodiments, the analysis unit is configured to identify head movements of the subject based on gyroscope and / or accelerometer and / or magnetometer data. Preferably, the head movements include rotation, for example around an axis passing through the center of the subject's head. In that case, preferably at least one of the N mandible movement classes indicates a change in head position. This allows for efficient differentiation between general head movements and movements of the mandible itself. In these embodiments, the system preferably includes both an accelerometer and a gyroscope.

[0078] In some embodiments, the analysis unit is adapted to apply one or more pre-processing steps to the gyroscope data, and / or the accelerometer data, and / or the magnetometer data. The one or more pre-processing steps are selected from the list comprising: applying a band pass filter, applying a low pass filter, an exponential moving average, and / or calculating the entropy of the frequencies of the gyroscope data, and / or the accelerometer data, and / or the magnetometer data. Applying low pass filtering improves the detection of micro-arousals.

[0079] In some embodiments, the analysis unit may include an interpretation module configured to interpret certain parameters measuring sleep quality and extent of sleep disordered breathing. Sleep quality parameters may include, for example, total sleep time (TST), sleep onset latency (SOL), first awakening from sleep onset (WASO), wakefulness index, sleep efficiency (SE), ratio of REM, non-REM sleep, latency of REM sleep, and other metrics of sleep quality indicators. Metrics related to sleep disordered breathing may include hourly incidence and cumulative duration of respiratory effort during sleep. The analysis unit may be configured to report the interpreted subject-specific parameters. The reporting may include providing an output to a device such as a computer or smartphone. The reporting may also include providing a visual or textual report of the subject-specific parameters, for example in the form of a hypnosis.

[0080] In some embodiments, at least one of the N mandibular motion classes indicates that the subject is awake, and a plurality of the N mandibular motion classes indicates that the subject is asleep. The incorporation of the classifications "sleeping" and "awake" in the method ensures that measurements taken while the subject is awake or asleep are interpreted accordingly. The interpretation can be performed using an interpretation module.

[0081] In some embodiments, at least one of the N mandibular motion classes indicates that the subject is in an N1 sleep state, and at least one of the N mandibular motion classes indicates that the subject is in a REM sleep state.Optionally, at least one of the N mandibular motion classes indicates that the subject is in an N2 sleep state, and / or at least one of the N mandibular motion classes indicates that the subject is in an N3 sleep state.

[0082] In some embodiments, at least one of the N mandibular movement classes indicates that the subject is in an N2 sleep state.

[0083] In some embodiments, at least one of said N mandibular movement classes indicates that said subject is in an N3 sleep state.

[0084] In some embodiments, one or more of the N mandibular movement classes are associated with sleep stage detection. The sleep stage detection can be further performed to establish the subject's specific sleep pattern. The sleep stage detection is preferably automated with different levels of resolution.

[0085] In a preferred embodiment, the sleep patterns may include (classified by increasing complexity): (1) Two-class scoring (i.e., binary): to detect the subject's wakefulness or sleep state; (2) 3-class scoring: when classifying the sleep stages including the subject's wakefulness, non-REM sleep stage, or REM sleep stage; (3) 4-class scoring: for classifying the subject's sleep stages, including wakefulness, light sleep (N1 and N2) stages, deep sleep (N3) stages, or REM sleep stages; (4) 5-class scoring: when classifying all sleep stages including the subject's wakefulness, N1 sleep stage, N2 sleep stage, N3 sleep stage, and REM sleep stage.

[0086] Examples 18 and 19 provide exemplary methods for automated detection of sleep stages with three-class scoring.

[0087] In some embodiments, at least one of said N mandibular movement classes is indicative of cortical activation.

[0088] In some embodiments, at least one of said N mandibular movement classes is indicative of subcortical activation.

[0089] In some embodiments, one or more of the N mandibular movement classes are indicative of obstructive apnea, obstructive hypopnea, respiratory effort related arousals, central apnea, and / or central hypopnea.

[0090] In some embodiments, one of the N mandibular movement classes is indicative of bruxism, and the measured rotational movement data exhibits mandibular movement amplitudes of at least 1 mm, or, if the movement is phasic, at an established frequency in the range of 0.5-5 Hz during at least three respiratory cycles, or mandibular movement amplitudes of greater than 1 mm sustained and tetanic for at least 2 seconds.

[0091] Bruxism during sleep is a frequent complaint of 5-10% of adults. It is often intermittent, fluctuating over time, sometimes disappearing for a few weeks and then recurring, either during the night or for several nights in a row. Bruxism is often recognized by the sleeper's partner in the form of an unpleasant and loud grinding sound. This may lead to symptoms of pain in the subject's face or temples and wear of the tooth enamel. Its origin is poorly understood, but obstructive sleep apnea is one of the possible causes.

[0092] In some embodiments, one or more of the N mandibular movement classes indicate loop gain, muscle gain that recruits the mandible during apnea or hypopnea or effort periods, passive collapse point after activation, and / or arousal point before activation.

[0093] Further provided herein is a method for assisting in characterizing a sleep disorder, such as sleep-disordered breathing (SDB), in a subject having a mandible, the method comprising the steps of: - receiving, at a data analysis unit, rotational movement data from a gyroscope placed on the subject's mandible via a data link; - storing, in a memory unit included in the data analysis unit, N mandibular movement classes (note: N is an integer greater than 1, and at least one of the N mandibular movement classes indicates an onset of a sleep disorder (e.g., an onset of sleep disordered breathing (SDB)). < j < N) a mandibular movement class is comprised of a j-th set of rotation values, each j-th set of rotation values ​​indicating at least one velocity, velocity change, frequency, or amplitude of mandibular rotation associated with the j-th class; - sampling data of the rotational movement during a sampling period using a sampling element included in said data analysis unit (thereby obtaining sampled data of the rotational movement); - deriving a plurality of measured rotation values ​​from the sampled rotational movement data using the data analysis unit; and - matching the measured rotation values ​​with N mandibular movement classes using said data analysis unit. Thus, sleep disorders in patients can be detected comfortably and efficiently.

[0094] In some embodiments, the method further comprises the steps of: - measuring acceleration using an accelerometer, said acceleration being indicative of movement and / or position of the subject's head and / or mandible; - transmitting measured acceleration data from the accelerometer to the data analysis unit using the data link; - sampling the measured acceleration data during a sampling period using a sampling element, thereby obtaining sampled acceleration data; - deriving a plurality of measured acceleration values ​​from the sampled acceleration data using the data analysis unit; and - matching the measured acceleration values ​​with N mandibular movement classes using said data analysis unit. Note: For each jth (1 < j < N) The mandible movement class includes a j-th set of acceleration values, each j-th set of acceleration values ​​indicative of at least one mandible or head movement associated with the j-th class. Using both the accelerometer and gyroscope, mandibular movement can be effectively distinguished from whole head movement.

[0095] In some embodiments, the method further comprises the steps of: - measuring magnetic field data using a magnetometer, variations in said magnetic field data being indicative of movement of the subject's head and / or mandible; - transmitting measured magnetic field data from the magnetometer to the data analysis unit using the data link; - sampling the measured magnetic field data during a sampling period using a sampling element included in the data analysis unit, thereby obtaining sampled magnetic field data; - deriving a plurality of measured magnetic field values ​​from the sampled magnetic field data using the data analysis unit; and - Matching the measured magnetic field values ​​with N mandibular movement classes using the data analysis unit. Note: < j < N) the mandibular movement classes are comprised of a j-th set of magnetic field data values; each j-th set of magnetic field data values ​​indicates at least one velocity or velocity change of mandibular movement or head movement associated with the j-th class.

[0096] In some embodiments, the method further comprises using the analysis unit to identify head movements of the subject based on gyroscope and / or accelerometer and / or magnetometer data.

[0097] In some embodiments, at least one of the N mandibular movement classes is indicative of bruxism. Also, the measured rotational movement data exhibits a mandibular movement amplitude of at least 1 mm, or, if the movement is phasic, a mandibular movement amplitude of greater than 1 mm at an established frequency in the range of 0.5-5 Hz during at least three respiratory cycles, or sustained and tetanic for at least 2 seconds. The combination of parameters is indicative of bruxism, thereby effectively detecting bruxism. In some embodiments, the frequency range is 1.0-4.5 Hz, or 1.5-4.0 Hz, or 2.0-3.5 Hz, or 2.5-3.0 Hz.

[0098] In the following, specific embodiments are described for matching data (e.g., preferably sampled rotation, acceleration, and / or magnetic field data) with N mandibular movement classes. These embodiments include extracting features from said data. The features include measured rotation values, and optionally measured acceleration values, and / or measured magnetic field values. After extracting the features, they are matched with one or more mandibular movement classes. Preferably, the mandibular movement classes that the features are matched with include central hypopnea, normal sleep, and obstructive hypopnea. Preferably, the features are matched with the mandibular movement classes with a SHAP score, and the matching is interpreted and explained.

[0099] In some embodiments, the features are selected from a non-exhaustive list including: central tendency (mean, median, and mode) of MM (i.e., mandibular movement, meaning rotation, acceleration, and / or position measured using a gyroscope, accelerometer, and / or magnetometer) amplitude; MM distribution (raw or envelope signal): skewness, kurtosis, IQR, 25th, 75th, and 90th percentiles; extreme values: minimum, maximum, 5th, and 95th percentiles of MM amplitude; variation trends: linear trends and coefficients of tensor product-based spline coefficients (S1, 2, 3, 4) from a generalized additive model to evaluate MM in function of time; and duration of each episode. It should be understood that such features refer to measured rotation, acceleration, and / or magnetic values, whether sampled and / or discretized. Preferably, the values ​​are sampled and discretized. It should be understood that the above list illustrates exemplary embodiments and is not to be considered as limiting the present system.

[0100] In some embodiments, extracting features includes isolating episodes. An episode is a sequence of mandibular motion data (preferably sampled rotational, acceleration, and / or magnetic data) that can be attributed to a single movement of the head and / or mandible. One particular type of episode is normal breathing, e.g., normal breathing for a predefined period of time. The predefined period may be, for example, 2-20 seconds, or 5-15 seconds, 30 seconds, or 10 seconds. The period of time can be adapted to the intended application, e.g., 30 seconds is suitable for identifying sleep stages, 10 seconds for identifying sleep bruxism and micro-arousals, and 20 seconds for identifying respiratory events.

[0101] In some embodiments, the features are extracted according to the following steps 1 to 4. 1. Obtain sampled mandibular movement data. The mandibular movement data includes sampled rotation values, and optionally sampled acceleration values ​​and / or sampled magnetic field values. Preferably, the sampling rate is 1.0-100.0 Hz, or 2.0-50.0 Hz, or 5.0-25.0 Hz, preferably 10.0 Hz. Preferably, the obtained sampled mandibular movement data is obtained for 10.0 minutes to 12.0 hours, or 20.0 minutes to 4.0 hours, or 30.0 minutes to 2.0 hours. 2. Characterize the timestamps of mandibular movement events. 3. For each timestamp ti, perform the following steps: 3. Determine if a.ti is the beginning of a mandibular movement event. 3. If b.ti is the beginning of a mandibular movement event, - assign ti to t_begin and then search for the end (t_end) of the mandibular movement event; and - index t_begin and t_end; 4. For each mandibular movement event E, perform the following steps: 4.a. Calculate the duration of the event: dt=(t_end-t_begin) Measure the statistical distribution of the mandibular movement data sampled during the event. Preferably, this involves calculating one or more features selected from the list including: minimum, maximum, mean, median, mode, 5th, 25th, 75th, 90th, 95th percentiles, skewness, kurtosis, and IQR. Additionally or alternatively, a General Additive Model (GAM) nonlinear model is used to estimate the MM amplitude and / or location with a spline function versus time t, and the coefficients of the spline function are extracted. Additionally or alternatively, a simple linear model is fitted to extract intercepts and slopes from mandibular movements including amplitude and / or position. Optionally, all features are concatenated. Next, the mandibular movement events are matched to mandibular movement classes.

[0102] In some embodiments, exploratory data visualization, one-way analysis of variance, and paired Student's t-test with Bonferroni correction are used to match mandibular movement events to mandibular movement classes. Preferably, in this procedure, the significance level is set at p=0.0001-0.01, more preferably p=0.001.

[0103] In some embodiments, machine learning methods (e.g., extreme gradient boosting, deep neural networks, convolutional neural networks, random forests) are used to classify the measured mandibular movement data into mandibular movement classes.

[0104] In some embodiments, the random forest algorithm employed uses between 20 and 5000, or between 100 and 2000, or between 200 and 1000, or 500 decision trees. In some embodiments, each decision tree is built on a random subset of the features.

[0105] In some embodiments, in model development (i.e., training of the artificial intelligence method), the measured mandibular movement data is randomly split into two subsets: a larger one for model development and a smaller one for model validation. In some embodiments, the larger set includes 60-80%, or 70% of the measured mandibular movement data. In some embodiments, the smaller set includes 20-40%, or 30% of the mandibular movement data. Preferably, a synthetic minority oversampling approach (SMOTE) is used on the training set before developing the model.

[0106] In some embodiments, model development involves evaluating the contribution of multiple features to classification using Lundberg's Shapley additive explanation (SHAP) method, which thus allows the predictions made by the employed machine learning model to be interpreted and makes the model explainable.

[0107] Certain aspects of the present disclosure may alternatively or additionally be expressed as follows: In some embodiments, the system comprises a sensing unit and a device for processing data related to possible disturbances during the subject's sleep. The processing device comprises an identification unit adapted to identify a first and a second measurement signal stream. The first signal has a frequency located in a first predetermined frequency range. The second signal has a value of at least one specific characteristic characterizing a head and / or mandible movement located in a second predetermined range of said values. The first and second predetermined frequency ranges consist of frequency values ​​of the head and mandible movements, respectively, of the subject, characterizing a sleep state of the subject. The identification unit is adapted to generate a trigger signal after observing the presence of the identified first and second signals in the first and second streams for a first predetermined period of time. The identification unit is also adapted to identify, after generating the trigger signal, a third signal in the first and second measurement signal streams, the frequency and / or value of the at least one specific characteristic characterizing a mandible movement and a change in head position of the subject. The identification unit is connected to an analysis unit adapted to be activated under control of the trigger signal. The analysis unit is also adapted to compare the third signal with a profile characterizing frequencies and / or values ​​related to sleep disorders and to produce a result of the comparison. The invention is based on the following concept: in the sleep of a subject, his breathing movements are controlled by nerve centers in the brain, which control the muscles of the head and mandible attached thereto, which in turn position the head and mandible of the subject. The accelerometer and gyroscope provide time streams of measurement signals characterizing the movements of the head and mandible, respectively. Using the identification unit, it is possible to identify from these streams of measurement signals those that characterize the sleep state of the subject and to activate the analysis unit to analyze the sleep disorders affecting the subject during actual sleep.

[0108] It can therefore be seen that the movement of the mandible is not only determined by the movement of the thorax, but also directly by the central nervous systems of the brain which control the muscles attached to it and which position the mandible, and which also control the position of the head.

[0109] In fact, the head may move due to tracheal traction, which is necessarily at respiratory frequencies, so measurements by both accelerometer and gyroscope are preferred. In fact, the gyroscope is more sensitive to the rotational movement of the mandible, which is acted by its own muscles under the direct control of the brain, than the accelerometer, which indicates the head movement that tracheal traction can produce. Besides the respiratory movement, there is another signal of large amplitude that is measured when the center is activated. However, the movement due to tracheal traction is a movement that is damped by the elasticity of the tissues that connect the mandible to the rest of the head, so that the movement can be transmitted passively. It is therefore a relatively imperceptible reflex of the diaphragm that produces the spinal drive, i.e. the tracheal traction, while the antagonist / agonist muscles of the mandible give a direct movement, especially by the action of the driving branch of the trigeminal nerve directly from the brain, i.e. the trigeminal drive. With the gyroscope, it is possible to measure well the rotational movement of the mandible, which is the result of the mandibular muscles and of the direct action of the brain on the mandible. Therefore, combining signals from the accelerometer and gyroscope can improve detection of the origin and nature of mandibular movement and improve determination of whether a person is sleeping.

[0110] Preferably, the sensing unit comprises a magnetometer adapted to measure the movement of the subject's head and / or mandible. The device or unit comprises a third input for receiving a third time stream of measurement signals from the magnetometer. The analysis unit is adapted to integrate the measurement signals from the magnetometer with the third signal. Using the magnetometer, the absolute position of the head and mandible can be measured.

[0111] Preferably, the sensing unit comprises an oximeter, and / or a thermometer, and / or an audio sensor, and / or an electromyography unit, and / or a pulse photoplethysmograph. The identification device or unit comprises a fourth and / or a fifth and / or a sixth and / or a seventh and / or an eighth input for receiving a fourth and / or a fifth and / or a sixth and / or a seventh and / or an eighth time stream of measurement signals from the oximeter, the thermometer, the audio sensor, the electromyography unit, the pulse photoplethysmograph, respectively. The analysis unit is adapted to integrate the measurement signals from the oximeter, the thermometer, the audio sensor, the electromyography unit, the pulse photoplethysmograph, respectively, with the third signal. The identification device or unit is then adapted to associate the measurement signals from the oximeter, the thermometer, and / or the audio sensor, the electromyography unit, and / or the pulse photoplethysmograph with the third signal. These measurement signals from the oximeter, and / or thermometer, and / or audio sensor, and / or electromyography unit allow more measurement signals to be taken into account and therefore improve the reliability of the analysis of the sleep disorder.

[0112] Preferably, the first predetermined range of frequencies is between 0.15 Hz and 0.60 Hz. The identification unit is adapted to identify the first signal over a period of at least two respiratory cycles of the subject. The second predetermined range is comprised of amplitude values ​​of the rotational movements of the mandible, for example with amplitudes of the order of 1 / 10th of a millimeter of normal breathing. The frequency range of 0.15 Hz to 0.60 Hz characterizes a situation in which the subject's head is, so to speak, quasi-immobile, thus reflecting a situation in which the subject is asleep or falling asleep.

[0113] Preferably, the analysis unit is adapted to identify, in the first and second streams, from among the third signal, signals characterizing head rotations about at least one axis extending through the subject's head, where head rotations are often closely associated with arousals, microarousals, cortical and / or subcortical activations during sleep and are indicative of sleep disorders.

[0114] Further provided herein is a method for automatically detecting sleep stages from data of mandibular rotation movements preferably recorded by a gyroscope. The method may be a machine learning based method according to one or more embodiments described herein. The method preferably includes the following steps: - providing sampled rotational movement data from at least one subject; the sampled data may be provided by one or more sampling and processing methods as described herein. - inputting the provided data into a machine learning classifier to generate a prediction score; and - determining a sleep stage based on the calculated score.

[0115] As will be appreciated, the preferred embodiments of the other methods described herein are also preferred embodiments of the method for automatic sleep or sleep stage detection. Data from this method may also be used and input into devices for other methods that may be therapeutic in nature.

[0116] In some embodiments, the sleep stages may include the following classes (categorized by increasing complexity): (1) two-class scoring (i.e., binary) to detect the subject's wakefulness or sleep state; (2) a three-class scoring to classify the subject's sleep stage, including wakefulness, non-REM sleep stage, or REM sleep stage; (3) a four-class scoring to classify the subject's sleep stage, including wakefulness, light sleep (N1 and N2) stage, deep sleep (N3) stage, or REM sleep stage; and (4) Five-class scoring to classify all sleep stages, including the subject's wakefulness, N1 sleep stage, N2 sleep stage, N3 sleep stage, and REM sleep stage.

[0117] Examples 18 and 19 show and discuss examples of methods for automatically detecting sleep stages with 3-class scoring.

[0118] Apart from detecting sleep-related disorders, the systems and methods described herein can also be used for the following exemplary applications: detection of sleep stages and / or monitoring of sleep quality in healthy subjects, elderly subjects, or subjects suffering from abnormal sleep patterns. Detection of sleep disorders, whether clinical or psychological in nature, can tailor treatment or meet the needs of the subject. Furthermore, studying the impact of sleep behavior on the clinical outcome of chronic diseases may provide new insights into the disease and treatment effects. The systems described herein can also be used in combination with other systems or methods. These systems may optionally be used in a therapeutic nature (e.g., respiratory devices (CPAP, BiPAP, adaptive support ventilation), mandibular advancement appliances, and oral devices, devices for percutaneous or implanted nerve and / or muscle stimulation, devices for correcting body and / or head posture and / or position during sleep, etc.). In some embodiments, the system can be connected to an alarm. Or the system may be connected to or comprise a device with alarm functionality.

[0119] Additionally or alternatively, the present invention may be described in the following numbered embodiments, in which the term "combination" is equivalent to the term "system," unless the context clearly indicates otherwise:

[0120] EMBODIMENT 1 A combination comprising a sensing unit and a device for processing data (e.g. a processing unit related to disorders that may occur during sleep of a subject). The sensing unit comprises an accelerometer adapted to measure head and / or mandible movements of the subject and a gyroscope adapted to measure mandible movements of the subject. The sensing unit is adapted to generate measurement signals based on measurements made. The device comprises first and second inputs for receiving first and second time streams of measurement signals from the accelerometer and the gyroscope, respectively. The processing device comprises an identification unit adapted to identify the first and second measurement signal streams. The first signal has a frequency located in a first predetermined frequency range. The second signal has a value located in a second predetermined range of values ​​of at least one characteristic characteristic characterizing the head and / or mandible movements. The first and second predetermined frequency ranges comprise frequency values ​​of the head and mandible movements, respectively, of the subject that characterize a sleep state of the subject. The identification unit is adapted to generate a trigger signal after observing the presence of the identified first and second signals in the first and second streams for a first predetermined time period. The identification unit is also adapted to identify, in the first and second measurement signal streams after generating the trigger signal, a third signal, the frequency and / or value of the at least one characteristic characteristic being representative of the mandibular movement and head position changes of the subject. The identification unit is connected to an analysis unit adapted to be activated under the control of the trigger signal. The analysis unit is also adapted to compare the third signal with a profile characterizing frequencies and / or values ​​associated with sleep disorders and to generate a result of the comparison.

[0121] EMBODIMENT 2 The combination according to embodiment 1, characterized in that the sensing unit includes a magnetometer adapted to measure movement of the subject's head and / or mandible, the device or unit includes a third input for receiving a third time stream of measurement signals from the magnetometer, and the analysis unit is adapted to integrate the measurement signals from the magnetometer with the third signal.

[0122] EMBODIMENT 3 3. The combination according to embodiment 1 or 2, characterized in that the sensing unit comprises an oximeter, and / or a thermometer, and / or an audio sensor, and / or an electromyography unit, and / or a pulse photoplethysmograph, the identification device or unit comprises a fourth and / or a fifth and / or a sixth and / or a seventh and / or an eighth input for receiving a fourth and / or a fifth and / or a sixth and / or a seventh and / or an eighth time stream of measurement signals from the oximeter, the thermometer, the audio sensor, the electromyography unit and the pulse photoplethysmograph, respectively, and the analysis unit is adapted to integrate the measurement signals from the oximeter, the thermometer, the audio sensor, the electromyography unit and the pulse photoplethysmograph, respectively, into a third signal.

[0123] EMBODIMENT 4 4. The combination according to any one of embodiments 1 to 3, characterized in that the first predetermined range of frequencies is 0.15 Hz to 0.60 Hz and the identification unit is adapted to identify the first signal over a period of at least two respiratory cycles of the subject.

[0124] EMBODIMENT 5 A combination according to any one of embodiments 1 to 4, characterized in that the second predetermined range of values ​​includes at least one head movement amplitude value indicative of a change in head position.

[0125] EMBODIMENT 6 The combination according to any one of embodiments 1 to 5, characterized in that the analysis unit is adapted to identify in the first and / or second streams signals characterizing a head rotation around at least one axis extending through the subject's head from among the third signals.

[0126] EMBODIMENT 7 The combination according to any one of embodiments 1 to 6, characterized in that the identification unit is adapted to identify movements of the first and second signal streams characterizing mandibular movements and head position changes of the subject, and the analysis unit is adapted to remove from the movement signal streams at least one feature used to identify information characterizing the movements.

[0127] EMBODIMENT 8 The combination according to any one of embodiments 1 to 7, characterized in that the processing device is adapted to apply a pre-processing to the first and / or second stream by applying thereto a band-pass filter, and / or a low-pass filter, and / or an exponential moving average, and / or a calculation of the entropy of the frequencies of the signal.

[0128] EMBODIMENT 9 The combination according to embodiment 7 or 8 when dependent on embodiment 7, characterized in that the analysis unit is adapted to verify whether, during a second time period, in particular 30 seconds, the at least one characteristic used to identify information characterizing the movement has a value characterizing a sleep state or a wakefulness state, respectively, and the analysis unit is adapted to produce a first data item indicative of a sleep state and a wakefulness state, respectively, if the at least one characteristic characterizing the movement and removed from the analyzed signals of the first and second received streams and used to identify the information has a value describing a sleep state or a wakefulness state, respectively.

[0129] EMBODIMENT 10 The combination according to any one of embodiments 7, 9 or 8 when dependent on embodiment 7, characterized in that the analysis unit is used to identify information characterizing the movements during a second time period, in particular 30 seconds, and is adapted to verify whether the frequencies and / or at least one characteristic removed from the analyzed signals of the first and second received streams have values ​​characterizing an N1 sleep state and a REM sleep state, respectively, and the analysis unit is adapted to produce second and third data items indicative of an N1 sleep state and a REM sleep state, respectively, if the frequencies and / or at least one characteristic characterizing the movements and removed from the analyzed signals of the first and second received streams have values ​​representative of an N1 sleep state and a REM sleep state, respectively.

[0130] EMBODIMENT 11 A combination according to embodiment 7, 9 or 10, characterized in that the analysis unit is adapted to verify whether at least one feature used to identify information characterizing the movement during a second time period, in particular 30 seconds, and removed from the analyzed signals of the first and second received streams has a value characterizing the N2 sleep state and the N3 sleep state, respectively, and the analysis unit is adapted to produce fourth and fifth data items indicative of the N2 sleep state and the N3 sleep state, respectively, if at least one feature characterizing the movement and removed from the analyzed signals of the first and second received streams and used to identify the information has a value representative of the N2 sleep state and the N3 sleep state, respectively.

[0131] EMBODIMENT 12 The combination according to any one of embodiments 1 to 11, characterized in that the analysis unit is adapted to verify whether at least one intrinsic characteristic of the analyzed signals of the first and second received streams has a level characterizing cortical activation and subcortical activation, respectively, during a third time period, in particular 3 to 15 seconds, and the analysis unit is adapted to produce a sixth data item indicative of cortical activation and subcortical activation if said at least one intrinsic characteristic of the analyzed signals of the first and second received streams has a level representative of cortical activation and subcortical activation, respectively.

[0132] EMBODIMENT 13 The combination according to any one of embodiments 1 to 12, characterized in that the analysis unit is adapted to verify whether at least one intrinsic characteristic of the analyzed signals has a level characterizing obstructive apnea, obstructive hypopnea, respiratory effort related arousal, central apnea, and central hypopnea, respectively, and the analysis unit is also adapted to produce seventh, eighth, and ninth data items indicative of obstructive apnea, obstructive hypopnea, respiratory effort related arousal, central apnea, and central hypopnea, respectively, if the at least one intrinsic characteristic of the analyzed signals of the first and second streams has a level representative of obstructive apnea, obstructive hypopnea, respiratory effort related arousal, central apnea, and central hypopnea, respectively.

[0133] EMBODIMENT 14 The combination according to any one of embodiments 1 to 13, characterized in that the identification unit is adapted to identify in the first and second streams values ​​of vibration frequencies and / or values ​​of at least one intrinsic characteristic indicative of variations not observed during a sleep state, to generate a neutralizing signal upon observing such variations and to supply said neutralizing signal to the analysis unit for neutralizing it.

[0134] EMBODIMENT 15 The combination according to any one of the preceding embodiments, characterized in that the analysis unit is adapted to verify, at a frequency established in the range of 0.5-5 Hz during at least three respiratory cycles, whether at least one characteristic feature of the analyzed signals of the first and second streams has increased by more than at least 1 mm, if said movement is phasic or if it has increased by more than 1 mm sustainedly and tonically for at least 2 seconds, and to produce a tenth data item indicative of bruxism during said verification.

[0135] EMBODIMENT 16 The combination according to any one of the preceding embodiments, characterized in that the analysis unit is adapted to capture one or more values ​​of the first and second streams providing access to a calculation of a loop gain of muscle gain recruiting the mandible during periods of apnea, hypopnea or effort from a point of passive collapse after activation and / or a point of awakening before activation. Working Example Example 1

[0136] In Example 1, reference is made to figure 1. Figure 1 shows a system according to the invention. The system comprises a sensing unit 1 and a device 10 for processing data, preferably a processing unit, related to disorders that may occur during sleep of a subject. The sensing unit comprises an accelerometer 2 adapted to measure, preferably in three dimensions, the movements of the subject's head and / or mandible. The sensing unit also comprises a gyroscope 3 adapted to measure, preferably in three dimensions, the rotational movements of the subject's mandible. According to one preferred embodiment, the sensing unit 1 also comprises a magnetometer 4, in particular in the form of a compass, and / or an oximeter 5, and / or a thermometer 6, and / or an audio sensor 7, and / or an electromyography unit 8, and / or a pulse photoplethysmograph 9. Other sensors, such as sweat sensors or nasal pressure sensors, may also form part of the sensing unit. The pulse photoplethysmograph works by transmission or reflection and provides access to the calculation of the pulse frequency and the changes in arterial tone.

[0137] The sensing unit is preferably small in size, for example, at most 5 cm long, 2 cm thick, and 1 cm high, so as not to disturb the subject's normal sleep. The sensing unit is preferably very small in overall size, lightweight and flexible, allowing good ergonomics. The signals produced by the sensing unit are very suitable for decoding using artificial intelligence. The diagnostic power of the measurements obtained by the sensing unit is comparable to that of a complete polysomnographic recording. Mandibular movements may occur preferentially on an axis, for example on the anterior-posterior axis, while the subject's head is turned to the right. Movements on other axes can be measured as well. The sensing unit is preferably intended for one-time use for hygienic reasons, but can of course also be re-used after re-adjustment.

[0138] Preferably, the head position is measured based on values ​​measured along three axes by the accelerometer 2, which measures acceleration values ​​relative to the gravity of the Earth, and therefore preferably integrates these measurements over time to obtain the head position, which would be a relative position if there was no initialization phase during application of the sensing unit to the human head. Said position can be expressed for example in Euler angles pitch, roll and yaw values, or gains with 15° tranches. The head position can also be expressed in the following terms: upright, lying, left, right, supine.

[0139] The following table shows various angle values ​​and their estimated head positions: JPEG2025020260000001.jpg39115

[0140] A magnetometer 4 is added to sense the orientation of the head, especially when the movement occurs perpendicular to the gravity vector. By combining the measurements from the accelerometer and magnetometer, the distance traveled can be calculated, and thus the absolute value of the head position can be obtained.

[0141] As for the head movements, the mandible movements are measured, preferably on three axes, with the aid of measurements from an accelerometer 2. The mandible movements are also measured with the aid of a gyroscope 3.

[0142] There are various types of head and mandible movements and the resulting changes in position. In the case of the mandible, the movements are rotational movements, for example at respiratory frequencies. However, lateral movements are possible during sleep, in the case of bruxism, chewing or oral dyskinesias, and the condyles of the mandible rotate in the glenoid cavity of the temporomandibular joint, but not about the same axis as in respiratory movements.

[0143] In the case of the head, the outcome of the movement is probabilistic, i.e. the position that the head will occupy at the end of the movement is not predictable after activation. The amplitude of the movement and the position change have different values. Therefore, if the amplitude of the head movement is large, the position changes of the mandible measured by the gyroscope are not examined, in such a case the subject will be awakened and no information is obtained about the subject's sleep disorder. Small amplitudes of the mandible movements captured by the gyroscope are observed when they are due to respiratory movements. The changes in the yaw angle are related to the head and indicate that it rotates from left to right. The changes in the pitch angle are related to the flexion or extension of the head, in addition to the fact that they provide information about the mandible movement, despite the use of other parameters. These values ​​of the captured signals are analyzed with the help of said analysis unit, as explained below.

[0144] Mandibular movement can be imposed by respiratory movements as well as non-respiratory movements. Thus, head movements when a person is sleeping can cause mandibular movement. Mandibular movement can be caused by tracheal traction or the human brain. Tracheal traction is the traction force that the chest exerts on the head. The traction is at the human respiratory frequency. Thus, when the head moves at the respiratory frequency, the mandible, which is attached to the head, follows the movement imposed by the head and moves at the respiratory frequency. This is a passive movement that follows the movement of the head. Mandibular movement can be directly and actively controlled by the brain, in which case the head does not move. When the brain controls mandibular movement, it is the mandibular muscle that is directly stimulated. Thus, it is useful to be able to clearly distinguish between mandibular movement controlled by the brain and mandibular movement controlled by tracheal traction.

[0145] A distinction is made between independent mandibular movements (IMM) during brain activation (e.g. at the end of a period of respiratory effort, during coughing, or spitting, or again when speaking in sleep) and respiratory mandibular movements (RMM) caused by the subject's breathing. There are also mandibular movements caused by bruxism and chewing. RMM-type mandibular movements are directly controlled by the subject's brain and do not lead to head movements. RMM-type movements are also caused by tracheal traction and are combined with head movements at the frequency of breathing. When RMM-type movements stop, normalize, or start, it is useful to observe whether the head moved at that time, with the help of accelerometer measurements. Bruxism-type movements often move the head and follow the activation shown by the accelerometer. This is because it actually captures this large amplitude movement, in contrast to the relatively fine rotational movements of the mandible that the gyroscope clearly shows.

[0146] The device 10 according to the invention for processing data related to sleep disorders comprises a first input 11-1 for receiving a first time stream F1 of measurement signals from an accelerometer 2, i.e. data of the measured acceleration. It comprises a second input 11-2 for receiving a second time stream F2 of measurement signals from a gyroscope 3, i.e. data of the measured rotational movement. It may also comprise a third input 11-3 for receiving a third time stream F3 of measurement signals from a magnetometer 4, i.e. magnetic field data. If the sensing unit also comprises an oximeter, the identification device comprises a fourth input adapted to receive a fourth time stream F4 of measurement signals from the oximeter, i.e. oximeter data. If the sensing unit also comprises a thermometer, the identification device comprises a fifth input adapted to receive a fifth time stream F5 of measurement signals from the thermometer, i.e. thermometer data. If the sensing unit also comprises an audio sensor, the identification device comprises a sixth input adapted to receive a sixth time stream F6 of measurement signals from the audio sensor, i.e. audio data. If the sensing unit also comprises an electromyography unit, the identification device also comprises a seventh input adapted to receive a seventh time stream of measurement signals F7, i.e. electromyography data, from the electromyography unit. If the sensing unit also comprises a pulse photoplethysmograph, the identification device also comprises an eighth input adapted to receive an eighth time stream of measurement signals F8, i.e. photoplethysmography data, from the pulse photoplethysmograph. In other words, measurement data from various sensors are transmitted from these sensors via a data link to the analysis unit.

[0147] The various inputs must not be physically distinct, since the various streams may be time division multiplexed and / or carried on different frequency carriers, and therefore the various input streams may be transmitted over a single data link.

[0148] The device includes a data analysis unit including an identification unit 12 adapted to identify in the first and second measurement signal streams F1 and F2 first signals whose vibration frequencies lie in a first predefined range of vibration frequencies and second signals whose values ​​lie in a second predefined range of values. The first and second predefined ranges respectively consist of vibration values ​​of the subject's head and mandible movements characterizing a sleep state of the subject. In case the sensing unit includes a magnetometer 4, the identification unit 12 is also adapted to identify in the third stream of measurement signals F3 a third signal whose values ​​lie in a third predefined range of the subject's head orientations whose values ​​can be observed during sleep. The identification unit is adapted to generate a trigger signal after observing that the first and second signals identified in the first and second streams are present during a first predefined period. The identification unit is also adapted to identify in the first and second measurement signal streams a third signal whose vibration frequencies and / or whose values ​​characterize a mandible movement and / or a change in head position of the subject after generating the trigger signal. The identification unit is connected to an analysis unit 13 adapted to be activated under the control of the trigger signal and adapted to compare the third signal with a profile characterizing frequencies and / or values ​​associated with a sleep disorder and to produce a result of the comparison.

[0149] In particular, the identification unit may be included in a data analysis unit which also includes a memory unit. The memory unit is configured to store N mandibular movement classes, where N is an integer greater than 1, and at least one of the N mandibular movement classes is indicative of an onset of sleep disordered breathing. < j <N) mandibular motion classes are composed of a j-th set of rotation values, each j-th set of rotation values ​​indicating at least one speed, speed change, frequency, and / or amplitude of mandibular rotation associated with the j-th class. Each j-th mandibular motion class also optionally includes a j-th set of acceleration values ​​and / or a j-th set of values ​​of magnetic field data. The data analysis unit includes a sampling element configured to sample the measured rotational motion data and optionally the measured acceleration data and / or the measured magnetic field data during a sampling period, thereby obtaining the sampled rotational motion data and optionally the sampled acceleration data and / or the sampled magnetic field data. The data analysis unit is configured to derive a plurality of measured rotation values ​​from the sampled rotational motion data, and optionally derive a plurality of measured acceleration values ​​and / or measured magnetic field values ​​from the sampled acceleration data and / or the sampled magnetic field data. The data analysis unit is further configured to match the measured rotation values ​​with the N mandibular motion classes. Optionally, the data analysis unit is further configured for matching the measured acceleration and / or magnetic field values ​​with the N mandibular movement classes. Thus, an onset of sleep disordered breathing is effectively detected.

[0150] With regard to the data link, the device and the sensing unit preferably communicate wirelessly, although of course a cable connection is equally possible. The device is preferably part of a computer located in a data processing center. Wireless communication is for example carried out with the aid of a telephone communication network, the sensing unit being equipped with for example a Bluetooth system that allows communication with a telephone. Thus, a stream of measurement signals produced by the sensing unit is transmitted to the device.

[0151] The invention is based on the following observation: that the mandibular movements are not only determined by the movements of the chest, but also by direct control from the nerve centres of the brain that control the muscles attached to the mandible and whose role it is to position the mandible, as shown in the literature. As observed, the position of the head, and especially the change of the head during sleep, can either stop all mandibular movements or start them, completely independent of said movements of the chest. That is, the mandibular movements can only follow the movements of the chest if the position of the head allows it and does not fixate it. The movements of the head can therefore act or paralyse the movements of the mandible. In this sense, they are nothing but epiphenomena of brain activation that characterise micro-arousals or arousals and that can have other effects on the movements of the mandible.

[0152] In fact, head movements affect the permeability of the upper airway by exerting a crushing force when it collapses in sleep or by activating / deactivating the muscle motor units of the upper airway. These head movements during sleep modify the permeability of the upper airway, so it is necessary to know and superimpose in time the movements of the mandible. These movements of the mandible can therefore be correctly analyzed and interpreted in terms of the variations in respiratory control that start from the airflow produced by the sleeping subject. In other words, sensing and analyzing the movements of the mandible taking into account the position of the head and its changes during sleep, whether during microawakening or awakening, is to take into account the brain control to position or reposition the mandible by activating / deactivating the muscles attached to the mandible. Besides the activation of the brain, the movements of the head position at the respiratory frequency are produced by tracheal traction, while the movements of the mandible at the same frequency are directly determined by the nerve center.

[0153] By operating like a lever the mobile bone formed by the mandible, the brain control tries to stiffen the upper airway by activating the tongue and the muscles of the pharynx attached to it in order to avoid apnea. For this purpose, the brain control relies on the muscles that raise and lower the mouth and open and close it during sleep at the respiratory frequency. The brain control can also act on the muscles that push the mandible forward at the respiratory frequency, or even act in combination with these combined muscle groups involved in movements in different directions.

[0154] Head position changes during sleep are often accompanied by arousals or microarousals that are also recorded, for example, with electrodes placed on the scalp. This records activity in the brain's cortex. After all, if there is head movement with changes in mandibular movement, the scalp electrodes may not record activation. The reason for this is that the activation remains subcortical and sometimes purely autonomic in the brainstem. These head movements are performed completely separately from chest movements.

[0155] A distortion of the neck is produced because this position of the head no longer coincides with that of the body, or because a change in head position is an epiphenomenon of a voluntary or involuntary toss under central nervous control. Analysis of the mandibular movements, in the vertical and horizontal planes, as a function of the resulting head position, provides information on the level of respiratory effort, and in particular its amplitude, that the control by the central nervous centers of the brain adopts upon a change in the resistance to the flow of air through the airways. A respiratory event is considered an increase in effort when the control from the central nervous center increases, and central when the control from the central nervous center decreases. The brain control that allows the organism to exit apnea must actuate the mandibular lever upwards in the vertical plane and forwards in the horizontal plane, ideally with the head in axial alignment with the body, to prevent compression of the upper airway. The (micro)arousal itself is identified by an independent large mandibular movement (IMM), the duration of which is measured and clearly differentiated from the mandibular movement that follows, whether respiratory or non-respiratory. Example 2

[0156] In Example 2, reference is made to Figures 2A and 2B.

[0157] During sleep, the results of the analysis that affect the measurement data stream from the sensing unit provide information on the brain's control state, and the changes in head position indicated by the signal from the accelerometer are often markers of that state change. Figures 2A and 2B show the streams during a change in head position of a human lying in bed. This movement cannot be superimposed on the movement of the mandible during wakefulness, and therefore on the conscious state during chewing, phonation, or swallowing, as studied by specialists outside of sleep medicine. The latter are related to chewing, phonation, and swallowing problems studied in dentistry, stomatology, maxillofacial surgery, orthodontics, orthodontics, logopediatrics, etc., in conscious subjects not in a sleep state.

[0158] FIG. 2A shows the change of the head, from left to right, first from a first position where the head is turned to the left, to a second position where the head is turned to the right. Then, a change is seen to a third position where the head is turned to the left again. The first stream F1 produced by the accelerometer is related to the three axes (Fx, Fy, Fz) of the three dimensions in which the measurements are made. The second stream F2 produced by the gyroscope is also related to the three axes. It can be clearly seen that at the moment of the head rotation, the two streams have a high amplitude peak. It can also be seen that when the head is in the first position, the streams F1 and F2 have a larger amplitude variation and a variation of the control strength, especially in the vertical direction y of the stream F1, than that shown by reference 1, indicating an increased brain control state. Moreover, this is also seen in the stream Ft, which indicates the movement of the chest. The analysis unit can therefore infer from these streams that the person is showing an increased and fluctuating respiratory effort.

[0159] When the head is rotated and in the second position, we see a significant decrease in the amplitude of stream F1, as well as stream F2. As shown in reference 2, the level of stream F1 decreases, which indicates that the mouth is open. We also see that the airflow F5 decreases and there is a possible loss of oxygen flow (reference 3). As shown in reference 4, we also see a decrease in amplitude in stream F2, which indicates that the brain's control amplitude is lost. All this indicates that the amplitude of the effort is decreased and that breathing is affected (see airflow F5). This further activates the brain and generates commands. A new change in its position occurs, with the head turning to the left. After this, we see a larger amplitude in stream F2 and an increase in airflow F5. Thus, as can be seen, the brain's control tends to normalize breathing.

[0160] Figure 2B shows that even slight changes in head position are caused by brain control. Figure 2B shows the changes when the head is rotated slightly to the right. Stream F1 first shows that the brain control state increases, as indicated by arrow 1, and breathing effort occurs. As shown in Reference 2, when the head position changes, it can be seen that the accelerometer (F1) shows an increase in amplitude and frequency indicating brain activation. In stream F8 (EEG) and stream F7 (EMG), brain activation is clearly seen for 30 seconds and is magnified here (Reference 2). Then, it can be seen that the level of stream F1 (Reference 3) indicates a brain control state with a decreased amplitude, with the mandible raised (mouth closed).

[0161] Surprisingly, the technique employed by the system according to the invention provides information on the nature of mandibular movement during sleep, its central origin and its central nervous control: the pharynx must be stiffened to maintain ventilation and thus oxygenation of the subject, but its head end must ideally remain aligned with the body, especially the trunk, during sleep. Therefore, mandibular movement must be interpreted as a function of head position and its changes, otherwise it would not be understood why it stops or starts or changes in amplitude during sleep. Example 3

[0162] In Example 3, reference is made to Figures 3A and 3B.

[0163] The techniques provided herein can be applied to the detection of bruxism. A known diagnosis of bruxism involves electromyography of the masseter and anterior temporal muscles, and possibly the anterior temporal muscles, during a polysomnography test in a laboratory. The test further includes audio-video recording. The demand for sleep recording is not proportional to the recording capacity of sleep laboratories, making the test expensive, cumbersome, and difficult to access. The recordings are performed overnight and are difficult to reproduce. Also, to track bruxism, it is necessary to make recordings over several nights, as they are not systematically reproduced every night and may remain intermittent. Therefore, it is necessary to carry out the recordings in the subject's home, under real life conditions, without disturbing the natural progression of sleep. In order to optimally control bruxism and verify the effectiveness of treatment, the results must be provided quickly.

[0164] Currently, bruxism is not detected at home because there are no technical solutions to do this. Proposed solutions such as surface electromyography of the masseter and temporalis muscles do not allow a definite diagnosis. In fact, the only recording of the masseter or temporalis muscle electromyographic activity may be affected by nocturnal feeding activity or because the fat medium of the muscle prevents the capture of its electromyographic (EMG) activity. Video recording allows the laboratory to verify that the mandibular movements and the resulting electromyographic activity are associated with bruxism.

[0165] The technical solution proposed by the invention consists in recording the mandibular movements, preferably in space, on three main axes with the help of said sensing unit, and then carrying out an algorithmic analysis of said signals with the help of said analysis unit, which allows to identify the mandibular movements that occur specifically and exclusively during episodes of bruxism, established by the detection of RMMA (rhythmic masseter activity), i.e. phasic but sometimes tonic activity, during surface electromyography of the masseter muscles. The stream of signals produced by said sensing unit is analyzed in said three axes, which also allows to capture the lateral movements imposed during bruxism and which may contribute to the wear of the enamel. The mandibular movements referred to as bruxism are the result of the simultaneous action of agonist and antagonist muscles, including not only the mandibular elevation group, such as the anterior temporalis muscle, but also the medial and lateral pterygoid muscles, both the vastus medialis and the vastus lateralis.

[0166] Figures 3A and 3B show the streams captured by the capture unit during bruxism. The recorded muscle EMG activity, seen as streams F7D and F7G, is confirmed to contribute to mandibular movements. Typical features of masseter and / or anterior temporalis EMG activity are reflected in the mandibular movements that are also the etiology of bruxism. The latter, in the form of modulated signals, are superimposed on the tonic (sustained) or phasic (rhythmic) EMG attacks of bruxism that produce them. The duration of this cycle or attack can be calculated.

[0167] Before the onset of bruxism, a period of effort, indicated by arrow 1, can be easily identified by the analysis of the mandibular movement and by a temporary arousal, indicated by arrow 2, accompanied by a cortical or simply subcortical activation of the autonomic nerves. Whether cortical or not, reflected only in changes in the cortical wave frequency of the EEG, as shown for example in stream F8, or subcortical and not visible in the EEG, the activation is well characterized by previous mandibular movements and often precedes the onset of bruxism, as described in the literature. As can be seen, the peak of phasic and / or tonic activity of the masseter muscles occurs simultaneously with the extreme position of the mandibular movement. This clearly confirms the relationship between muscle recruitment and the movement of the mandibular mobile bone. In Figure 3A, in stream F1, a period of effort, indicated by arrow 1, is seen, followed by an activation, indicated by arrow 2, and then the mandibular movement due to bruxism, indicated by arrow 3. Figure 3B is an enlarged view of the 10-second period indicated by arrow K in the upper right corner of Figure 3A. Figure 3B shows the synchrony between EMG activity of the right masseter (F7D) and left masseter (F7G) and mandibular movement during bruxism.

[0168] Here we see that the resumed activity of stream F7 (EMG) of the right masseter (F7D) is synchronized with the activity of the left masseter (F7G) and with the activity of the mandibular movement due to bruxism. As clearly shown in this figure, after a period of effort clearly shown in F1Z and F2X, the mandibular position of abnormal amplitude changes at the respiratory frequency. In F1Z, a large movement accompanied by a head movement follows, while in gyroscopic F2X, the movement showing cortical activation is followed by four rotational movements at high frequency (1 Hz) corresponding to the onset of bruxism. The effort period is then reproduced.

[0169] The head and mandible movements analyzed through their intrinsic properties, i.e., notably the frequency and morphological properties of the signal stream, can be differentiated and sequentially ordered in time as a function of their generating mechanism. These properties can be observed, for example, by analyzing the amplitude, area or gradient of the measured signals. They are, for example: - Movements associated with respiratory effort, followed by Movements associated with transient cortical or subcortical activation, followed by - movements associated with distinct bruxism or chewing movements, such as the number of attacks during a bruxism cycle, the length of the cycle between two attacks, and the duration of attacks.

[0170] Mandibular movement is produced by the agonistic / antagonistic action of muscles to raise and lower the mandible. The latter is directly controlled by the cranial nerve center core of the trigeminal drive branch. Here, mandibular movement can be sensed by the change in the angle the mandible makes while moving relative to a plane, for example, when moving perpendicular to the horizontal plane.

[0171] Mandibular movements may start or stop only if the head position changes, even if the thorax movements continue. Head position changes always occur simultaneously with cortical or subcortical microarousals and therefore disturb the control by the cranial nerve centers. Mandibular movements may continue at the respiratory rate during sleep even if abdominal and thoracic movements are absent, i.e., even if the diaphragm muscles that perform the expansion of the chest and abdomen during inspiration controlled by the spinal nerves are no longer functional or have stopped functioning. Mandibular movements may then be exerted in another plane, for example in the horizontal plane, in the form of anterior-posterior or posterior-anterior movements, i.e., in a plane other than the plane of rostral-caudal traction in which the tracheal traction is affected.

[0172] The first stream provided by the accelerometer, as well as the second stream provided by the gyroscope, show a tonic phase movement at the respiratory frequency of the mandibular position that is upward, i.e. in the direction opposite to that exerted by the traction forces produced by the tracheal traction. This upward and forward movement is produced by the anterior temporal and masseter muscles, respectively, and by the contraction of the pterygoid muscles, especially by the upper muscle groups.

[0173] When the respiratory effort begins and the amplitude of the mandibular movement increases due to the increase in central respiratory control, the direction imposed on the mandibular movement may also be in a plane other than the vertical plane, which was the plane of pull. This is due to the action of certain muscle groups that are recruited more than others, for example the pterygoid group, which is recruited more than the pterygoid group. Movements at respiratory frequencies may occur in more horizontal directions that are captured by the inertial unit, which is composed of an accelerometer and a gyroscope. In fact, if the effort is observed only in the vertical plane, periods of effort may escape the signal analysis. The movement may also occur mainly in one direction (vertical or horizontal) rather than in another direction.

[0174] The shape of the respiratory movement, and in particular its acceleration gradient, changes as a function of the recruited muscle group. During vertical movements, when the masseter muscles are active, the direction of movement during inspiration is upwards, opposite to the direction observed when the antagonist muscles dominate, causing a decrease in movement, a situation that may lead to changes in the waveform of the movement.

[0175] This analysis of the stream delivered by the sensing unit makes it possible to verify the fact that the movement of the mandible during inspiration is downwards when the activity of the depressor muscles dominates, and upwards when the activity of the elevator muscles dominates. This information is obtained by analyzing the changes in the captured speed and acceleration, which allows access to the level and nature of the subject's reaction to avoid respiratory episodes and to a greater or lesser recruitment of the mandible elevator muscles to stabilize the upper airway. Example 4

[0176] In Example 4, reference is made to FIG.

[0177] The observed streams are the identification of four features that explain the behavior of the mandible during the attack. These features allow us to understand the mechanism of respiratory attack in a specific position of the head, in a specific stage of sleep, for the subject, and the reaction of the brain to release itself. In addition to explaining the progression of the attack, information on the risk of relapse in the short and long term can be identified. These characteristics have a predictive value, for example, in the case of high values ​​of the amplitude of the response to the disturbance, called loop gain, i.e. high response to disturbance. The loop gain is shown in Figure 4. In this figure, arrow 1 indicates the collapse point of stream F1, i.e. the solution where the exercise of brain control is no longer exercised, so that the mandible passively lowers under the influence of local anatomical constraints, such as the weight determined by the subject's obesity. Arrow 2 indicates the mandible movement, which is also seen at the same time in stream 2. The peak-to-peak amplitude of the mandible movement at the beginning of arrow 2 is low. Then, afterwards, the mandible lowers during the opening of the mouth. This can be seen at the level of stream 1, which lowers, resulting in an increase in the peak-to-peak amplitude. The level of stream 1 then reaches a level indicated at 3, corresponding to the arousal point, followed by a much larger amplitude peak indicated at arrow 4. This large amplitude movement allows the measurement of the loop gain that accompanies the closing of the mouth, as indicated by the peaks in streams F1 and F2, and the maximum value reached by stream 1 during which the mouth closes again. The loop gain indicates the response to the disturbance. It is calculated as the ratio of the difference between the points of interest indicated by arrows 4 and 3 for the numerator, and arrows 3 and 1 for the denominator.

[0178] A high risk of spontaneous recurrence of episodes is seen with short apneas, especially in the central form. The duration of episodes can be predicted by evaluating the muscles of the upper airway, especially the phasic gain. A low muscle gain means that there is a risk of the episode lasting longer than in the case of a high gain. The arousal point, which is the lowest point of the mandibular position just before the activation that ends the episode, also allows the prediction of the duration of the episode. If the position is not lowered much, there is a risk of the episode being repeated, which may occur periodically. Also, the influence of anatomical constraints, such as those related to weight or local accumulation of adipose tissue in the upper airway, can be measured by calculating the position of the mandible based on the measurements of the accelerometer, especially at the time of micro-arousal or mandibular descent immediately after arousal when the latter is still dominant over the central (collapse point). Example 5

[0179] In Example 5, reference is made to FIGS.

[0180] The stream of measurement signals produced by the sensing unit may contain noise that affects the measurement signal. This may be used to pre-process the stream when it is received by the device. The principle of the pre-processing is simply to produce an enhanced signal. From the analysis, the skilled person could know that during a period of enhanced brain control state, the position of the mandible, and therefore its velocity and acceleration, changes periodically at a frequency of the same order as the respiration rate, i.e. between 0.15 Hz and 0.60 Hz, with approximately the same values. For example, by low-pass filtering of the measurement signals from the accelerometer and gyroscope, it is possible to isolate the signal related to micro-arousals by retaining only the lower frequencies of the band of respiration frequencies. Figure 5 shows that by applying the pre-processing, micro-arousals representing activation are suppressed compared to periods of enhanced brain control state. A clear peak is seen in the signal during each micro-arousal. The application of the pre-processing is possible, for example, by applying a 6th order Butterworth filter, well known in the field of digital signal processing.

[0181] Conversely, periods of enhanced brain control can be captured by filtering one of the captured signals with a bandpass filter corresponding to the respiration frequency band. The result of applying this type of filter to the signal from the gyroscope is shown in Figure 6, which shows that the signal is high during periods of effort.

[0182] Characteristics that can be used to identify information about a stream of measurement signals are, for example: - Head and mandible position (e.g. roll, pitch, and yaw angles) - Acceleration along the mandible and head axes - Rotational speed along each axis of the mandible and head - the standard of rotational velocity around one or more axes of the mandible and head (in space, if a vector u has coordinates (x, y, z), its standard is written as follows: (x 2 + y 2 +z 2 ) 0.5 ) - Acceleration standards along one or more axes of the mandible and head - The median of the values ​​measured over 10 or 30 seconds or defined over two activations - Values ​​measured over 10 or 30 seconds or the average of values ​​defined over two activations - The maximum value measured over 10 or 30 seconds or defined over two activations - The minimum of the values ​​measured over 10 or 30 seconds or defined by two activations - Standard deviation of values ​​measured over 10 or 30 seconds or defined over two activations - Exponential moving average of measurements (with half-lives of 5, 60, 120, and 180 seconds) - Fourier transformation and integration over all frequencies of the measurements, respiratory frequency band (0.15-0.60 Hz) and low frequency band (0-0.10 Hz) - Fourier transformation and identification of the energy maximum or second energy maximum frequency of the measured value - Shannon entropy over a 90 second window of measurements - Time offsets of rotational velocity and acceleration signals of the mandible, head, and other properties to take into account the past and future.

[0183] It is likewise possible to combine the above methods with each other.

[0184] Once the characteristics have been identified in the stream of measurement signals, the analysis unit can proceed to analyze them. For that purpose, an artificial intelligence is used, for example invoking a random forest type algorithm. The features thus extracted from the entire series of signal fragments, for which the polysomnographic results are known, are injected into the algorithm in parallel with the expected results in order to produce a model allowing a pattern recognition type classification of new fragments.

[0185] A signal pattern is a particular state of a signal sequence, which can be represented physically or mathematically through parameters. Pattern recognition is the process of identifying (classifying) specific patterns in the signal using automated learning algorithms based on previously acquired information or statistical parameters extracted from the signal.

[0186] Deep learning is an automated machine learning technique that involves models inspired by the structure of the human brain called artificial neural networks. These networks are composed of multiple layers of neurons that allow the extraction of information in data and the production of results. The technique is highly effective on unstructured types of data such as images, sequences, and biological signals.

[0187] Automated learning (or statistical learning) is a branch of artificial intelligence whose goal is the application of statistical modeling techniques that give machines (computers) the ability to learn information from data, thereby improving their performance in solving tasks, without being explicitly programmed for each task.

[0188] Artificial intelligence (AI) is a set of technologies aimed at enabling machines to simulate intelligent activities.

[0189] The development of these models can proceed, for example, as follows. 1) 200 subjects will be equipped with sensing units and simultaneously undergo polysomnography, the reference clinical test in the field of sleep. 2) Then, each random forest model is trained using the signals captured from 40 of these subjects: the signals from the sensing units and the subset of features obtained after the preprocessing step are injected in combination with the reference results of the sleep studies in a random forest algorithm, which produces a classification model based on this input data. 3) The remaining subjects are then used for model validation: signals from the sensing units corresponding to these subjects are injected into the model produced in the previous step to produce results, and these results are compared with the results obtained by polysomnography. If the results obtained by the model and the results obtained by polysomnography are deemed to be in sufficient agreement, the model is deemed valid. Otherwise, the development is resumed from step 2 of this section.

[0190] In order to be able to reliably identify disorders occurring during the subject's sleep, it is preferable to be able to observe that the subject has actually entered a sleep stage. Once it is detected that the subject is actually in a sleep stage, it is also possible to establish the subject's sleep stage so that the signals present in the stream of measurement signals can be correctly interpreted. Upon falling asleep, a mandibular respiratory frequency of, for example, 0.15 Hz to 0.60 Hz is assumed. To verify a stable sleep state, said respiratory frequency must be present for several tens of seconds in succession. Example 6

[0191] In Example 6, the different sleep stages of the subject are described. In particular, Table 1 (described below after the examples) shows the different sleep stages of the subject and their relationship to the mandibular movement and head position. The wakefulness state is essentially characterized by the unpredictable movement of the mandible in that state, whereas the sleep state without sleep disorders is characterized in the subject by a rotational movement of the mandible at the respiratory frequency. To detect the wakefulness state and the respective sleep state, the analysis unit preferably works with an analysis window of 30 seconds and pre-processes the first and second streams with a band-pass filter and / or an exponential moving average. To extract a profile characterizing the wakefulness state and the respective sleep state, for example the normalized average level is taken into account, which is in fact higher in the wakefulness state than in the sleep state.

[0192] Also, during sleep, the following stages are distinguished: N1, N2, N3, and REM (rapid eye movement). In the N1 sleep stage, the mandibular movement changes with the respiratory rate, and in adults, peak-to-peak amplitude variations of limited duration are often observed, often lasting a few minutes. While the head position is generally stable, the mandibular position may remain unpredictable or change periodically. To detect the N1 sleep stage using a processing device, it is preferable to use an analysis window of 30 seconds to ensure the continuity of the movements. A pre-processing of the first and second streams by calculating the entropy of the frequency of the signal can be used. The normalized average level is considered in the first approach as a profile characterizing the N1 sleep stage, but other approaches can be used to improve the accuracy of the analysis. In the N1 stage, the normalized average level is higher than in the N2 or N3 stages.

[0193] The analysis unit is adapted to verify whether, during a second time period, in particular 30 seconds, the normalized average and variance of the amplitude and frequency of the received first and second streams have levels characterizing the sleep state of N1. The analysis unit is adapted to produce a second data item indicative of the sleep state of N1 if the normalized average and variance of the amplitude and frequency of the received first and second streams have levels characterizing the sleep state of N1.

[0194] In N2 and N3 sleep stages, the amplitude and / or frequency fluctuation of brain control becomes lower and lower from N2 to N3. Therefore, in normal subjects, there is practically no mandible or head movement in these stages. To detect N2 or N3 sleep stages using a processing device, an analysis window of 30 seconds is preferably used to ensure the continuity of movement. Pre-processing using a low-pass or band-pass filter is also preferably used. The normalized average level is considered in the first approach as a profile that characterizes the N2 sleep stage. In the N2 and N3 stages, the normalized average level becomes lower and lower, respectively. The normalized median level can also be used to identify the N2 or N3 stage or other statistical measurement techniques.

[0195] The analysis unit is adapted to verify whether, during a second time period, in particular 30 seconds, the normalized mean and / or normalized median of the received first and second streams have a level characterizing the N2 and N3 sleep states, respectively, and if the normalized mean and / or normalized median of the received first and second streams have a level characterizing the N2 and N3 sleep states, respectively, the analysis unit is adapted to produce fourth and fifth data items indicative of the N2 and N3 sleep states, respectively.

[0196] In humans, the REM stage is characterized by unpredictable movements of the mandible. To detect this type of stage with a processing device, it is preferable to use an analysis window of 30 seconds to ensure the continuity of the movements. In adults, this type of movement of unpredictable frequency and / or amplitude often lasts longer in the REM stage than in the N1 stage. The duration of such movements during the N1 stage is often limited to a few minutes. The direction of the mandible movement during brain activation is often negative due to the mouth opening. During the REM stage, fluctuations of the mandible movement at the respiratory frequency are observed, with peak-to-peak amplitude fluctuations that are not periodic. The head position usually does not change during the REM stage. The detection is performed in a similar way to the N1 period, the aim being to observe respiratory instability during mandible movements. It is common to enter the REM stage without cortical activation that the EEG can capture and without head movements. Thus, the accelerometer does not measure anything, but the gyroscope observes changes in mandibular rotation. This shows the importance of having both the signal from the gyroscope and the signal from the accelerometer to correctly observe the transition to the REM stage. The end of the REM stage is often closely related to brain activation observed by the accelerometer and gyroscope, which observe the independent mandibular movement (IMM) and, if applicable, head movement. The normalized average level can be taken into account as a first approach. For example, the amplitude and frequency variations are also examined. During the first 15 minutes of sleep, the detection of REM allows the diagnosis of hypersomnia. Example 7

[0197] In Example 7, reference is made to FIG.

[0198] For example, comparative analysis of the signal amplitude and / or movement frequency value variations and / or other statistical characteristics of the signal, alone or grouped into a classifier, such as a random forest type classifier, during sleep can be applied to perform statistical inference and distinguish the various stages. For that purpose, FIG. 14 shows a spectrogram of the distribution of mandibular movement frequencies to distinguish the stages. In FIG. 14, the vertical axis represents the amplitude density and the horizontal axis represents the frequency. These specific characteristics of each sleep stage can also be identified by deep machine learning. This algorithmic and / or statistical approach can also be used to characterize respiratory and non-respiratory movement onsets.

[0199] The following table shows an example of the variation of the amplitude level of the mandibular rotation signal during different stages of sleep. In this table, "interval" means the interval between the upper level of percentile 2.5 and the lower level of percentile 97.5, "amplitude" means the difference between the maximum and minimum values, and "variation" is a measure of the spread of the considered values. The measurement is based on 1000 samples taken during 30 seconds in each stage. JPEG2025020260000002.jpg46131

[0200] The analysis unit is adapted to verify whether the normalized average and variance of the amplitude and frequency of the received first and second streams during a second time period, in particular 30 seconds, have levels characterizing a REM sleep state, and the analysis unit is adapted to produce a third data item indicative of a REM sleep state if the normalized average and variance of the amplitude and frequency of the received first and second streams have levels characterizing a REM sleep state.

[0201] The identification unit is adapted to identify in the first and second streams movement signals characterizing the rotation of the mandible and / or the movement of the head of the subject. The analysis unit is adapted to analyze these movement signals, for example by applying a band-pass filter and an exponential moving average or a measure of the entropy of the frequency of the signal to these signals. For example, by applying a band-pass filter to the respiratory frequency and this exponential moving average, for example with a half-life equal to 5, 60, 120 or 180 seconds, to the first and second streams of the supplied signal, with a first observation period of 30 seconds, the analysis unit can observe whether the signal is unstable. In this case, a wakefulness state is observed. On the other hand, if the signal is stable, a sleep state is observed.

[0202] The analysis unit is adapted to apply an exponential moving average over a second period of time, between 30 seconds and 15 minutes, in particular 3 minutes, of the first and second streams as a profile characterizing a sleep state. In some analyses, the second period of time may be 30 minutes. The analysis unit is adapted to verify whether the exponential moving average has a substantially constant value during the second period, and to produce a first data item indicative of a sleep state or a wakefulness state, respectively, if the value is substantially constant or not constant, respectively.

[0203] The identification unit is adapted to identify in the first and second streams movement signals characterizing the rotation of the mandible and head of the subject. The analysis unit is adapted to calculate the entropy of the frequencies of these movement signals. By applying the entropy function to the first and second streams of the supplied signals, for example with an analysis window of 90 seconds and an observation period of 30 seconds, the analysis unit is able to observe a normalized average level. If the level is high, an N1 or REM sleep state is observed as a function of the level value.

[0204] The identification unit is adapted to identify in the first and second streams movement signals characterizing the subject's mandibular rotation and / or head movement. The analysis unit is adapted to apply a band-pass or low-pass filter to these movement signals. By applying the band-pass filter, for example at the respiratory frequency, or the low-pass filter (for example below 0.10 Hz) to the first and second streams of the provided signals and an observation period of 30 seconds, the analysis unit can observe normalized mean and / or median levels. As a function of the levels, N2 or N3 sleep status is observed.

[0205] Brain activation in the form of microarousals can last from 3 to 15 seconds and can be of cortical or subcortical type. Brain activation leading to arousals lasts for more than 15 seconds. Corticocerebral activation in REM sleep can be accompanied by repeated lowering of the mandible as a characteristic feature. In the case of cortical activation, the corticobulbar reflex is activated and multiple sudden movements of the mandible of large amplitude or long duration are observed. The reflex amplifies the movement. In the case of subcortical activation, this reflex is not activated and only one sudden movement of smaller amplitude can be observed, accompanied by a discontinuity in the frequency of the mandible relative to the respiratory frequency acted upon. This movement may be of much lower amplitude and shorter duration than when the corticobulbar reflex is activated. This movement is therefore often less noticeable and its identification can be aided by the detection of a concomitant movement of the head, which can only occur over a very short distance. Example 8

[0206] In a further embodiment, reference is made to Table 2, which shows the characteristics of cortical and subcortical brain activation. As a result of the cortical activation, the mandible suddenly opens and closes with large amplitude for 3 to 15 seconds. If the cortical activation occurs during sleep, it is generally accompanied by a change in the subject's head position. The analysis unit analyzes the amplitude and duration of this movement in a 10 second window using the first and second data streams.

[0207] Subcortical activation is characterized by discontinuities in the frequency and shape of the mandible movement changes, while the mandible remains stable most of the time. The analysis unit analyzes the amplitude and duration of this movement in a 10 second window using the first and second data streams. The analysis can be performed on continuous variables as well.

[0208] Thus, the analysis unit verifies whether the amplitudes of the received signals of the first and second streams have levels characterizing cortical and subcortical activation, respectively, during a third time period, in particular 3 to 15 seconds, and is adapted to produce a sixth data item indicative of cortical activation and subcortical activation, respectively, if the amplitudes of the received signals of the first and second streams have levels characterizing cortical and subcortical activation, respectively.

[0209] To detect the presence or absence of respiratory events or non-respiratory movements, the analysis unit analyzes the changes in mandibular position, the peak-to-peak amplitude of mandibular movements, the variations in the peak-to-peak amplitude of mandibular movements, which indicate the variations in the amplitude of brain control, and the frequency of mandibular movements. If low amplitudes are observed, i.e., amplitudes corresponding to those observed during normal respiratory movements, in the presence of stable central (mandibular movements occurring at a continuous and stable degree of mouth opening), there are no episodes that should be considered as sleep disorders.

[0210] If high respiratory control amplitudes are observed, e.g., amplitudes corresponding to movements of more than 0.3 mm, i.e., amplitude changes greater than those observed during normal movements, then an increase in movement or respiratory effort is presumed, which may indicate a sleep disorder.

[0211] A central respiratory onset is presumed if a significant decrease in the amplitude of respiratory control is observed, i.e., the centrality of control is stable or unstable, e.g., low, e.g., of the order of 0.1 mm, or zero, for at least 10 seconds or two respiratory cycles.

[0212] By measuring the gain of the upper airway muscle response during an episode, its obstructive nature, i.e., significant respiratory effort, can be diagnosed, as opposed to a central character in which there is no respiratory effort or the respiratory effort falls below a level considered normal. The analysis allows apneas and hypopneas to be characterized as obstructive or central. Normal levels of respiratory effort are pre-measured for each stage of sleep during periods of normal breathing during sleep.

[0213] Changing the head position can result in a change in the configuration of the transition events between sleep and wakefulness, with or without a change in the sleep stage. The gain of the muscle response during the onset is calculated by measuring the peak-to-peak amplitude change during the phasic mandibular movement at the respiratory frequency during the onset. This is a measure of the peak-to-peak amplitude difference between the start and end of the onset period during the phasic movement, and can already be calculated from a single respiratory cycle providing a gain value. This change can be minimal, on the order of 0.1 mm, or even less, but can reach 3 cm. This change may be accompanied by a change in the absolute position of the mandible. This means that the mouth is more or less open when the phasic displacement is applied. This change can occur in any direction between horizontal and vertical, taking into account the head position during sleep. Example 9

[0214] In Example 9, reference is made to Table 3, which shows a typical operation of the brain control for detecting respiratory events and non-respiratory movement episodes. As can be seen, to detect obstructive apneas and hypopneas, the analysis unit uses, for example, the median and / or mean values ​​of the first and second streams of measurement signals. To increase the reliability of the analysis, an observation time of at least two respiratory cycles or 10 seconds would be desirable. Obstructive apneas and hypopneas are characterized by large brain control amplitudes at respiratory rates that can be repeated periodically or aperiodically, and finally, during brain activation, large mandibular movements. In particular, the distribution of amplitude values ​​of mandibular movements in the streams of interest is analyzed.

[0215] To detect respiratory effort related arousals (RERA), the analysis unit proceeds as before. To detect central apnoeas and hypopnoeas, the observation period will also be at least two respiratory cycles or 10 seconds. Example 10

[0216] In Example 10, reference is made to FIGS.

[0217] Bruxism may be detected using the median, mean, maximum or other statistics of mandibular rotation and its acceleration over an observation period of, for example, 30 seconds.

[0218] Brain control after brain activation with or without head position change Stable at low amplitudes; · High or "enhancing" centrally enhanced; mouth closed, if the latter is obstructive, the onset is corrected; Low or "reduced" central and enhanced; open-mouthed, onset-imposed, obstructive; Low or "reduced" central, decreased; imposed onset, central; High or "heightening" central augmentation when the lateral pterygoid muscle is recruited.

[0219] If the head position remains the same but the level of respiratory control changes, the position of the mandible and its changes continue to provide information about the level of respiratory control. Figure 7 shows signals indicative of cortical cerebral activation, plotting the mandible movement measured by the accelerometer and gyroscope. From left to right in Figure 7, what is seen first are some vibrations indicating a movement of the mandible at a certain frequency. This movement of the mandible is caused by breathing with some effort. The subject must make an effort to make the air pass through the upper airway. This can be seen in the amplitude of the signal from the gyroscope. In particular, reference 1 shows micro-arousals caused by cortical activation. What is seen next are strong vibrations indicating a movement of larger amplitude following brain activation. What is seen next is an increase in the level of the signal. This indicates that the mouth is closed and the mandible has risen by a few tenths of a millimeter. If the mandible rises and the respiratory frequency of its movement is still detected with an amplitude larger than the normal value, it can be assumed that there is a persistent obstructive episode as observed here. If the amplitude of the movements decreases, respiratory control is not elevated above normal and respiratory effort is said to be normalized. Reference 2 shows that microarousals caused by subcortical activation produce lower amplitude signals than cortical activation.

[0220] The set of results obtained after processing of the signals by the analysis unit may, for example, be presented in the following way: · Hypnotic: evolution of sleep stages and moments of wakefulness / sleep transition during the recording; Start and end times of recording, time spent in bed and / or time lying down; Total sleep time; various efficacy indicators; Sleep fragmentation; e.g. number and indicators of microarousals and arousals (activations), number and indicators of wake / sleep transition changes; Number and index of respiratory and non-respiratory movement episodes; For example, repeated central respiratory episodes accompanied by periodic, regular, gradually increasing and decreasing fluctuations in cerebral control amplitude, the duration of which is measured to be more than 40 seconds, may suggest that a periodic type of breathing is developing, possibly in the context of heart failure; Events of a cyclic nature may be obstructive episodes or may be cyclic recurring obstructive episodes (e.g., high loop gain or high arousal); Recurrent subcortical activation independent of respiratory effort suggests that it is associated with limb movement.

[0221] The position of the head influences the frequency and nature of the onset of respiratory and non-respiratory movements occurring during sleep. Changes in head position during sleep always occur simultaneously with activation from the brain. During the latter, the head finds a new position and the mandible, which during this change moved with large amplitudes by multiple repeated movements, finds a new position where it is then subjected again to respiratory drive, the amplitude of which is a measure of the level of central control. Thus, as shown in Figure 7, there is an association of onsets. Thus, active respiratory control with central activation, possible changes in the position of the head and, with it, possible changes in the position of the mandible, as well as changes in the amplitude of the respiratory movements of the mandible, are integrated in the brain. The relationship between activity and brain activation was investigated. If the level of control activity changes, for example through a change in the peak-to-peak amplitude of the respiratory movements of the mandible, this is first of all the result of a change in the central activation and control state in the brainstem. Moreover, while respiratory activity is captured by said gyroscope during the rotational movements of the mandible, central activation is captured by said accelerometer during the linear movements of the head.

[0222] Head position changes and the associated cerebral activation determine the risk of an episode occurring, whether a respiratory or non-respiratory motor episode, by altering the level of control and therefore the type of episode. Head position changes, which are necessarily accompanied by cerebral activation, can modify the airflow conditions in the airways, in particular by altering the muscular retention of the upper airway caliber, which may expose the airways to mechanical crushing forces due to the new orientation assumed by the head.

[0223] The movement and repositioning of the mandible during cortical or subcortical activation during onset can be explained as follows. (1) The mandible is passive, or falls in degeneration of activation without tonic and / or phasic support of musculature, while central motor control is inhibited for a period of several respiratory cycles. After mouth closure, the mandible relaxes its position by passive mouth opening, i.e., the mandible is no longer supported, as the tonic portion of musculature considered to support it is lost over a variable distance, but with a significant gradient (>1 / 10 mm / s). Measurement of the distance between the lowest point recorded before mouth closure and the subsequent change in gradient is a marker of passive collapsibility of the pharynx as it follows from activation with loss of central nervous control. This situation may continue for a time equivalent to several (up to 5) respiratory cycles. This relaxation does not occur because central control (tonic part continuity) of the musculature controlling the position of the mandible is not lost, and the mouth may remain closed or substantially closed. (2) The mandible then shows a muscle response gain in the form of phasic and / or tonic, which repositions the mandible at the frequency of respiratory control during the onset, before a new activation is induced. This is followed by a resumption of muscle activity controlling the position and movement of the mandible. This resumption of muscle activity may be manifested by a change in the slope, which represents a new position with or without respiratory movement, i.e. with or without a phasic component, i.e. with at least one peak-to-peak amplitude measurable above the background noise of the measurement (>0.05 mm). This latter movement indicates a resumption of respiratory movement, i.e. a change in the respiratory frequency, and therefore a respiratory effort that allows to qualify the obstructive onset as central or mixed, according to the rules of evaluation. The central movement allows to identify whether the degree of mouth opening is stable, increasing or decreasing, while the peak-to-peak amplitude of the respiratory movement reflects the degree of the current effort. (3) The central or amplitude point is the lowest, from which the movement of closing the mouth is executed. The first movement determined by activation is similar to the threshold of arousal. This movement can be downwards, for example in REM, or when the mouth does not open because the respiratory effort is exerted by the activity of the lateral pterygoid and masseter muscles, which hold the mouth forward and high, this movement is proof of activation. The latter can be of cortical or subcortical, or subcortical and then cortical series, or if the mandible did not open much during the onset, it can suddenly open during activation, probably due to the activity of the lateral pterygoid, but in most cases, the mouth was opened during the onset and is therefore brutally closed by activation. (4) As shown in Figure 4, the mandibular position point follows at the maximum distance during activation from the arousal point. The distance separating them is a measure of the amplitude of mandibular movement during activation. Its value is measured and compared with the level of respiratory effort deployed during onset before activation via the change in the amplitude of respiratory movement from the beginning of the resumption of effort to the arousal point. The ratio of these values ​​is a measure of the degree of mandibular loop gain. Example 11

[0224] In Example 11, reference is made to FIG. 8. FIG. 8 shows an example of a first measurement signal stream F1 (measured by the accelerometer) and a second measurement signal stream F2 (measured by the gyroscope) in a situation where the subject suffers from obstructive apnea. In this figure, F5n indicates the nasal flow and F5th indicates the oronasal thermal flow. As observed there, during the period T1-T2 and following the apnea shown by control 1, the signals are not stable. At the beginning of the period, it can be seen that the signal provided by said gyroscope has a smaller amplitude than at the end of the episode. During the episode, central control is strengthened, since it is necessary to fight the obstruction that causes the apnea or hypopnea. During the same period T1-T2, it can be seen that the accelerometer (control 2) and the gyroscope (control 3) indicate a respiratory effort followed by brain activation (control 4).

[0225] As shown in the signal analysis, in the case of obstructive apnea between T1 and T2, the movement of the mandible at the respiratory frequency is observed on the accelerometer (F1), increasing in amplitude from peak to peak as the result of the mandible position increasingly descends from one respiratory cycle to another (A ). At the same time (C), it can be seen that the angular velocity of the rotational respiratory movement, whose amplitude increases, indicates that the effort itself is increasing. Note the effect of activation on the movement of the mandible measured by said accelerometer at the height of the letter B. Said activation causes an upward movement of the mandible, resulting in the closing of the mouth. On this occasion, the mandible takes a new position. At the level of the letter D on the gyroscope, it can be seen that this movement of closing the mouth is not purely rotational. The change in the signal state at the height of the letters B and D coincides with the resumption of ventilation upon brain activation (microarousal). Example 12

[0226] In Example 12, reference is made to FIG. 9. FIG. 9 shows, with arrow 1, an example of a first measurement signal stream F1 and a second measurement signal stream F2 in a situation where the subject suffers from obstructive hypopnea. Arrow 0 indicates the state of arousal. FIG. 9 also shows a sixth stream F6 captured by an audio sensor indicating the presence of snoring, a seventh stream F7 detected by the electromyogram of the jaw, and an eighth stream F8 detected by the electroencephalogram. A series of mandibular movements (R) of increasing amplitude are observed for several seconds, each time indicating a cortical or subcortical activation. These movements are accompanied by changes in the peaks of streams F6, F7, and F8. Indeed, the electromyogram and electroencephalogram signals clearly show that in this case there is brain activation. The obstructive hypopnea is indicated by arrows 2 and 3, with arrow 2 indicating an effort and opening of the mouth and arrow 3 indicating an effort and rotation of the mandible. The hypopnea is followed by activation in the form of microarousals indicated by arrow 4. High values ​​of respiratory mandibular movement during micro-arousals reflect high respiratory effort, which is further accentuated by snoring. Thus, in the stream F1 from the accelerometer and the stream F2 from the gyroscope, we see that snoring causes the mandibular rotation with the mouth open. Thus, FIG. 9 shows that brain activity may be recorded by the accelerometer and gyroscope measuring mandibular movement during periods of effort at respiratory frequencies, as well as during brain activation, but at frequencies that are no longer typically respiratory. Also in FIG. 9, the number 0 indicates the subject's awake state. Example 13

[0227] In Example 13, reference is made to Figure 10. Figure 10 shows an example of a first measurement signal stream F1 and a second measurement signal stream F2 when a subject suffers from mixed apnea. As in Figure 8, Figure 10 shows an increase in the angular velocity of the mandible at a frequency corresponding to the respiratory frequency. Number 1 shows a loss of respiratory flow closely associated with a loss of control and effort shown by number 2, followed by a recovery of cerebral control and effort shown by number 3. Example 14

[0228] In Example 14, reference is made to Figure 11. Figure 11 shows an example of a first measurement signal stream F1 and a second measurement signal stream F2 when a subject suffers from central apnea. Peaks F indicate head and mandible movements during rebreathing. It can also be seen that between peaks F there is, so to speak, no mandible movement. Number 1 indicates a lack of respiratory flow which goes hand in hand with a lack of effort, indicated by number 2, and with an activation and resumption of effort, indicated by number 3. Example 15

[0229] In Example 15, reference is made to figures 12 and 13. Figure 12 shows an example of a first measurement signal stream F1 and a second measurement signal stream F2 in the case where the subject suffers from a temporary loss of all control of brain origin, which is characteristic of central hypopnea. The loss is characterized by a passive opening of the mouth due to the lack of muscular support. Thus, as can be seen in streams F1 and F2, the signals show no activity between the peaks. On the other hand, at the moment of the peak, a large amplitude of mandibular movement is observed. Towards the end of the peak, a movement corresponding to a non-respiratory frequency is seen, which is the result of brain activation, which then causes a micro-arousal. Number 1 shows a period of hypopnea during which a reduction in flow rate is clearly visible in stream F5 from the thermistor. Numbers 2 and 3 show the loss of mandibular movement in streams F1 and F2 during a period of central hypopnea. Figure 13 shows an example of a first measurement signal stream F1 and a second measurement signal stream F2 in the case where the subject experiences a prolonged respiratory effort that ends with brain activation. It can be seen that the signal from the accelerometer F1 shows a large movement of the head and mandible, at the position indicated by H. Then, while the stream F2 remains practically constant, the level drops in F1 from said accelerometer. This indicates that in any case there is a movement of the mandible, which slowly goes down. This is followed by a high peak I, which is the result of a change in the position of the head during the activation that ends the active period. The number 1 indicates this long period of effort, marked by snoring. It can be seen that the effort increases over time, as indicated by the number 2. The effort ends with an activation of the brain that results in a movement of the head and mandible, indicated by the letter I, as indicated by the number 3.

[0230] The analysis unit has memory models of these various signals that are the result of processing them using artificial intelligence as described above, and uses these results to process the stream and generate a report on the analysis of the results.

[0231] It has been found that the accelerometer is particularly suitable for measuring head movements, whereas the gyroscope, which measures rotational movements, is particularly suitable for measuring rotational movements of the mandible. Thus, brain activations that lead to rotation of the mandible without changing the position of the head can be detected by the gyroscope. On the other hand, IMM type movements are detected by the accelerometer, especially when the head moves in this case. RMM type movements are detected by the gyroscope, which is very sensitive to it. Example 16

[0232] In the further example, reference is made to an exemplary procedure for feature extraction, data processing, and data description used in the methods and apparatus provided herein, which is shown diagrammatically in FIG.

[0233] Specifically, feature extraction, data processing, and description were performed in the R statistical programming language (8), and machine learning experiments were performed using the sci-kit learn and SHAP packages in the Python language.

[0234] Twenty-three different features were extracted from the raw signal of mandibular movement for each episode, or for each 10-s period of normal breathing. These features included: central tendency of MM amplitude (mean, median, mode); MM distribution (raw or envelope signal): skewness, kurtosis, IQR, 25th, 75th, and 90th percentiles; extreme values: minimum, maximum, 5th, and 95th percentiles of MM amplitude; trend of variation: linear trend and coefficients of tensor product-based spline coefficients (S1, 2, 3, 4) from a generalized additive model to evaluate MM in function of time; and duration of each episode.

[0235] The impact of different features on the model classification into central hypopnea, normal sleep, and obstructive hypopnea can be illustrated by the SHAP score, which measures the average marginal contribution across all possible associations with other features to classify the three target labels. The higher the SHAP score, the more significant contribution the feature provides. Lundberg's Shapley additive explanation (SHAP) method integrates Shapley's score and local interpretation approaches (Marco Tulio Ribeiro, Sameer Singh, Carlos Guestrin. "Why should i trust you? Explaining the predictions of any classifier". In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM. 2016, 1135-1144.) in cooperative game theory (1953) (Lloyd S Shapley. "A value for n-person games". In: Contributions to the Theory of Games 2.28 (1953), 307-317.) to provide the best solution to date for explaining black-box models. SHAP theory considers input features as "players" in cooperative games whose "payout" correctly predicts the target label (i.e., central or obstructive hypopnea). The SHAP algorithm allows us to combine each feature value with other features in a random order to form a coalition and assign a payout (the SHAP score) to each feature value depending on its contribution to the overall prediction. The SHAP score is the result of averaging the change in prediction that the coalition gets when a new feature is added. Essentially, the SHAP score of a feature value is the average marginal contribution of that feature value across all possible coalitions of a particular prediction.

[0236] Specifically, the features are extracted as follows. 1. Load a raw MM data series, e.g., with a sampling rate of 10 or 25 Hz, with the series having a significant duration, e.g., 30 minutes to 8 hours; 2. Characterize the timestamps of obstructive and central hypopnea onset; 3. For each timestamp ti, perform the following steps: 3.a. Determine whether the TI is the beginning of an obstructive or central hypopnea episode. 3.b. If ti is the beginning of an obstructive or central hypopnea episode: - assign ti to (t_begin), then search for the end (t_end), and - index t_begin and t_end and extract the raw data series into a temporary holder named "Onset E"; 4. For each symptom E, perform the following steps: 4.a. Calculate the onset period dt = (t_end-t_begin), 4.b. Determine the distribution of the measured parameters during the episode; - minimum, maximum, mean, median, mode, percentiles 5, 25, 75, 90, and 95, skewness, kurtosis, and IQR; - Fitting a GAM nonlinear model to estimate the MM amplitude and / or location by a spline function of time t and extracting the coefficients of the spline function; - Fit simple linear models and extract the intercept and linear slope; - Concatenate all features and labels by matching the measurement data with mandibular movement classes.

[0237] After feature extraction, the extracted features and corresponding target labels were integrated into a tabular dataset.

[0238] Exploratory data visualization, one-way ANOVA, and pairwise Student's t-tests with Bonferroni correction were then performed to classify the mandibular movement characteristics into three groups (normal breathing, obstructive, and central hypopnea). The significance level was set at a very strict criterion for null hypothesis testing (p = 0.001) (10).

[0239] For model development, the data were randomly split into two subsets: a larger set (70%) for model development and a smaller set (30%) for model validation. Because the original training set was not balanced between central (minority class) and obstructive hypopnea (majority class), a synthetic minority over-sampling technique (known as SMOTE) was applied on the training set before model development.

[0240] A multi-class classification rule was created to classify the three groups using the 23 input features, which consisted of a random forest algorithm combining 500 different decision trees, each built with a random subset of the five features.

[0241] The contents of the random forest model were then analyzed to assess the importance of each feature and the possible associations that contributed to the classification (potential combinations between them to distinguish between obstructive and central hypopnea).To assess the contribution of each feature to the prediction, Lundberg's Shapley additive explanation (SHAP) method, which is well known in the art, is employed.

[0242] These methods allow, inter alia, the detection of obstructive hypopnea and central hypopnea. Example 17

[0243] In this further example, reference is made to Figures 16 and 17, which show the analysis of mandibular movement data captured by a magnetic sensor. As such, the data analysis is similar to that of mandibular movement data acquired by an accelerometer and / or gyroscope in addition to a magnetic sensor.

[0244] Figure 16 shows the 18 most important MM signal features derived from magnetometer measurements, ranked by their global impact on the model's predictions. The bars show the average SHAP score for each feature stratified by the three target labels: central hypopnea (dark grey), normal (light grey), and obstructive hypopnea (grey). The SHAP score measures the average marginal contribution across all possible associations with other features to classify the three target labels. The higher the SHAP score, the more important the contribution that the feature may provide.

[0245] Figure 17 shows the interpretation of onset based on extracted features and SHAP score. Specifically, Figure 17 shows the SHAP score scale and the probability of the target label. Figure 17 is composed of two general regions: region a) and region b). Region a) contains extracted features that support the prediction of the target label, and region b) contains extracted features that point in a direction away from the target label. Example 18

[0246] In this further example, reference is made to Fig. 18, which shows an exemplary method for determining sleep stages from mandibular movement data captured by a gyroscope and an accelerometer. The steps described below correspond to the reference numbers in Fig. 18.

[0247] Specifically, these steps are: (1) Using the system of the present invention including a gyroscope and an accelerometer, record the mandibular movement during the subject's sleep. The resulting data pack includes 6-channel raw signals obtained by the 3-axis accelerometer and gyroscope sensor. The raw data may further include recordings from other devices suitable for measuring sleep stages, such as EEG, EOG and EMG signals for sleep stages, 6-channel MM signals obtained by the 3-axis accelerometer and gyroscope sensor.

[0248] (2) After passing through a pre-processing and feature generation module, the raw data is split into successive 30 second long windows. The pre-processing consists of producing time series sampled at 0.1 Hz and 0.034 Hz ​​(30 second sliding window) from the sleep score sequence and sensor and PSG acquired time series. The pre-processing is performed in two steps: the series or sequence is split and then a feature extraction function is applied to each window. Handcrafted feature extraction can be used as input data for machine learning experiments. For example, the feature generation module extracted a set of 1728 features from the six channels of the MM activity signal using sliding windows centered on every 30 seconds. The extracted features include signal energy in the low frequency band (0-0.1 Hz), high frequency band (>0.3 Hz), or respiratory frequency band (0.2-0.3 Hz), exponential moving averages with several half-lives, entropy of energy in several frequency bands, statistical features applied to the above features: central tendency (mean, median), extreme values ​​(minimum, maximum), quartiles, standard deviation, and normal standardized values ​​of all the above features.

[0249] (3) The extracted feature set is fed into a machine learning classifier that produces a soft prediction score (i.e., probability) and a binary output for each target label according to the specific classification task. The automated sleep stage task was approached at three levels of complexity. The task targets are the three basic sleep stages: wake, non-REM (including N1, N2, and N3), and REM. Feature selection and hyperparameter tuning were performed using cross-validation, where the input data was randomly split into folds at the participant level. The final model was trained on the full training set using only the most relevant features and optimized hyperparameter values. Due to the imbalanced ratio between the target labels, the training data was balanced with a synthetic minority oversampling technique (SMOTE) before each training session. Machine Learning Algorithm: The Extreme Gradient Boosting (XGB) classifier is adopted as the core algorithm for all three classification tasks. The XGB classifier is optimized during the training process by minimizing a regularized objective function that combines a convex loss function (based on the difference between predicted and target outputs) and a penalty term for the model complexity. Model training: The learning goal is set to multi-class classification, aiming to classify three target labels according to a specific task. The training implied a dropout-multiple additive regression tree (DART) booster and a histogram-optimized approximate greedy tree construction algorithm. Log-loss was chosen as the evaluation metric (thus optimizing the balanced accuracy between the three target classes). To prevent overfitting, the learning rate (eta, or step size reduction) parameter is set to 0.01. This shrinks the feature weights, making the boosting process more conservative. The output of the model is a softmax function that produces a probability score for each target label, and the final decision (to assign only one label per 30 seconds) is reached by applying the argument function that gives the maximum value to these three probability scores.

[0250] (4) Depending on the period-by-period agreement between the model predictions and the reference PSG scoring on an unseen validation dataset, the most satisfactory solution is adopted for implementation.Further quantitative evaluation was performed to verify whether the selected algorithms could provide reliable estimations of sleep quality scores such as TST, sleep efficiency, and REM ratio.The model selection was based on the following criteria: Per-class agreement assessment: The normalized confusion matrix allows us to assess the per-class performance of a model for a given multiclass classification task. The rows are the truths derived from manual PSG scoring, and the columns show the results of the automated algorithmic scoring. The diagonal cells of the confusion matrix show the true positive rates per class. Precision (or positive predictive value), defined as true positives / (true positives + false positives), measures the ability of the model to correctly identify positive cases. Recall (also called sensitivity, hit rate, or true positive rate) indicates the usefulness of the model (defined as the proportion of correct classifications among all subject examples), where recall = true positive predictions / all positive examples; The F1 score is a composite metric defined as the harmonic mean of per-class recall and precision. 2 × (precision rate × recall rate) / (precision rate + recall rate) The F1 score has an intuitive meaning and indicates the accuracy of the model (how many periods are correctly classified) and how robust it is (how low the misclassification rate is). Since the real data shows an unbalanced ratio between sleep stages and all labels are equally important, a classifier that obtains an equally high F1 score for all classes is employed. Global and per-period agreement assessment metrics: Balanced Accuracy (BAC) measures the average true positive and true negative rates between target classes. Cohen's Kappa coefficient measures the strength of agreement between the model's classification and the actual observations (manual PSG scoring). It can be interpreted as six levels of agreement strength: <0: weak, 0-2: slight, 0.2-0.4: fair, 0.41-0.6: moderate, 0.61-0.8: fair, 0.81-1: almost perfect.

[0251] (5) The prediction data from the selected (three-class task) model is passed through an interpretation module. The first submodule (Calculate Sleep Score) converts the predicted sleep stage sequence into a quantitative score. The definitions of these quantitative scores are given in the table below: JPEG2025020260000003.jpg171140

[0252] (6) Creating hypnosis scenarios: A customized function converts a sequence of discretely encoded labels (e.g., 2=awake, 1=REM sleep, 0=non-REM sleep) into hypnosis scenarios. The graph shows step lines representing the values ​​of the individual sleep stages as a function of time, which simulates traditional hypnosis scenarios obtained from manual PSG scoring. Example 19

[0253] Example 19 continues the experiment of Example 18. Specifically, the method shown in Example 18 was performed on a group of 96 participants who were randomly assigned to a training subset (n=68, 70%) and a validation subset (n=28, 30%). Both subsets represented healthy individuals aged 18-53 years.

[0254] The mandibular movements were recorded during the subjects' sleep using the system of the present invention, which includes a gyroscope and an accelerometer. The resulting data pack contained 6 channels of raw signals obtained with the above-mentioned 3-axis accelerometer and gyroscope sensors. The raw signals were used to develop the automatic sleep stage model. Reference data was also recorded using devices suitable for determining sleep stages, such as EEG, EOG, EMG, etc. The latter data was used to determine the accuracy of the applied model.

[0255] Instead of using deep learning models, we followed a traditional framework, which means handcrafted feature extraction and structured data-driven algorithms. Handcrafted feature extraction allows for greater control and understanding of the input data compared to black-box models like convolutional neural networks. For classification tasks, we employed XGBoost. This algorithm offers several advantages over traditional methods (LDA, SVM, RF), including being computationally and resource efficient, allowing for very fast training and execution speeds.

[0256] Subject subsets: Polysomnography (PSG) profiles from a group of 96 participants showed normal sleep activity in both subset groups, with median sleep efficiency of 89.4% and 87.3%. Within each set, the data structure also shows an imbalance in the ratios between the three sleep stages: except for the mostly normal wakefulness label, non-REM sleep predominated over REM sleep in both groups (92.3 vs. 7.7 in the training set and 79.9 vs. 20.1 in the validation set). This suggests that data balancing techniques are needed in model development and that performance metrics should be interpreted with care during model validation.

[0257] Coring 3 Classes: The current model aims to classify awake (no sleep), non-REM sleep, and REM sleep. The model obtains balanced accuracy among the three classes (82.9%, 74.9%, and 82.5% for awake, non-REM, and REM sleep, respectively). The model also has a fair strength of agreement (kappa = 0.71). With an F1 score of 0.86, it is optimal for detecting awake periods. Based on the distribution of awake, non-REM, and REM instances in the model, it turns out that identifying awake is easier than distinguishing non-REM from REM, since awake instances are well clustered and clearly separated from other instances, whereas most REM labels are more variable and mixed with other non-REM or awake points. As suggested by this pattern, nonlinear algorithms such as random forest, XGboost, and deep neural networks may be considered to successfully separate the three classes.

[0258] Concordance analysis of sleep quality indices: The three-class task sleep stage algorithm can automatically classify every 30 seconds as awake, non-REM, or REM. The output was then transformed by a second algorithm to provide an estimate of the sleep quality indices. These indices can be classified into three main categories: a) time-based indices, which measure the cumulative time (in minutes) during sleep (TST) or during a particular sleep stage such as awake, REM, or non-REM; b) ratio-based indices, which are estimated as a percentage of a particular sleep stage (REM, non-REM) over the entire sleep period; c) latency-based indices, which measure the time elapsed from the start of recording to sleep onset (sleep onset latency) or from sleep onset to the first REM period (REM latency).

[0259] The automated sleep stage algorithm quantitative scores were measured according to the table shown in Example 18. The difference between the standard scores of the PSG profile and the automated sleep stage algorithm quantitative scores is shown in the table below: JPEG2025020260000004.jpg20781

[0260] These data indicate that the 3-class-based scoring algorithm can measure total sleep time with acceptable accuracy (median difference was only -7.15 minutes; 97.5% CI -20.34 to +4.38) compared with the reference method (manual PSG scoring). This agreement was also good for measuring sleep efficiency (median difference was -1.29%; -3.03 to +0.01).

[0261] Conclusion: We investigated the feasibility of using the mandibular movements recorded during the sleep of subjects using the system of the present invention including a gyroscope and an accelerometer. The results show that automated sleep stage detection based on data measured with a gyroscope configured to measure rotational movements provides better performance at all three resolution levels (in the case of three-class scoring) compared to the state of the art system including only an accelerometer.

[0262] JPEG2025020260000005.jpg215133JPEG2025020260000006.jpg21578

[0263] JPEG2025020260000007.jpg21584

[0264] JPEG2025020260000008.jpg214135JPEG2025020260000009.jpg21423

Claims

1. A system for detecting sleep events of a subject having a head and a mandible, the system including a sensing unit, a data analysis unit, and a data link configured to be attached to the mandible of the subject; wherein the sensing unit includes a gyroscope configured to measure rotational movement of the mandible of the subject over a period of time, and an accelerometer configured to measure an acceleration indicating movement and / or position of the head and / or mandible of the subject over a period of time; the data link is configured to transmit data of the rotational movement measured by the gyroscope and data of the acceleration measured by the accelerometer to the data analysis unit; wherein the data analysis unit includes a memory unit configured to store N mandibular movement classes, where N is an integer greater than 1, and one or more of the N mandibular movement classes indicate sleep events related to sleep disorders; wherein each jth (1 < j < N) mandibular movement class includes a jth set of rotation values, and each jth set of rotation values indicates at least one speed, speed change, frequency, and / or amplitude of mandibular rotation associated with the jth class, and / or each jth (1 < j < N) mandibular movement class includes a jth set of acceleration values, and each jth set of acceleration values indicates at least one of mandibular movement and / or head movement associated with the jth class; wherein the data analysis unit includes a sampling element configured to sample data of the rotational movement and data of the acceleration measured during the same period, thereby obtaining sampled data of the rotational movement and the acceleration; wherein the data analysis unit derives a plurality of measured rotation values and acceleration values from the sampled data of the rotational movement and the acceleration; compares the measured rotation values and acceleration values with one or more of the N mandibular movement classes indicating the sleep events; generates diagnostic information for diagnosing a sleep disorder related to the sleep event based on one or more of the N mandibular movement classes indicating the sleep event that have been compared; and is configured to create a report including the diagnostic information for diagnosing a sleep disorder related to the detected sleep event, system.

2. The system according to claim 1, further comprising a magnetometer, the magnetometer being adapted to measure magnetic field data, and variations in the magnetic field data indicating movement of the subject's head and / or mandible, wherein the data link is configured to transmit the magnetic field data measured from the magnetometer to the data analysis unit; wherein each j-th (1 < j < N) mandibular movement class includes a j-th set of values of magnetic field data, and each j-th set of values of magnetic field data indicates at least one velocity or change in velocity of the movement of the mandible or head associated with the j-th class; wherein the sampling element is further configured to sample the magnetic field data measured over a certain period of time, thereby obtaining the sampled magnetic field data; wherein the data analysis unit is further configured to derive a plurality of measured magnetic field values from the sampled magnetic field data; and wherein the data analysis unit is further configured to match the measured magnetic field values with one or more of the N mandibular movement classes indicative of the sleep event, a system.

3. One or more of the N mandibular movement classes indicate a change in head position; the data analysis unit is configured to identify the movement of the subject's head based on the data of the rotational movement and acceleration, and further configured to identify the movement of the head from the movement of the subject's mandible, the system according to claim 1.

4. One or more of the N mandibular movement classes are characterized by a predetermined frequency range consisting of the frequency indicating the respiration of the subject, and the at least one predetermined frequency range consists of a frequency of 0.15 Hz to 0.60 Hz, the system according to claim 1.

5. One or more of the N mandibular movement classes indicate a respiratory disease event associated with a respiratory disease; the respiratory disease includes obstructive apnea, mixed apnea, obstructive hypopnea, respiratory effort related arousal, central apnea, and / or central hypopnea, the system according to claim 1.

6. The data analysis unit analyzes and observes one or more of the changes in the mandibular position, head position, peak-to-peak amplitude of mandibular movement, dispersion of peak-to-peak amplitude of mandibular movement, frequency of mandibular movement, and / or dispersion of frequency of mandibular movement over a certain period of time; and If one or more of the peak - to - peak amplitudes of the mandibular movement and / or the change in the frequency of the mandibular movement are observed, detect the presence of a sleep event The system according to claim 1, configured for this purpose **Claim 7** One or more of the N mandibular movement classes indicate bruxism events related to bruxism disorder, where the measured rotational movement data shows a mandibular movement amplitude of at least 1 mm, and when the movement is phasic, a frequency established in the range of 0.5 - 5 Hz during at least 3 respiratory cycles, or a mandibular movement amplitude of more than 1 mm continuously and rigidly for at least 2 seconds. The system according to claim 1 **Claim 8** The system according to claim 1, wherein the report containing the diagnostic information includes the hourly incidence rate and / or the cumulative duration of the sleep event **Claim 9** A method for detecting a sleep event of a subject having a head and a mandible, comprising Measuring the rotational movement of the mandible of the subject using a gyroscope and measuring the acceleration indicating the movement and / or position of the head and / or mandible of the subject over a certain period using an accelerometer, where the gyroscope and the accelerometer are arranged on the mandible of the subject; Receiving, by a data analysis unit, the data of the rotational movement measured from the gyroscope and the data of the acceleration measured from the accelerometer via a data link; Storing, in a memory unit included in the data analysis unit, N mandibular movement classes, where N is an integer greater than 1, and one or more of the N mandibular movement classes indicate sleep events related to sleep disorders; where each j - th (1 < j < N) mandibular movement class includes a j - th set of rotation values, and each j - th set of rotation values indicates at least one speed, speed change, frequency, or amplitude of the mandibular rotation associated with the j - th class, and / or each j - th (1 < j < N) mandibular movement class includes a j - th set of acceleration values, and each j - th set of acceleration values indicates at least one of the mandibular movement and / or the head movement associated with the j - th class; Sampling, using a sampling element included in the data analysis unit, the data of the rotational movement and the measured acceleration data during the same period, thereby obtaining the sampled rotational movement and acceleration data Deriving a plurality of measured rotation and acceleration values from the sampled rotational movement and acceleration data using the data analysis unit; Collating the measured rotation and acceleration values with one or more of the N mandibular movement classes indicative of the sleep event using the data analysis unit; Generating diagnostic information for diagnosing a sleep disorder associated with the sleep event based on one or more of the N mandibular movement classes indicative of the collated sleep event; and Creating a report including diagnostic information for diagnosing a sleep disorder associated with the detected sleep event A method comprising.

10. The method according to claim 9, wherein Measuring magnetic field data using a magnetometer, wherein fluctuations in the magnetic field data indicate movement of the subject's head and / or mandible; Transmitting the measured magnetic field data from the magnetometer to the data analysis unit using the data link; Here, each j-th (1 < j < N) mandibular movement class includes a j-th set of values of magnetic field data; each j-th set of values of magnetic field data indicates at least one velocity or change in velocity of mandibular or head movement associated with the j-th class; Sampling the magnetic field data measured over a certain period using a sampling element included in the data analysis unit, thereby obtaining sampled magnetic field data; Deriving a plurality of measured magnetic field values from the sampled magnetic field data using the data analysis unit; and Collating the measured magnetic field values with one or more of the N mandibular movement classes indicative of the sleep event using the data analysis unit A method further comprising.

11. One or more of the N mandibular movement classes indicate a change in head position; the method further comprising identifying movement of the subject's head based on the rotational movement and acceleration data and differentiating movement of the subject's head from movement of the subject's mandible, the method according to claim 9.

12. One or more of the N mandibular movement classes are characterized by a predetermined frequency range consisting of the frequency of the subject's respiration, the at least one predetermined frequency range consisting of frequencies from 0.15 Hz to 0.60 Hz, the method according to claim 9.

13. 10. The method of claim 9, wherein one or more of the N mandibular movement classes are indicative of a respiratory disease event associated with a respiratory disease; the respiratory disease comprising obstructive apnea, mixed apnea, obstructive hypopnea, respiratory effort related arousals, central apnea, and / or central hypopnea.

14. 10. The method of claim 9, wherein one or more of the N mandibular movement classes are indicative of a bruxism event associated with a bruxism disorder, and wherein the measured rotational movement data exhibits a mandibular movement amplitude of at least 1 mm, if the movement is phasic, at a frequency established in the range of 0.5-5 Hz during at least three respiratory cycles, or a mandibular movement amplitude of more than 1 mm sustained and tetanic for at least 2 seconds.

15. 10. The method of claim 9, Analyzing and observing changes in one or more of mandibular position, head position, peak-to-peak amplitude of mandibular movement, variance of peak-to-peak amplitude of mandibular movement, frequency of mandibular movement, and / or variance of frequency of mandibular movement over a period of time; and detecting the presence of a sleep event when a change in the peak-to-peak amplitude of mandibular movement and / or the frequency of mandibular movement is observed. The method further comprising:

Images