SYSTEM AND METHOD FOR AUTOMATIC DETECTION OF HAZARDOUS SITUATIONS

Description

[0001] The invention relates to a system and a method for automatically detecting hazardous situations within a monitoring area for objects in a medical environment. Furthermore, the invention relates to a program with program code for implementing the method according to the invention.

[0002] In the medical setting, it is a well-known problem that confused patients endanger their own treatment by, for example, removing bandages, withdrawing catheters, or interacting with complex medical devices. To prevent such patient behavior in the patient's own best interest, a camera-based monitoring system for monitoring bed spaces within a hospital is known to be used.

[0003] US 10,090,068 describes a method for monitoring, in a clinical setting, whether a patient's hand has entered a predetermined zone within a monitoring area for a predetermined minimum period. The predetermined zone can be, for example, in the area of ​​a catheter or in the area of ​​a patient wound. The described method thus makes it possible to detect the patient's hand entering a dangerous area.

[0004] CN 106 233 349 A describes a method for monitoring a patient within a medical monitoring area by means of a monitoring system having a 3D camera device, comprising the following steps: generating a point cloud of the monitoring area by means of the monitoring system, evaluating the point cloud to detect predefined objects, in particular persons, determining the location of at least one detected object in the monitoring area, comparing the determined location of the at least one detected object with at least one predefined value for the location of the detected object.

[0005] US 10 043 360 B1 describes a system to deter organized retail crime.

[0006] WO 2019 / 011 769 A1 describes a method, a device, and a computer program for acquiring optical image data of a patient environment and for detecting a patient check. The method comprises detecting a patient based on the image data, detecting at least one further person based on the image data, and determining at least one geometric relationship between the patient and the at least one further person. Determining the geometric relationship includes determining an orientation or a viewing direction of the at least one further person relative to the patient. The method further comprises detecting the patient check based on the geometric relationship.

[0007] The object of the present invention is to enable an improved automated detection of hazardous situations for objects in the medical environment, in particular an improved avoidance of the issuance of false alarms and in particular an early detection of relevant hazardous situations.

[0008] According to the invention, to solve this problem, a system for automatically detecting hazardous situations within a monitoring area for objects in a medical environment is proposed, comprising a sensor unit, a processor unit and an output unit.

[0009] The sensor unit has a number of optical sensors and is configured to determine a sensor signal and output it in real time. Furthermore, the number of optical sensors can be arranged in the medical environment and configured such that the sensor signal comprises sensor data indicating a sequence of three-dimensional representations of the monitoring area over a monitoring period.

[0010] The processor unit is configured to receive the sensor signal and, based on the indexed three-dimensional representations, to determine a number of objects to be protected within the monitoring area. Furthermore, the processor unit is configured to detect, in particular to detect automatically, hazardous objects within the monitoring area based on the indexed three-dimensional representations and to determine a number of object pairs by assigning an object to be protected from the number of objects to be protected to a hazardous object.Furthermore, the processor unit is designed to monitor the positions of the object to be protected and the endangering object of a respective object pair within the monitoring area over the course of the monitoring period and to assign a current risk value to a respective object pair depending on the two current positions and a group of parameters resulting from an evaluation of previous evaluation data, wherein the resulting parameters are at least a distance between the objects of an object pair and a direction of movement of the objects of an object pair relative to one another.Furthermore, the processor unit is further designed to calculate the current hazard values ​​up to the current point in time for each object pair over a temporal evaluation interval, in particular over a predetermined temporal evaluation interval, and to output a corresponding alarm signal if a hazard value of an object pair calculated in this way exceeds a predetermined threshold value.

[0011] The output unit is designed to receive the alarm signal and to trigger an alarm depending on the alarm signal.

[0012] Within the scope of the invention, it was recognized that in order to avoid false alarms and for a particularly early detection of dangers, a current situation within the monitoring area must be qualitatively evaluated.

[0013] For this purpose, the invention defines the objects to be protected and detects hazardous objects within the monitoring area. Furthermore, for such a qualitative evaluation, the system according to the invention takes into account not only current data but also parameters resulting from previous evaluation data.

[0014] In the context of the invention, contact between the object to be protected and the object posing a risk is considered to be a hazardous situation to be avoided, since such contact can have adverse consequences for the patient's health in everyday medical practice, for example due to incorrect operation of a medical device or due to the patient tearing out a tube or catheter.

[0015] The system according to the invention advantageously ensures that hazardous situations within the monitoring area can be quickly identified. In particular, there is no need to wait until a hazardous situation arises, for example, due to the insertion of a patient's hand into a hazardous area, such as near a catheter. Thus, by monitoring the two current positions in combination with the evaluation of previous data, a particularly meaningful hazard value is determined early on. From this, the hazard magnitude characterizing the current hazard for a respective pair of objects is derived.

[0016] By using a predetermined threshold, it is advantageous to specify the user-specific probability of a hazard occurring at which a corresponding alarm signal is issued. If the threshold is predetermined in such a way that the alarm signal is issued even at a low probability of a hazard, this would, in turn, lead to a higher number of false alarms than if the alarm signal were only issued at a moderate probability of a hazard.

[0017] The number of objects to be protected can be determined, for example, automatically, manually or through gesture control.

[0018] The detection of hazardous objects is particularly preferably automated. Such automated detection advantageously allows for rapid recognition of these objects, as there is no need to wait for manual input. Furthermore, it enables the detection of hazardous objects that a user of the system might not have considered posing any risk at all. This automated detection is preferably carried out by executing an object detection algorithm that scans the monitored area for predefined objects. Such algorithms are well known and are therefore not explained in detail below. Automated detection can also be based on object properties, such as a detected object speed or an object's direction of movement. In particular, automated detection can be based on an evaluation of a sequence of three-dimensional data.Alternatively or additionally, the hazardous objects can be detected manually or by gesture control and thus defined for the processor unit.

[0019] According to the invention, the sensor data can be data indicating a two-dimensional arrangement, which allows a three-dimensional representation of the monitored area to be determined. The sequence of three-dimensional representations according to the invention results from a temporal sequence of individual sensor data to the sensor signal output in real time. Three-dimensional representations can also be three-dimensional point clouds, according to the invention. These allow for significantly reduced storage and transmission costs compared to the transmission of complete images. Furthermore, the use of three-dimensional point clouds allows for a particularly simple combination of data from several optical sensors to create a three-dimensional representation. Advantageously, the use of three-dimensional point clouds also allows for a particularly simple determination of a distance between objects by determining the distance of the corresponding points from the point cloud.

[0020] In particular, the three-dimensional point cloud can be reduced to relevant parts of the surveillance area through automated pre-filtering, for example based on recorded brightness contrasts, so that large monochrome and planar areas, such as a floor, can be removed from the recorded sensor data.

[0021] The detailed structure of a sensor unit according to the invention with a number of optical sensors for providing a sequence of three-dimensional representations is known and such sensor units are already commercially available, so that a detailed explanation of the functioning of this sensor unit and of the at least one optical sensor from the number of optical sensors is omitted below.

[0022] The previous evaluation data includes data collected at a time earlier than the current time and suitable for evaluating the current hazard situation. In particular, the previous evaluation data can be based on the previous positions of the object to be protected and the object posing a risk in a respective pair of objects relative to each other.

[0023] In the context of the invention, an alarm is understood to mean the output of information that informs the recipient of the output of the presence of an alarm condition. In this sense, an alarm can include a visual and / or acoustic output and / or a logging of the alarm condition. Triggering an alarm according to the invention can also include the output of alarm information corresponding to the alarm signal, through which a recipient of the output can be informed of the presence of an alarm condition.

[0024] For objects in a medical environment, the monitoring area might be, for example, a designated patient area on a hospital ward. In another example, this area might be a designated area for patients in a residential facility.

[0025] Preferred embodiments of the system according to the invention and further aspects of the invention are described below.

[0026] In a preferred embodiment, the assignment between an object to be protected and a hazardous object depends on the current position of the hazardous object within the monitoring area relative to the object to be protected. This avoids the need to monitor an unnecessarily large number of object pairs by the system according to the invention. Thus, an object generally known as a hazardous object, such as a hand, is usually only dangerous if this object is located near an object to be protected or is moving toward the object to be protected. Furthermore, the assignment according to this embodiment enables the formation of object pairs in which different objects to be protected are each assigned to different hazardous objects to form object pairs.This takes into account that a hazardous object located near a first object to be protected may pose a serious threat to that object, while it may not currently pose a threat to a second object to be protected that is far away and no corresponding pair of objects needs to be monitored.

[0027] In a particularly preferred embodiment, the parameters resulting from an evaluation of previous evaluation data, in particular previous positions, are at least a distance between the objects of an object pair and a direction of movement of the objects of an object pair relative to one another. In this embodiment, a future distance between the object to be protected and the object posing a risk within an object pair can be deduced particularly advantageously. Thus, a change in the distance between the objects can be used to deduce a speed between these objects. In a preferred variant, a higher current risk value results for a positive speed with which the objects move towards one another than for a lower speed.In a further variant of the embodiment, the current hazard value is configured such that when the object to be protected and the threatening object move away from each other, the hazard level of this pair of objects stagnates or decreases compared to a previous hazard level. The direction of movement is preferably determined by evaluating the change in distance over the course of the previously detected positions. In a variant of this embodiment, the distance is determined as a separation between the geometric centers of gravity of the two objects in an object pair. In an alternative variant of this embodiment, the distance is determined as a separation between the closest display points of a respective object in the object pair.In a further variant of this embodiment, the processor unit calculates a contact time based on the distance and direction of movement of the two objects in an object pair relative to each other. The contact time at which the two objects in an object pair would touch if the movement remained constant is calculated. In this variant, the contact time also represents a parameter resulting from the evaluation of previous evaluation data. Calculating the contact time is particularly advantageous if one of the objects is temporarily obscured from the sensor unit, for example, if the hazardous object is behind another object, such as under a ceiling.Given a currently calculated contact time, which represents a prediction of the duration until a future hazardous event occurs, an alarm signal can still be issued in good time and a corresponding alarm can be triggered in the event of such a temporary concealment of the hazardous object or the object to be protected.

[0028] In a particularly advantageous embodiment, the processor unit assigns each object pair an importance value that results from a predetermined relevance level of the object to be protected and / or from a predetermined danger level of the endangering object, and wherein the current danger value of the object pair is dependent on the importance value. Such an importance value can take into account that certain object pairs can lead to a very high danger to patient health, whereas other object pairs pose only a minor danger. For example, a wound that is currently healing can be assigned a particularly high relevance level as an object to be protected, so that any danger to this wound and the corresponding healing process quickly leads to high danger values ​​and thus a high danger amount and a rapid alarm via the alarm signal.Furthermore, a hand, as a hazardous object, can be assigned a higher hazard level than a head, so that each pair of objects with the hand is generally assigned a higher importance value because a corresponding interaction with an object to be protected poses a particularly high risk to patient health. In a preferred embodiment, the hazard level and relevance level are assigned at least partially manually. In an alternative or supplementary embodiment, the hazard level and relevance level are assigned by a predetermined assignment to predetermined objects stored in a memory of the processor unit.The importance value is determined from the relevance level and the hazard level, for example, by a sum of the two levels, a product of the two levels, an average of the two levels, a maximum value of the two levels, and / or another suitable algebraic assignment rule. Alternatively, the importance value is derived directly from the relevance level or directly from the hazard level.

[0029] In an alternative embodiment to the previous embodiments, the importance value results from an assignment of importance value to a predetermined object pair stored in a memory of the processor unit.

[0030] In a further advantageous embodiment, the current hazard value of an object pair also depends on whether a direct path between the object to be protected and the hazard-posing object is blocked by another object. This advantageously takes into account the fact that blocking the path between the two objects of an object pair results in the risk of contact between the two objects being very low, at least for a certain period of time. Accordingly, blocking the path preferably leads to a current hazard value that reduces the determined hazard value.

[0031] InIn a preferred embodiment, the monitoring of an object pair by the processor unit is terminated if the calculated hazard value falls below a lower threshold. The lower threshold is designed such that the originally hazardous object poses no danger to patient health.

[0032] It is particularly advantageous to determine the number of objects to be protected and / or the number of objects posing a risk using automated object recognition. Methods for automated object recognition are generally known, so they will not be explained in detail below. Known automated object recognition algorithms can be used to reliably detect predetermined objects. This makes it particularly advantageous to define common medical objects from the medical environment that are to be protected before using the system according to the invention, such as a tube, a catheter, a supply cable, bandages, and / or medical devices. If the processor unit detects such an already defined object in the monitoring area, it can be automatically defined as the object to be protected.

[0033] InIn a further advantageous embodiment, the determination of the number of objects to be protected is at least partially realized through interaction with a user of the system. In In a preferred variant, the interaction with the user involves a manual definition of the object to be protected, for example by marking this object on a graphical user interface, in particular by marking it with a mouse click or by touch. In In another advantageous variant, the interaction with the user is gesture control, for example, where medical personnel within the monitoring area trigger the activation of monitoring for an object defined by the gesture using a predetermined gesture. InIn a further advantageous variant, the interaction is a predetermined treatment action within the monitoring area, so that, for example, the processor unit recognizes a bandage as an object to be protected by the fact that it is applied, or that a tube is recognized as an object to be protected by the fact that it is used during integration into a patient. All variants of this embodiment have the advantage that even for unusual shapes and types of objects to be protected, which possibly cannot be detected by automatic object recognition, the user can define them as objects to be protected. Furthermore, in this embodiment, the user can ensure that no object to be protected is overlooked by the processor unit, which is particularly advantageous for patient safety.

[0034] InIn one embodiment of the system according to the invention, the temporal evaluation interval is the time interval formed by a determination time for the determination of the corresponding object pair and the current time. In This embodiment advantageously takes into account all hazard values ​​determined since the objects were assigned to an object pair. This allows for particularly accurate qualitative evaluation of the monitored area over time, allowing particularly reliable conclusions to be drawn about the future behavior of the objects in an object pair relative to each other.

[0035] InIn an advantageous embodiment, the system according to the invention further comprises a user interface which is connected to the processor unit and which is designed to enable manual control of the definition of an object to be protected, the detection of a hazardous object and / or the assignment of the two objects to an object pair. Such a user interface can be implemented, for example, by a keyboard, a touch display, a computer mouse, an optical sensor and / or a joystick. The operation of the processor unit via the user interface according to this embodiment advantageously enables particularly reliable definition of the object to be protected and / or particularly reliable detection of the hazardous object and / or particularly reliable assignment of the two objects to an object pair.In particular, errors that may occur during automated determination, recognition and / or assignment are avoided.

[0036] According to the invention, the object to be protected is at least one object from the group: infusion, medical device, cable, hose, tube, bandage, wound. Such objects regularly require protection in everyday clinical practice, so their manual and / or automatic designation as objects to be protected is particularly advantageous. In this context, a person is understood to mean, for example, the entire patient, so in this case, an alarm is provided for any contact with the patient, especially by unauthorized persons.

[0037] In a particularly advantageous embodiment of the system according to the invention, the sensor unit is further configured to output sensor data relating to a specified object currently to be protected to the processor unit less frequently as part of the sensor signal than sensor data relating to a detected object currently posing a risk. This embodiment advantageously utilizes the fact that the objects to be protected, such as the group of objects according to the previous embodiment, are objects that typically move very slowly or not at all in a clinical environment. As a result, this embodiment advantageously allows a data stream between the sensor unit and the processor unit and / or a storage space for received sensor data to be reduced.

[0038] In a further advantageous embodiment, the processor unit is configured to filter the received sensor signal and use only the filtered signal for further processing, in particular for further monitoring of object pairs. In this case, the filtered signal preferably only indicates areas of the monitoring area in which a movement or a change in the monitoring area has occurred since the last time step.

[0039] According to the invention, the hazardous object is at least one object from the group: a patient's hand, a patient's foot, a patient's head, an animal, or a bed rail. Such objects can advantageously be detected particularly reliably by automation. Furthermore, in a clinical environment, these objects can lead to an interaction with an object to be protected, which regularly poses a threat to the patient's health.

[0040] In a particularly advantageous embodiment, the alarm signal indicates the object to be protected and / or the object posing a risk. In a variant of this embodiment, the triggered alarm indicates the object to be protected and / or the object posing a risk. This allows a user of the system according to the invention to detect the existing dangerous situation as soon as they perceive an alarm triggered by the alarm signal and, if necessary, take particularly rapid precautions to safeguard the patient's health.

[0041] Particularly advantageously, different pairs of objects triggering the alarm signal can also lead to different types of alarms. For example, an alarm triggered by a stranger as a dangerous object can at least additionally be reported to the security service of the corresponding hospital. In another example, an alarm triggered by a bed rail as a dangerous object is only issued as an alarm within the corresponding room. In another example, an alarm triggered by an animal as a dangerous object is advantageously issued throughout the entire corresponding hospital ward, since this animal could also pose a danger to other patients.

[0042] In a further embodiment, the processor unit is configured to output a further corresponding alarm signal if the calculated alarm value of the object pair exceeds a further predetermined threshold, wherein the further predetermined threshold is greater than the predetermined threshold. This allows different alarm levels to be triggered depending on the current threat to the patient's health. For example, a local alarm perceptible only in the patient room is output when the predetermined threshold is reached, and a global alarm perceptible even in a medical personnel area is output when the further predetermined threshold is reached.

[0043] InIn a further embodiment of the invention, the alarm signal triggers a logging of the currently existing dangerous situation within a corresponding memory of the system according to the invention. This advantageously allows impairments to patient health due to a previous dangerous situation to be retrospectively reconstructed.

[0044] In a preferred embodiment of the system according to the invention, the alarm signal triggers an alarm for the object to be protected and / or the endangering object, corresponding to the pair of objects triggering the alarm signal. In one variant of this embodiment, the alarm for the object to be protected leads to an adaptation of this object that reduces the risk to patient health. For example, in one variant, a medical device is adapted as the object to be protected so that it does not accept any accidental input from the user. In another example, an audio message is sent as an alarm to the person constituting the endangering object, informing them of the risk to patient health.

[0045] According to a further aspect of the invention, the above-mentioned object is achieved by a method for the automatic detection of hazardous situations within a monitoring area for objects in the medical environment.

[0046] The method according to the invention comprises the following steps: Determining a sensor signal comprising sensor data indicating a sequence of three-dimensional representations of the monitoring area over a monitoring period, and outputting the sensor signal in real time, receiving the sensor signal, determining a number of objects to be protected within the monitoring area based on the indexed three-dimensional representations, detecting, in particular automatically detecting, hazardous objects within the monitoring area based on the indexed three-dimensional representations, determining a number of object pairs by assigning an object to be protected from the number of objects to be protected to a hazardous object, assigning a current hazard value to a respective object pair depending on the current positions of the object to be protected and the hazardous object and on a group of parameters resulting from an evaluation of previous evaluation data,wherein the resulting parameters are at least a distance between the objects of an object pair and a direction of movement of the objects of an object pair relative to each other, - calculating the current hazard values ​​for a respective object pair to a respective hazard amount over a time evaluation interval up to the current time, issuing a corresponding alarm signal if the hazard amount thus determined exceeds a predetermined threshold value, receiving the alarm signal and triggering an alarm depending on the alarm signal.

[0047] Advantageously, the method according to the invention allows a particularly rapid detection of a hazardous situation while avoiding false alarms.

[0048] In particular, calculating the current hazard values ​​for a respective pair of objects over the temporal evaluation interval is advantageous, since by taking into account the recent past of the monitoring area, conclusions can be drawn particularly reliably about the future development of the positions of a respective pair of objects.

[0049] Furthermore, it is advantageous that the hazardous objects are preferably detected automatically, as this ensures that non-hazardous objects are not overlooked during, for example, manual entry of these objects.

[0050] In a particularly preferred embodiment of the method according to the invention, the calculation of the current hazard values ​​comprises summing these current hazard values ​​up to the current point in time to the respective hazard amount of the object pair. This summation represents a particularly simple mathematical operation that enables particularly fast processing by the processor unit through particularly fast calculation of the current hazard values. In a preferred variant of this embodiment, the current hazard value can also be negative, so that the summed hazard amount can decrease again depending on the actions within the monitored area, for example, if the hazardous object and the object to be protected move away from each other again after an initial approach.A detailed description of how such a determination of the risk amount can be realized can be found in the description of . Figs. 3 to 5 .

[0051] In a further embodiment of the method according to the invention, this method is stopped if medical personnel are detected by the processor unit within the monitoring area. Such detection can be performed, for example, by reading an identification number carried by the medical personnel. Preferably, the method according to this embodiment is continued together with the previously determined risk values ​​as soon as the medical personnel have left the monitoring area.

[0052] Furthermore, the above-mentioned object is achieved by a program with a program code for carrying out the method according to at least one of the above-mentioned embodiments of the method according to the invention when the program code is executed on a computer, a processor or a programmable hardware component.

[0053] The program can also execute only a portion of the data processing according to the invention. Preferably, at least the function of the processor unit and the output unit are controlled by a program and / or coordinated parts of the program. In particular, the processing of the sensor data into the sensor signal is controlled in embodiments of the program according to the invention by a separate part of the program within a processor of the sensor unit. Preferably, the program according to the invention is executed by a processor of the system according to the invention. Alternatively, the program is executed at least by a first processor of the processor unit and by a second processor of the output unit.

[0054] The invention will now be explained in more detail with reference to advantageous embodiments schematically illustrated in the figures. These show in detail: Fig. 1 is a schematic representation of a first embodiment of a system according to the invention; Fig. 2 is a schematic representation of a second embodiment of the system according to the invention; Figs. 3, 4, 5 are schematic illustrations of various hazardous situations, wherein the object to be protected and the hazardous object approach each other ( Fig. 3 ), whereby the object to be protected and the object at risk move away from each other ( Fig. 4 ), and where there is another object between the object to be protected and the object at risk ( Fig. 5 ); Fig. 6 a flowchart of a first embodiment of a method according to the invention.

[0055] Fig. 1 shows a schematic representation of a first embodiment of a system 100 according to the invention.

[0056] The system 100 for automatically detecting hazardous situations within a monitoring area 102 for objects in a medical environment comprises a sensor unit 110, a processor unit 120 and an output unit 130.

[0057] The sensor unit 110 has a number of optical sensors 112 and is designed to determine a sensor signal 114 and output it in real time. The number of optical sensors 112 can further be arranged in the medical environment and is designed such that the sensor signal 114 comprises sensor data that indicates a sequence of three-dimensional representations of the monitoring area 102 during a monitoring period. In the illustrated embodiment, the number of optical sensors 112 is installed within a housing, wherein the different positions of the optical sensors 112 indicate a three-dimensional representation of the monitoring area 102 by the sensor data. In an embodiment not shown, the at least one sensor from the number of optical sensors is designed to determine the three-dimensional representations via a time-of-flight measurement within the monitoring area.In a further embodiment not shown, the at least one sensor from the number of optical sensors is designed to determine the three-dimensional representations via a so-called structured-light measurement, i.e. an output of a known radiation pattern in the non-visible wavelength range and a reception of corresponding reflections.

[0058] In the illustrated embodiment, the sensor signal 114 is output to the processor unit 120 wirelessly via a radio connection, for example, via WLAN, Bluetooth, BLE, or ZigBee. In an embodiment not shown, the sensor signal is output via a cable, for example, within the framework of a bus system, in particular an Ethernet system.

[0059] The processor unit 120 is configured to receive the sensor signal 114 and, in a first processing step 121, to determine a number of objects 122 to be protected within the monitoring area 102 based on the indicated three-dimensional representations. In the present case, only one wound 104 on the leg of the patient 106 is defined as the object 122 to be protected. This definition was performed by gesture-controlled recognition of the wound 104 by the processor unit 120. The processor unit 120 automatically recognized that the wound 104 had been treated and protected by a bandage (not shown), so that the area of ​​the wound 104 is an object to be protected.In a further embodiment not shown, the object to be protected is automatically identified by evaluating the sensor data such that an object from the following group of predetermined objects is automatically designated as the object to be protected or is automatically suggested to a user of the system as an object to be protected. This group of predetermined objects includes at least: infusion, medical device, cable, hose, tube, bandage, wound.

[0060] Furthermore, the processor unit 120 is configured, in a second processing step 123, to perform an automated detection of hazardous objects 124 within the monitoring area 102 based on the indexed three-dimensional representations. In this case, the hand 108 and the head 109 of the patient 106 are detected as hazardous objects. For the automated detection, the processor unit 120 searched the monitoring area for objects belonging to a predetermined group of hazardous objects stored in a memory of the processor unit 120. This group includes at least the following objects: foot of a patient 106, bed rail, hand 108 of a patient 106, head of a patient 106, and animal.In an embodiment of the invention not shown, the hazardous objects are defined by a manual selection on a user interface of a user interface of the system and are thereby recognized by the processor unit.

[0061] After temporarily storing objects 122 to be protected and objects 124 at risk, the processor unit 120 determines a number of object pairs 105 in an assignment step 125 by assigning an object 122 to be protected from the number of objects to be protected to a hazardous object 124. In doing so, the processor unit 120 recognizes that the head 109 of the patient 106 cannot be brought near the wound 104 and therefore does not pose a danger to the wound 104 as an object to be protected. Therefore, the processor unit 120 determines the combination of hand 108 and wound 104 as the only relevant object pair 105 to be monitored.

[0062] Furthermore, the processor unit 120 is designed to monitor the positions of the object to be protected 122 and the endangering object 124 of the only monitored object pair 105 in this exemplary embodiment within the monitoring area 102 during the monitoring period as part of a monitoring process 127, and to assign a current hazard value to a respective object pair 105 depending on the two current positions and a group of parameters resulting from an evaluation of earlier evaluation data. Finally, over a temporal evaluation interval, the current hazard values ​​up to the current time are calculated for each object pair 105, and a corresponding alarm signal 126 is output if a thus calculated hazard value of an object pair 105 exceeds a predetermined threshold value. A concrete exemplary embodiment for such a calculation of the hazard values ​​is described in the context of the Figures 3 to 5 described.

[0063] In the illustrated embodiment, the alarm signal 126 is output via a cable. In an embodiment not shown, this output occurs via a wireless connection, as is known in various forms in the field of communications technology.

[0064] The output unit 130 is designed to receive the alarm signal 126 and to trigger an alarm 132 depending on the alarm signal 126. In the illustrated embodiment, the alarm is an acoustic alarm provided by a loudspeaker 134. In an alternative or supplementary embodiment not shown, a visual alarm is triggered by the output unit. In a further embodiment not shown, a logging of the dangerous situation is triggered by the output unit. In a further embodiment not shown, the alarm is issued by an external device that is not part of the system according to the invention, but the output unit 130 is designed to trigger the corresponding alarm by a trigger signal on the external device.

[0065] In the illustrated embodiment, all units of the system 100 according to the invention have separate housings, since all units are arranged at a distance from one another. This makes it possible, for example, to provide the processor unit at a central location in the medical environment and to operate a number of systems according to the invention with a corresponding number of different sensor units using a common processor unit.

[0066] Fig. 2 shows a schematic representation of a second embodiment of the system 200 according to the invention.

[0067] The System 200 differs from the one in Fig. 1The system 100 shown is distinguished, among other things, by the fact that the sensor unit 210 comprises two separate cameras with a corresponding number of optical sensors 212, 212' and two partial monitoring areas 103, 103'. Combined processing of the correspondingly acquired sensor data is possible in that the partial monitoring areas 103, 103' have a common monitoring area 102 as an intersection. This indexes a single sequence of three-dimensional representations of this common monitoring area 102.

[0068] Furthermore, the System 200 differs from the System 100 in that Fig. 1 that processor unit 120 and output unit 230 are arranged in a common housing 240. The housing 240 also includes a user interface 250 as an additional component of the system 200.

[0069] The user interface 250 is signal-connected to the processor unit 120. The user interface 250 is designed to enable manual control of the definition of an object to be protected, the detection of a hazardous object, and / or the assignment of the two objects to an object pair. For this purpose, the user interface 250 has a touch display 252 for receiving a user input 254. In the present case, the definition of an object to be protected occurs such that the current monitoring area 102 is displayed on the touch display 252, and the object to be protected can be defined by touching this object on the touch display 252. The manual detection and storage of a hazardous object occurs analogously.Manually assigning two objects to an object pair via the touch display 252 can be achieved, for example, by swiping one object over another, by touching both objects simultaneously, by quickly tapping two objects on the touch display in quick succession, and / or by making a corresponding note in a database of the user interface 250. In an embodiment not shown, the user interface is designed as a keyboard, a computer mouse, an optical sensor, and / or a joystick. In this case, the user interface 250 outputs a corresponding input signal 256 to the processor unit 120 in accordance with the user input 254.

[0070] In the illustrated embodiment, the output unit 230 is connected to the processor unit 120 via a cable. Unlike the system 100 of Fig. 1, the output unit 230 is designed to output a trigger signal 236 to a communication network 260 of the hospital in accordance with the received alarm signal 126 and to trigger an alarm within this communication network 260. Both the alarm signal 126 and the correspondingly output trigger signal 236 indicate the object to be protected and / or the object posing a risk. As a result, an alarm can be triggered by the communication network 260 that is specifically tailored to the resulting hazardous situation. For example, only one alarm device can be triggered within the relevant hospital room, or in a serious case, such as a threat from an animal, a group of alarm devices distributed across a hospital ward can be triggered.

[0071] Figures 3, 4 and 5show schematic illustrations of various hazardous situations, with the object 122 to be protected and the hazardous object 124 approaching each other ( Fig. 3 ), whereby the object to be protected 122 and the endangered object 124 move away from each other ( Fig. 4 ), and wherein between the object 122 to be protected and the endangered object 124 there is another object 570 ( Fig. 5 ).

[0072] Based on the Figures 3 to 5 The following describes the assignment of the same pair of objects 105 to a current risk value and the calculation of the current risk values ​​to a risk amount in accordance with the Fig. 1 illustrated preferred embodiment.

[0073] In the three hazardous situations shown, the object 122 to be protected is, as already explained in the examples from Fig. 1 and 2explained, a wound 104 on the knee of the patient 106. The endangering object 124 is again the hand 108 of the patient 106 and the object pair under consideration is therefore formed by the wound 104 and the hand 108 of the patient 106.

[0074] The current hazard value assigned to this object pair 105 at the current time g ( t, x, p 1, ..., pn ) depends on the current time step t, the current distance vector x, which has the absolute value D of the distance between the two objects of the object pair and is oriented along a direction defined by the current position of the object 122 to be protected and the current position of the endangering object 124. Furthermore, the hazard value g is dependent on several parameters p 1 , ..., pn , which result from the evaluation of earlier evaluation data. These parameters are preferably the absolute value D of the current distance vector and / or a current direction of movement R of the two objects relative to one another and / or an importance value assigned to the object pair 105 by the processor unit and / or a parameter which indicates whether there is a clear path between the endangering object 124 and the object 122 to be protected.

[0075] The geometric center of gravity of the respective object is used by the processor unit 120 as the current position of the respective object. In an embodiment not shown, the distance vector is formed by the closest pair of sensor data points of the respective object. Preferably, the geometric center of gravity of the object is not recalculated with each new data set of sensor data; particularly preferably, this center of gravity is calculated only once for an object within the monitoring area, and is then held at a fixed position relative to the circumference of the object. This advantageously reduces the calculation time for the processor unit.

[0076] From the current positions and a change in these positions over time, in addition to the distance D, a speed of the two objects of an object pair 105 relative to each other and a direction R of this movement can be determined as parameters for the hazard value g. In the illustrated embodiment, the processor unit 120 evaluates, in addition to the direction of movement, the proportion r of the movement of the hazardous object in the direction of the object to be protected. This is done by projecting the corresponding movement vector of direction R onto an axis A formed by the two objects.

[0077] In Fig. 3The portion r is configured such that it points from the hand 108 toward the wound 104. Given this movement of the two objects of the object pair 105 relative to each other, the current hazard value g will be greater than the hazard value determined in an earlier step, so that the corresponding calculated hazard amount is currently greater than at the earlier point in time. Should the hazard amount thereby become greater than the predetermined threshold, the system 100 would issue a corresponding alarm signal.

[0078] In Fig. 4The portion r is designed such that it points away from the hand 108 in a direction away from the wound 104. Given this movement of the two objects of the object pair 105 away from each other, the current hazard value G will be smaller than the hazard value determined in an earlier step, so that the corresponding calculated hazard amount is currently smaller than at the earlier point in time. Therefore, the triggering of the alarm signal becomes less likely with the movement shown than if the two objects remained in their current position.

[0079] In the illustrated embodiment, the calculation of the hazard values ​​is realized by summing them up for the evaluation interval to the hazard amount. This results in the hazard amount B calculated according to B t , T , x → , p 1 , … , p n = ∑ i = t − T i = t g i , x → , p 1 , … , p n . Here, T is the temporal evaluation interval for the monitored object pair. T specifies a fixed number of discrete time steps into the past over which the hazard value g is summed. In another advantageous embodiment, the hazard value B is calculated according to B t , T , x → , p 1 , … , p n = ∑ i = Z i = t g i , x → , p 1 , … , p n . Here, Z is the point in time at which an assignment between the endangered object and the object to be protected has taken place and a corresponding object pair is available for monitoring.

[0080] The hazard value g is preferably a rational number that is positive if the hazardous object is moving in the direction of the object to be protected, or if the distance between these two objects has currently decreased faster than in the last time step, or if the object to be protected is assigned a particularly high relevance level, or if the hazardous object is assigned a particularly high hazard level. The dependence of the hazard value g on the distance can be continuous, for example, realized by a function that increases monotonically with decreasing distance. The hazard value g is zero or negative if the distance between the hazardous object and the object to be protected increases, as in Fig. 4Furthermore, the hazard value is zero or negative if an object blocks the direct path between the endangering object and the object to be protected, as is the case in Fig. 5 by an object 570 located between the hand 108 and the wound 104, namely a tray. A negative risk value or one that stagnates at zero ensures that the risk amount does not increase, since, in particular, when calculating by summation according to the present embodiment, no positive number is added to the previous risk amount.

[0081] In a concrete calculation example for the Fig. 3In the hazardous situation shown, after an initial movement of the hand 108 in the direction of the wound 104, a hazard value was changed from 0 to 1, so that the hazard amount was also 1. After the hand 108 then remained in one position for some time, the hazard value was changed back to 0 and the hazard amount remained at 1. With the movement shown starting again, the hazard value was changed to 2, since now in addition to the movement, the distance is also below a corresponding threshold value, so that the hazard value is additionally increased by 1, corresponding to the inverse dependency g D = 0.5 / D with the specific distance D = 0.5 m starting from the threshold value of 0.5 m. As a result, the hazard amount is now 3. In this case, the alarm signal is issued from the threshold value of the hazard amount of 4.5.Therefore, at a new time step in which the hand moves towards the wound, this alarm signal will be triggered unless there is a prior movement away from the wound, which would lead to a negative hazard value and a corresponding reduction in the hazard amount.

[0082] Preferably, the hazard value is linearly dependent on the importance value assigned to the respective object pair by the processor unit. The importance value is preferably proportionally dependent on the predetermined relevance level of the object to be protected and / or the predetermined danger level of the endangering object.

[0083] In a preferred embodiment, the determination of specific rational numbers for the hazard level and / or the relevance level and / or the importance value and / or the danger value is carried out by manual user input and / or by a predetermined assignment of values ​​provided by the manufacturer and stored in a memory of the processor unit. This allows individual preferences to be taken into account when assessing a hazardous situation.

[0084] Fig. 6 shows a flowchart of a first embodiment of a method 600 according to the invention.

[0085] The method 600 for automatically detecting hazardous situations within a monitoring area for objects in a medical environment comprises a sequence of steps described below: A first step 610 comprises determining a sensor signal comprising sensor data indicating a sequence of three-dimensional representations of the monitoring area over a monitoring period, and outputting the sensor signal in real time.

[0086] A subsequent step 620 includes receiving the sensor signal.

[0087] A next step 630 comprises determining a number of objects to be protected within the monitoring area based on the indexed three-dimensional representations.

[0088] A step 640 which can be executed in parallel to step 630 comprises a detection, in particular an automated detection, of dangerous objects within the monitoring area based on the indexed three-dimensional representations.

[0089] A further step 650 comprises determining a number of object pairs by an assignment between an object to be protected from the number of objects to be protected and a dangerous object.

[0090] A subsequent step 660 comprises assigning a current hazard value to a respective object pair depending on the current positions of the object to be protected and the hazardous object and on a group of parameters resulting from an evaluation of previous evaluation data.

[0091] A next step 670 comprises calculating the current hazard values ​​for a respective object pair to a respective hazard amount over a temporal evaluation interval up to the current time.

[0092] A further step 680 comprises issuing a corresponding alarm signal if the thus determined hazard amount exceeds a predetermined threshold value.

[0093] A final step 690 includes receiving the alarm signal and triggering an alarm depending on the alarm signal.

[0094] Steps 610 and 620 are typically executed repeatedly in real time, i.e. in short successive time steps.

[0095] Steps 630 and 640 may be performed in parallel or sequentially in any order.

[0096] Step 650 is executed once during the monitoring of an object pair, whereas steps 660 and 670 describe the actual monitoring process and are executed repeatedly in successive time steps, so that the respective hazard amount is always offset against a hazard value adapted to the current situation in the monitored area. Steps 680 and 690 are preferably executed only once consecutively during the monitoring of an object pair, namely after the hazard amount exceeds the predetermined threshold.

[0097] In In a preferred embodiment, the calculation of the current hazard values ​​according to step 670 is carried out by summing these hazard values ​​over the time evaluation interval up to the current point in time.

[0098] The method according to the invention preferably comprises, in a further step, deactivating the method 600 if the processor unit detects the presence of medical personnel within the monitoring area through the sensor signal. In a preferred example of this variant, the method 600 is continued with the data detected before the medical personnel were deactivated if the detected medical personnel have left the monitoring area. In a variant of this embodiment, the medical personnel are identified by other persons through automatic detection of an identification number carried by the medical personnel. This is possible, for example, by a key card carried on which the identification number is stored and / or by a corresponding arrangement of identification information on the clothing of the medical personnel.In an alternative or complementary example, the identification information is provided manually by the medical staff. List of reference symbols

[0099] 100, 200System 102Monitoring area 103, 103'Partial monitoring areas 104Wound 105Object pair 106Patient 108Hand 109Head 110, 210Sensor unit 112, 212, 212'Number of optical sensors 114Sensor signal 120Processor unit 121First processing step 122Object to be protected 123Second processing step 124Endangering object 125Assignment step 126Alarm signal 127Monitoring process 130, 230Output unit 132Alarm 134Loudspeaker 236Trigger signal 240Housing 250User interface 252Touch display 254User input 256Input signal 260Communication network 570Further Object 600Procedure 610, 620, 630, 640, 650, 660, 670, 680, 690Procedure steps DDistance between monitored objects RDirection of movement rProportion of movement AAxis between monitored objects

Claims

1. System (100) for automatically detecting hazardous situations within a monitoring region (102) for objects in a medical environment, comprising - a sensor unit (110) having a number of optical sensors (112), which is designed to determine a sensor signal (114) and output it in real time, wherein the number of optical sensors (112) can also be arranged in the medical environment and designed such that the sensor signal (114) comprises sensor data which indicate a sequence of three-dimensional representations of the monitoring region (102) over a monitoring period, - a processor unit (120) which is designed - to receive the sensor signal (114), - to establish a number of objects (122) to be protected within the monitoring region (102) on the basis of the indicated three-dimensional representations, wherein the number of objects (122) to be protected is at least one object from the group: infusion, medical device, cable, hose, tube, bandage, wound (104), - to detect hazardous objects (124) within the monitoring region (102) on the basis of the indicated three-dimensional representations, wherein the hazardous objects (124) are at least one object from the group: hand (108) of a patient (106), foot of a patient (106), head (109) of a patient (106), animal, bed rail, - to determine a number of object pairs (105) by an assignment between an object (122) to be protected from the number of objects (122) to be protected and a hazardous object (124), - to monitor the positions of the object (122) to be protected and the hazardous object (124) of a relevant object pair (105) within the monitoring region (102) during the monitoring period, - to assign a current hazard value to a relevant object pair (105) depending on the two current positions and a group of parameters resulting from an evaluation of previous evaluation data, wherein the resulting parameters are at least a distance (D) between the objects of an object pair (105) and a direction (R) of movement of the objects of an object pair (105) relative to one another, and - to sum up each of the current hazard values up to the current point in time for each object pair (105) over a temporal evaluation interval, and to issue a corresponding alarm signal (126) if a thus calculated hazard amount of an object pair (105) exceeds a predetermined threshold value, - an output unit (130) which is designed to receive the alarm signal (126) and to trigger an alarm (132) depending on the alarm signal (126).

2. System (100) according to claim 1, wherein the assignment between an object (122) to be protected and a hazardous object (124) depends on a current position of the hazardous object (124) within the monitoring region (102) relative to the object (122) to be protected.

3. System (100) according to at least one of the preceding claims, wherein the processor unit (120) assigns each object pair (105) an importance value which results from a predetermined relevance level of the object (122) to be protected and / or from a predetermined hazard level of the hazardous object (124), and wherein the current hazard value of the object pair (105) depends on the importance value.

4. System (100) according to at least one of the preceding claims, wherein the current hazard value of an object pair (105) also depends on whether a direct route between the object (122) to be protected and the hazardous object (124) is obstructed by another item (570).

5. System (100) according to at least one of the preceding claims, wherein the establishment of the number of objects (122) to be protected and / or the detection of hazardous objects is achieved by means of automated object detection.

6. System (100) according to at least one of the preceding claims, wherein the establishment of the number of objects (122) to be protected is at least partially achieved by means of an interaction with a user of the system (100).

7. System (100) according to at least one of the preceding claims, wherein the temporal evaluation interval is the time interval which is formed by a determination time for the determination of the corresponding object pair (105) and the current time.

8. System (200) according to at least one of the preceding claims, further comprising a user interface (250) which is connected to the processor unit (120) and which is designed to allow manual control of establishing an object (122) to be protected, detecting a hazardous object (124) and / or assigning the two objects to an object pair (105).

9. System (200) according to at least one of the preceding claims, wherein the alarm signal (126) indicates the object (122) to be protected and / or the hazardous object (124).

10. Method (600) for automatically detecting hazardous situations within a monitoring region (102) for objects in a medical environment, comprising the steps of - determining a sensor signal (114) comprising sensor data which indicate a sequence of three-dimensional representations of the monitoring region (102) over a monitoring period, - outputting the sensor signal (114) in real time, - receiving the sensor signal (114), - establishing a number of objects (122) to be protected within the monitoring region (102) on the basis of the indicated three-dimensional representations, wherein the number of objects (122) to be protected is at least one object from the group: infusion, medical device, cable, hose, tube, bandage, wound (104), - detecting hazardous objects (124) within the monitoring region (102) on the basis of the indicated three-dimensional representations, wherein the hazardous objects (124) are at least one object from the group: hand (108) of a patient (106), foot of a patient (106), head (109) of a patient (106), animal, bed rail, - determining a number of object pairs (105) by an assignment between an object to be protected from the number of objects (122) to be protected and a hazardous object (124), - assigning a current hazard value to a relevant object pair (105) depending on the current positions of the object (122) to be protected and the hazardous object (124) and depending on a group of parameters resulting from an evaluation of previous evaluation data, wherein the resulting parameters are at least a distance (D) between the objects of an object pair (105) and a direction (R) of movement of the objects of an object pair (105) relative to one another, - performing a calculation on the current hazard values for a relevant object pair (105) to obtain a relevant hazard amount over a temporal evaluation interval up to the current time, wherein performing a calculation on the current hazard values comprises summing up these current hazard values up to the current point in time to obtain the relevant hazard amount of the object pair (105), - outputting a corresponding alarm signal (126) if the thus determined hazard amount exceeds a predetermined threshold value, and - receiving the alarm signal (126) and triggering an alarm (132) depending on the alarm signal (126).