WO2012131171A1

WO2012131171A1 - Arrangement for monitoring and predicting a fit or a seizure

Info

Publication number: WO2012131171A1
Application number: PCT/FI2012/050327
Authority: WO
Inventors: Katja KÄÄRIÄ; Ari Nikkola
Original assignee: Vivago Oy
Priority date: 2011-03-31
Filing date: 2012-03-30
Publication date: 2012-10-04
Also published as: FI126095B; EP2691021A4; EP2691021A1; FI20115308A; FI20115308A0

Abstract

The invention relates to an arrangement for monitoring or predicting a fit, the arrangement comprising a wearable, non-invasive monitoring device, means for creating a personal baseline pattern and means for analysing changes in the collected data with respect to the personal baseline pattern. It is characteristic of the invention that the arrangement further comprises means for updating the baseline pattern and means for comparing the updated version of the baseline pattern with the history. The invention further comprises a corresponding software product for monitoring or predicting a fit.

Description

ARRANGEMENT FOR MONITORING AND PREDICTING A FIT OR

A SEIZURE

The invention relates to an arrangement and a software product for monitoring and predicting a fit. The arrangement and software product implement a method for measuring the activity of a person and for warning of the risk of a fit. To carry out the method, a person wears a non-invasive monitoring device for collecting information on the person's kinetic motor activity and/or physiological properties. From the patent publication US 2010/0145236 Al is known a system and an apparatus for continuous monitoring of movement disorders. The apparatus can be used for simultaneous monitoring in multiple locations and for comparing the monitoring results with the cardinal motor symptoms of a particular illness.

US2007123754 describes a motor function monitoring system, for example, for indicating the risk of a heart attack to remote caregivers. In the solution described in the publication, the requirement for data communication is high and the system is not easily operable without a continuous remote data communication link.

The aim of the invention is to provide an arrangement and software product which make possible versatile observation of patients by means of an easy- to-use, wearable monitoring device, and which make remote alarms possible even without continuous or almost continuous data communication links, and which are still able to indicate long-term changes and to alarm when the change should be brought to the attention of the care personnel. A further aim is to provide a wearable device with improved usability, the battery life of which is good and the alarm properties of which are still adapted to the properties of the person monitored. This aim is achieved by means of the characteristics disclosed in the appended claims 1 and 12. The dependent claims disclose preferred embodiments of the invention. In the following drawings:

Figure 1 shows diagrammatically a typical signal based on the minute averages of the raw data obtained from the movement sensors. In the signal, the difference in motor activity in the waking state and in sleep is clearly distinguishable. The greater this difference, the better the night/day rhythm. On the horizontal axis in the Figure are the hours of one day.

Figure 2 shows the most essential stages of the method as a flow chart.

The method is preferably implemented by means of a wrist-held monitoring device which comprises sensors for detecting movements and/or physiological properties, a memory for storing the raw data collected from the sensors, a program-controlled processor, a program/algorithms which controls/control the processor to generate a personal baseline pattern from the raw data collected on the basis of its long-term monitoring. The same monitoring device further comprises a program/algorithms which controls/control the processor to detect and analyse changes from the personal baseline pattern. The monitoring device is programmed to alarm the desired party when the change from the personal baseline pattern exceeds a predetermined limit or correlates with a specific other baseline pattern which represents motoric or physiological cardinal symptoms of a particular illness. Such illnesses are, for example, epilepsy and shock relating to diabetes. It is typical of the fit types to be monitored that they are associated with a changed motor state which can be detected by measurement. The features of the movement associated with a fit (e.g. amplitude and/or frequency) differ from the features of the normal movements of the said person. On the basis of the alarm, the care personnel is able to take the necessary more specific laboratory tests and carry out examinations in order to make a diagnosis. It is characteristic of the solution according to the invention that the features of a person's normal movements are first detected and from these is formed a personal baseline pattern by means of long-term monitoring. From the motion signal to be measured are then sought features which differ from the baseline pattern formed. The differing feature may be either modelled in ad- vance or a feature associated with a fit may be identified separately for each user. The method is, therefore, adaptive. When a feature or features computed from the motion signal exceed a predetermined threshold value, a deviation from the personal baseline pattern is identified. This indicates a potential fit or an increase in the risk of a fit. If false alarms are to be mini- mised, the identification limit is set close to the average limits of measured fits. If it is to be ensured that every possible fit is monitored, the identification limits are set close to the parameters of normal movement. Following an alarm, the care personnel is preferably also able to use the measurement data or values derived from it to help in making the diagnosis.

The baseline pattern is updated, or can be updated, continuously with data obtained from monitoring, which data is also used for detecting features deviating from the baseline pattern in such a way that a detected deviation, which exceeds the alarm limit or other predetermined limit, is not used for updating.

From the data obtained by monitoring is easy to identify the awake time and the sleep time. Baseline patterns are formed separately for both of these, that is, an awake time baseline pattern and a sleep time baseline pattern. These typically follow the personal night and day rhythm. Fits can thus be detected selectively also only on the basis of the data collected on kinetic motor activities, possibly combined with skin humidity and temperature measurement. The pulse can also be measured either by auscultation or electrically. Furthermore, it is possible to use various wireless wearable sensor devices, which preferably pack the measurement data and preprocess it before transmission. In this way can normally only be sent change data and confirmations of an unchanged state. In this way can be achieved a tolerably long battery life and the failure of one sensor device will not prevent other operations. The data can, for example, be first collected into a wrist-held sensor device, or all sensor devices can communicate independently with the base station. If the data is first collected into a wrist-held device, it may use data from other sensors for alarms also independently, or the wrist-held device can at the same time also function as, for example, an ordinary pulse meter or running odometer by means of an acceleration sensor attached to the ankle or shoe, and simultaneously use data from other sensors in accor- dance with the invention.

The updated version of the baseline pattern is preferably compared with the history, that is, the stored data. Changes in the patient's activity, motor functions or other measurable variables can then be detected. For example, slowly occurring weakening of motor functions can be detected or, for example, the advancing of a circulatory disorder in the extremities over a longer period. In that case, the trained model is compared with old measurements going back at least weeks, preferably months or years. It is then very easy to compare, for example, the development of the sleep rhythm or of the amount of exercise, for example, also in support of fitness or sports training.

Using baseline patterns for long-term comparison facilitates, for example, the data processing of a wrist-held measuring and alarm device, because normally, the measurement results are only compared with one or a few base- line patterns during continuous activity. On the other hand, from the point of view of storing measurement data, baseline patterns formed by means of continuous learning are already in a compact and easily comparable form, and thus by comparing the baseline patterns with historic models describing the past, long-term follow-up can be carried out using less resources than by comparing the measurement data itself with old measurements. Thus, the comparison can also be carried out on a wrist-held device, or at least the data required can be carried along in the wrist-held device.

In practice, the invention according to claim 1 makes it possible that also the wrist-held device is capable of comparing measurement data, even over a longer period, because it is sufficient to store and compare only the baseline patterns with one another over a long period and not the data itself. For example, a slow decrease or increase in activity or, for example, a change in a particular frequency component is shown as a change in the updating adaptive baseline pattern, and when the current baseline pattern is compared with the old ones, the slow change can be detected in the baseline patterns themselves. Knowledge of the change may help a physician to make the correct diagnosis even before clear symptoms appear. For example, many chronic diseases can be seen as changes long before the symptoms are normally detectable.

Also the baseline pattern values alone can be transmitted for comparison to the system used by the physician. In this way, for example, the life of the batteries can be extended, because there is less data to be transmitted. In this case, a small change is shown as a change in the last baseline pattern compared with the previous ones and will finally result in an alarm when the long-term change is significant enough. The physician may also use the changed data to aid diagnosis.

The wrist-held device may also comprise storage space for storing the meas- urement data itself, in which case it can be loaded for processing outside the wrist-held device, for example, in connection with recharging the batteries. The invention still reduces the need for data processing, because only relatively new measurement data needs to be processed when updating the baseline pattern, if the baseline pattern is also updated on the wrist-held device. If the reference data is only updated at regular intervals, for example, daily or weekly, the wrist-held device only has to compare the latest baseline patterns with the measurement data. In this case, the baseline patterns are not compared until after the computation of the most recent baseline pattern, before loading the baseline pattern into the wrist-held device. The old measurement data can then simply be left in the non-volatile memory until something is written over it again and/or until the data has been loaded in store for further processing. If measurement data is collected, for example, on an SD card or a similar non-volatile memory, its mere storage does not increase power consumption. For example, a few gigabytes of memory may be reserved for storing the measurements and new measurements can be stored over the oldest measurements.

The updating of the baseline pattern can then be done, for example, for the past few days or weeks. The baseline patterns can also be stored daily or weekly, and similarly, comparison can also be done, for example, daily or weekly. Comparison over a longer period can still only be done on the basis of the collected data or the baseline patterns, for example, in connection with a physician's consultation, where, for example, the trend over several years can be followed. In that case, the wrist-held device compares, for example, only the changes within the past year for an alarm. The baseline pat- terns can also be formed by external data processing by transferring the data every now and then for processing, and by loading the baseline pattern in the wrist-held device or other wearable device. This is advantageous at least in connection with product development. The wrist-held device is, however, then less independent, at least unless the wrist-held device itself also up- dates the latest baseline pattern ands makes a comparison between forming the baseline patterns. The baseline patterns and measurement data stored in the wrist-held device are preferably accessible and are carried along by the patient, whereby each attending physician can easily view them. The baseline patterns can be up- dated in several places, also without a data communication link to a centralised database, because the baseline patterns can be stored in the wrist-held device itself. This facilitates data transfer, for example when moving outside the scope of the patient's normal healthcare district. The base station forwarding the alarm may also make contact by using a non-continuous con- nection, for example, a short message, land line or mobile call, or fax. The base station is then easy to move or carry along. A portable base station is preferably made with PDA or mobile phone software, in which case it is easy to use GPS or base station positioning for informing of the alarm site. The follow-up of long-term trends by means of the baseline patterns and the adaptive baseline pattern updated by the wrist-held device itself make possible, for example, a wrist-held device with highly independent alarm functions. The device may also alarm locally with sound and/or light, or also make contact with a base station for relaying the alarm. Thus, the system does not require a particular infrastructure for generating an alarm. It is much easier to introduce a base station system which merely relays an alarm, than a more extensive system, where data has to be processed in several places. The alarm alone does not even require the base station to identify the user, nor does the base station have to store any data. It suffices that the device or base station receiving the message is able to forward the message either to a predetermined address or an address determined by the alarm message. The base station of, for example, a nursing institution or a home is thus also able to relay the alarm messages of visitors either to the address determined in the message transmitted by the wrist-held device and/or to the nearest alarm address used by the base station. Preferably, the wrist-held devices do not have to be formed into a pair with the respective base station merely to deliver the alarm.

Since the alarm can be given independently, the wearable device does not have to be in continuous data communication to transmit the measurement data. This saves the device's batteries considerably compared to remote monitoring of measurement data via a base station. Furthermore, the base station can relay the alarm, for example, as an SMS message, an e-mail, a telephone call, etc. The information on the recipient of the message can then be specified on the wrist-held device, which informs the base station in the message, for example, of the number of the attending physician or the numbers for an SMS message. Thus, the base station raising the alarm may be, for example, a smartphone, which receives the message from the wrist-held device and transmits it, for example, as an Internet message, an SMS mes- sage, a telephone call or a video call. The software required for alarms only is simple and thus it can be installed in several telephones.

The monitoring device is typically a device worn on the wrist, which makes it easy to use and it does not complicate the person's normal life. The adapta- tion of the monitoring device can be carried out in a specific learning mode, where typical features of both a normal motor state and of a fit are identified. The adaptivity may be continuous, whereby the accuracy of detection can be improved. The monitoring device transmits the alarm and other data by a radio link to the base station. In the monitoring device is preferably also integrated a manual alarm function. The measuring sensor is, for example, a 3D acceleration sensor. One or more different types of acceleration sensors (dynamic and static acceleration sensors) can be used; in addition can be used a 6D sensor which also detects all rotational motions. In this way can, with relative reliability, be detected, for example, a vehicle driving situation when us- ing a steering wheel, and also the driver's level of alertness can be measured, for example, by monitoring slight corrective movements and speed variations (that is, the longitudinal acceleration of the vehicle). The temperature and humidity of the skin can also be measured. In the method can be used a fixed base station or a mobile base station. With the alarm data can be transmitted information on the location of the person or base station. The monitoring device can also transmit the data directly in a mobile network or other fixed networks. The mobile network part of the monitoring device does not have to be in operation continuously, but can be switched on only for the alarm, whereby the mobile network part will only wear the batteries when an alarm is transmitted. As a base station can be used, for example, a mobile data terminal, such as a mobile phone or a small portable computer with a data line, for example, a tablet device in which is installed a suitable software means, and which comprises the required data communication links, such as 3G Internet and a Bluetooth connection, ANT+ or the like for measurement and alarm data.

The procedures described above up to the decision to alarm can thus be carried out on the wearable monitoring device. It thus remains the function of the base station to relay the general activity level and alarms to the desired location.

In the following is described an embodiment of the invention for monitoring an epileptic fit in connection with a fixed based station in an apartment:

- Movement is measured with a 3D acceleration sensor and the monitoring device is a watch-like device worn on the wrist.

- The monitoring device is connected to the base station with an 869 MHz radio link, which is similar to those in security telephones. Other data transfer means, such as Bluetooth, WLAN, can also be used, in which case a smartphone loaded with suitable software can be used as the base station. For the data transfer of possible additional sensor devices can be used ANT+ data transfer, whereupon the ready-to-use sensors of third parties, such as pulse-measuring breast bands, can also be used.

- The base station is in contact with a corresponding 869 MHz radio link to a mobile phone with a positioning function.

- The base station transmits the data to a server, in which is collected the required information on the person.

- The detection of a deviating state takes place in such a way that the monitoring device first collects information on a person's motor state over, for example, a two-week period, separately for night-time and daytime. On the basis of the data collected are computed the features of normal movement: frequency content and the amplitudes of differ- ■ ent frequencies.

- Of this data is formed a normalised vector which represents the person's normal movement. For example, 10 different typical vectors of normal movement are stored, which form the personal baseline pattern.

- Data is also collected during fits and a feature vector describing the motor state relating to a fit is computed from these.

- In a fit detection state is computed the distance between the vectors of the baseline pattern and the fit vector.

- When the predetermined threshold value is exceeded, an alarm is given.

The base station can also be a wearable mobile device. In that case, the base station can also be, for example, a smartphone or a tablet computer or the like equipped with mobile Internet or other data line, whereby the sensor can operate directly via a Bluetooth link with the base station and the base station can give the alarm by means of mobile Internet or the telephone network. The base station may also be, for example, a separate device with, for instance, an 869 MHz radio link for a sensor device and a possible interphone, and which uses a wire or wireless connection to a mobile phone or mobile Internet for forwarding the alarm. The detection of a change and the decision to transmit an alarm are carried out separately at different stages. Criteria for generating an alarm are, for example, the degree of the change and the duration of the deviating state. These criteria may be personal. The criteria can be changed on the basis of activity data in such a way that the alarm limit can be lowered on the basis of the activity data, for example, by teaching the system afterwards to react differently to different activity signal types. For example, the system can be taught to identify the driving of a car or other vehicle, taking a sauna, the user's exercise activities and simi- lar situations where there is reason to change the alarm criteria for the sake of safety or to avoid false alarms. In such a case, the historic data in the user's measurement data is compared with the daily routine described by the user and from this is afterwards identified, for example, the driving of a car, during which a potential fit is more dangerous than at other times and the risk of fit can then be warned of more easily. Similarly, for example, a detected risk of falling asleep or a decrease in alertness may set off an alarm if the person's motor and vibration patterns indicate the use of a steering wheel and the vibrations of the car engine at the same time. In that case, tfie person does not have to notify the system of the use of the vehicle, but the use of the vehicle's control devices can be identified merely from the acceleration measurements of the wrist. Alertness can be measured, for example, by 6D measurement of the fine motor functions of steering wheel use and variations in speed and direction during driving, where sharp, infrequent steering reactions indicate a decrease in alertness (or, for example, intoxica- tion). Similarly, the slackening of steering and a simultaneous steady change of speed usually indicate that falling asleep has already taken place. Usually, by teaching the motion features of activities, the selectivity and reliability of the system can be improved and the identification of the activities also helps care personnel in interpreting the data and in using it for patient monitoring.

In identifying motion features, feature vectors can be created for different time periods in such a way that the shortest compared period examined is represented by the length of the period due to the frequency of the variable measured and by the length of the vector required for monitoring the phenomenon indicating an illness, when processing the measurement data corresponding to the prevailing state of activity. The sample rate must then be at least twice the measured frequency and the length of the time domain sample vector must be sufficient for making reliable feature identification. From these short-term measurement vectors can be extracted features, for example, the average intensity of the activity and the most significant frequency, and of these first features extracted from the short-term vectors can further be assembled vectors representing longer periods, by means of which the day-rhythm and activity during a few hours or days can be followed. Fur- thermore, from the vectors of a longer period can be extracted features, for example, the amounts of sleep time and awake time, statistical values for the heaviness and consistency of sleep. These third vectors are further used for monitoring ways of life and sleep rhythm and activity over a longer period of time. The data collected by the measuring device itself can also be stored as a whole, in which case all of the above-mentioned vectors can also be created afterwards from the start. On the other hand, the measuring device can only transmit preprocessed extracted features computed from the first vectors, or a lossy or lossless data compression method can be used to reduce the amount of data transmitted or stored. For data transfer can also be used wire transfer (e.g. USB), whereby the batteries of the wrist-held device can be recharged simultaneously with data transfer.

Claims

1. An arrangement for monitoring and predicting a fit, the arrangement comprising:

- a wearable, non-invasive monitoring device for collecting data on a person's kinetic motor activity and/or physiological properties,

- means for creating a personal baseline pattern on the basis of long- term follow-up of monitoring data collected on the features of move- ments and/or physiological properties,

- means for analysing changes in the data collected with respect to the personal baseline pattern and for generating an alarm, at least when the difference to the personal baseline pattern exceeds a predetermined limit,

characterised in that

- the arrangement further comprises means for updating the baseline pattern and means for comparing the updated version of the baseline pattern with the history, whereupon from the changes in the baseline pattern compared to earlier baseline patterns can be monitored changes that take place slowly.

2. An arrangement as claimed in claim 1, which comprises means for updating the baseline pattern continuously with data obtained from monitoring, the said arrangement also being used for monitoring features deviating from the baseline pattern in such a way that a detected deviation, which exceeds the alarm limit or other predetermined limit, is not used for updating.

3. An arrangement as claimed in claim 1 or 2, characterised in that the arrangement is adapted to identify from the data obtained from monitoring at least the awake and sleep times and to create baseline patterns separately for each, that is, a personal baseline pattern for awake time and a personal baseline pattern for sleep time.

4. An arrangement as claimed in any of the claims 1 to 3, which is adapted to compare the data obtained from monitoring simultaneously to the personal baseline pattern and to at least one other predetermined pattern which represents the motoric cardinal symptoms of a particular fit or changes in physiological properties.

5. An arrangement as claimed in any of the claims 1 to 4, which comprises a wrist-held monitoring device.

6. An arrangement as claimed in claim 6, wherein the monitoring device is adapted to compare the baseline patterns with one another and the moni- tored data, and the monitoring device is furthermore adapted to generate an alarm.

7. An arrangement as claimed in any of the claims 1 to 6, which is adapted to monitor the movements of the monitoring device with at least one 3D or 6D acceleration sensor or sensors.

8. An arrangement as claimed in any of the claims 1 to 7, which is adapted to compute a baseline pattern on the basis of the frequency content of normal movements and/or the amplitudes of the frequencies.

9. An arrangement as claimed in claim 8, which is adapted to form a vector from the frequency and amplitude data, which represents the normal movements of the person, and to store several typical vectors of the normal movements, and/or to also collect data at the time of fits and to compute a feature vector from these, which represents the motor state relating to the fits, and is adapted to identify a fit by computing the distance between the normal vectors and the fit vector.

10. An arrangement as claimed in claim 9, the vectors created by which comprise time series measurement data or characteristic features computed from these, for example, as a frequency domain presentation, correlation values or classification values created by means of known pattern recognition and signal processing methods, such as the self-organizing feature map (SOM).

11. An arrangement as claimed in claim 1, which is adapted to detect and analyse with respect to a personal baseline pattern, using a correlation method or known pattern recognition and signal processing methods, such as the self-organizing feature map (SOM).

12. A software product for monitoring and predicting a fit, the software product comprising:

- means for reading data produced by a wearable, non-invasive monitoring device on a person's kinetic motor activity and/or physiological properties,

- means for creating a personal baseline pattern on the basis of long- term follow-up of monitoring data collected on the features of movements and/or physiological properties,

- means for analysing changes in the collected data with respect to the personal baseline pattern and for generating an alarm, at least when the difference to the personal baseline pattern exceeds a predetermined limit,

characterised in that

- the software product further comprises means for updating the baseline pattern continuously and for comparing the updated version of the baseline pattern with the history, whereupon from the changes in the baseline pattern compared to earlier baseline patterns can be detected changes that take place slowly.