Patent Grant
10682090
The subject matter disclosed herein relates to adverse event reduction in the absence of direct, continuous human observation and hands on interaction.
Pressure ulcers form when weight is concentrated in one or more areas over time, such as when a patient is relatively immobile. In particular, pressure ulcers develop when the skin's structure loses its healthy resilience and blood circulation is inadequate. Patients in a weakened condition laying in a bed where their body weight tends to concentrate in a particular location, such as, for example, a hip can be susceptible to developing pressure ulcers. While pressure ulcers for certain patient population are not avoidable, clinical guideline and practice suggest that moving the patient's body orientation from time to time allows the movement of body weight to other pressure points, lowering the probability of pressure ulcer development.
With this in mind, a pressure ulcer prevention or treatment protocol may stipulate that, for example, a patient is to be re-oriented every hour or at some other defined interval. An underlying assumption of such a protocol is that certain patients may be immobile and therefore are not moving on their own accord. An example case would be a sedated patient on a ventilator. When the risk for ulcers is clearly identified, care providers can implement and manage the pressure ulcer protocol(s), such as by routinely adjusting the resting position of the patient.
In addition, a patient who is susceptible to the condition but who is not properly identified as being so or who is not properly managed (whether or not properly identified as being at risk) may be susceptible to pressure ulcers or other adverse events due to failure to adhere to the appropriate care and prevention protocols. Hence, monitoring and implementation of such a protocol may be an issue.
In one embodiment, a protocol compliance sensing system is provided. In accordance with this embodiment, the protocol compliance sensing system includes a sensing system configured to detect skin surfaces of a patient using one or both of optical or range sensors, to detect movement patterns that concentrate pressure on the skin of the patient, and to generate an alert based on when the detected movement patterns are inconsistent with compliance with a pressure ulcer protocol.
In a further embodiment, a protocol compliance sensing system is provided. In accordance with this embodiment, the protocol compliance sensing system includes: a multimodality sensing system, comprising: a first sensing modality configured to detect skin surfaces of a patient using one or both of optical or range sensors, a second sensing modality configured to determine a centroid of mass of the patient; and a protocol adherence component configured to use both the detected skin surface and the centroid to detect movement patterns that concentrate pressure on the skin of the patient, and to generate an alert based on when the detected movement patterns are inconsistent with compliance with a pressure ulcer protocol.
In another embodiment, a patient management system is provided. In accordance with this embodiment, the patient management system includes: a classifying component configured to assign one or more patients into respective cohorts and to continuously or periodically assess placement of the one or more patients in the respective cohorts; a data integration component configured to receive and process one or more of protocol compliance data, patient characteristic data, or wound characteristics data from one or more sensor systems configured to monitor the one or more patients and to generate an assessment of each patients performance with a presently assigned cohort; and a communication component configured to provide feedback to the one or more patients based on the generated assessments.
These and other features, aspects, and advantages of the present invention will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
The present approach relates to approaches for assessing patient motion with respect to a protocol or other standard. In certain implementations, the motion is assessed using a multi-modality sensing arrangement suitable for sensing or assessing motion in various contexts that may be complementary to one another (e.g., providing qualitatively distinct motion information) such as measurements related to both a centroid of patient mass and a skin surface (external bound or boundary) of the patient. Further, sensor data from different modalities may be assessed in different manners so as to facilitate motion determination, such as using motion queues to determine relative motion of a surface in a computer vision context or a grid of squares or cubes to determine motion using ranging techniques (e.g., Doppler radar).
Such a system may be used to validate that a patient has been moved on a prescribed schedule by assessing and/or tracking the observed movement or activity relative to a protocol or schedule indicating the needed movement by the patient to avoid pressure ulcers or some other movement related condition. Non-compliance with the protocol or schedule may result in a notification, such as a message, alert, or other indicator, being provided to a caregiver. Such a notification may be made via a visual or audible notification channel or message prompting a caregiver to reorient the patient.
In certain implementations, motion sensing using systems as described herein may be used in the providing care over groups or cohorts of patients. By way of example, sensing and assessment techniques described herein may be used in the characterization of patients into different treatment or monitoring, in the ongoing assessment of maintaining or moving patients within or between such cohorts, and in approaches for incentivizing patient participation in protocol compliance efforts and/or self-healing or self-treatment programs. By way of example, movement and/or compliance information acquired using sensing systems as described herein may be used by computer-implemented treatment or assessment systems which may use one or more of patient's physical descriptors, indications of tissue health and/or healing, and care protocols, to determine the care plans and actions most likely to improve tissue health for the patient and/or for similarly attributed patients. Such systems may then group like patients into cohorts having similar or the same treatment regime and may monitor patient progress to assess whether a given patient is properly placed or should be moved to a different treatment cohort. Further, such approaches may provide feedback or incentives to patient to encourage patient compliance with treatment protocols and, in turn, patient healing.
Though a healthcare context is used in the present examples, the present approach may be utilized in other contexts where similar issues may arise, including but not limited to veterinary contexts, home health contexts, and mechanical apparatus contexts.
With respect to certain context provided below in explaining the present approach, the present disclosure incorporates the following references, in their entirety and for all purposes, as well as other references listed below: “SYSTEM AND METHOD FOR PROTOCOL ADHERENCE”, published as U.S. Patent Application No. 2012/0154582; “SYSTEM AND METHOD TO SCHEDULE RESOURCES IN DELIVERY OF HEALTHCARE TO A PATIENT”, issued as U.S. Pat. No. 8,027,849; and “SYSTEM AND METHOD TO MANAGE DELIVERY OF HEALTHCARE TO A PATIENT”, published as U.S. Patent Application No. 2009/0089093, some or all of which may aid in the understanding of certain concepts discussed herein.
In accordance with these approaches, various examples of motion measurement and assessment approaches are described below to provide useful context with respect to protocol adherence approaches as discussed herein. By way of example,
Using an example of a patient bed 102 (as seen from above in
In one such implementation, the placement and degree of the granularity defined by lines 120, 122 is a set point in the system. This placement and granularity of the grid 30, as defined by lines 120, 122 may be configured and/or adjusted to achieve a balance between detection specificity and acquisition of sufficient data for computation of momentum, motion, surface contour of the body and bed. Likewise, the relevant time units over which motion is assessed or aggregated relative to grid elements, and which define another dimensional aspect of the system, may be configured and/or adjusted to achieve a desired level of sensitivity and responsiveness.
By way of example, and turning to
In certain implementations, motion queues may be created when a relative change in position occurs (e.g., a change in the position of two patient surfaces (such as may be determined by a computer vision system) relative to one another, such as at flexible parts of the human body), as shown in
As noted above, in certain implementations the sensing and protocol adherence system may develop customized or patient-specific care plans or may classify different patients into different cohorts based on patient characteristics, where each cohort receives is handled in view of a protocol adapted to the characteristics shared by the members of the respective cohort. In such an approach, and turning to
These five factors which range from physical descriptors 90 of the patient 100, to a patient's mental state 88, actual motion 86 by body anatomical location, the care protocol 84 and clinical measures 82 of tissue health may be used to assess the comparative effectiveness of care plans or protocols, both at the patient and cohort level, so that the presently discussed method of improving tissue health is continuously improved in terms of weight density over time for areas of the body and patterns of loading for ever more specific guidance based upon these attributes. In certain implementations, regression and data mining are used to correlate the comparative effectiveness and designs of experiments to enrich these classifiers, to identify outliers, and to routinely reassess patient classifications into respective cohorts.
As noted above, with respect to the acquired and tracked data and its use in patient and cohort assessment, in certain implementations it is a goal to reduce the rate of clinical decline and to enable more rapid clinical indicator outcome improvement. With this in mind, and turning to
Three example initial condition clinical outcomes are described to illustrate the clinical outcome objectives of the disclosed invention. A first patient 72 presents with tissue health indicator 78 of 0.5, a second patient 70 presents with a 0.4 indicator, and a third patient 68 begins at 0.1. In the depicted example, third patient 68 has significantly less tissue health in a given anatomical location than first patient 72 does. As tissue health is a local indicator specific to an anatomical location, these temporal indicators are also local. The indicator can be a cumulative one for the patient or specific to one area of the body. There may be a plurality of concurrent indicators for the same patient. First patient 72 with comparatively superior tissue health on, as an example, their hip than third patient 68, it can be appreciated that the reverse could be the case for an upper body indication where third patient 68 has a tissue health measure in that area that is superior to that of first patient 72.
Using a first patient 72 as an example, in certain implementations it is a goal to avoid a rapid deterioration 66 of tissue health from which clinical recovery becomes significantly more challenging than in other contexts, such as a slow or gradual rate of decline 64. Said positively, a goal is to rapidly improve or increase 62 the clinical outcome at the highest rate achievable. With this in mind, assignment of a patient to a cohort and/or assignment (or customization) of a care protocol to a patient may be based on such considerations and taking into account other patient specific factors, such as those noted above, so as to achieve the best outcome.
It can be appreciated that the time scale 74 may be on the order of days for an acute inpatient condition and on the order of months or years for patients 100 in therapy or assisted living. A longer duration patient, for example, may be third patient 68 who is held stable (lines 60) over a year even at a reduced vitality, but does not degenerate. In such a scenario, the patient may be able to attain improvement (e.g., line 62) through monitoring and adherence to care protocols in response to actual clinical state, as discussed herein.
While tissue degradation is localized and specific to an individual patient 100, it can be appreciated that impacting (e.g., improving) a broader cohort's health may also desirable. For example, improvement of the health of a larger cohort may include cohorts such as some or all patients in a hospital, an insured population, residents in assisted living, or any other meaningful aggregation of patients (such as based upon similar physical or demographic characteristics, movement limitations, wound susceptibility, and so forth) where there is a desire to focus on pressure ulcer reduction tissue health improvement and an allocation of resources is made to effect the desired change.
With this in mind, turning to
In the depicted example, at a time period 40 later than an initial time period 46, the clinical conditions are again mapped by attribute for the patient population being monitored. Patients of like classifiers 48 who comparatively raise their clinical outcome an order of magnitude of statistical significance (e.g., patients 38) may be assessed and used to refine treatment, monitoring, and/or protocol aspects (e.g., setting the grid 30 sizes being monitored, the type of clinical measure, the frequency, pattern, and duration of rotation, the allowed weight loading and its duration, the effectiveness of patient room setup, and/or care aides such as foam fittings or vibratory or pressure distribution apparatus) for current or future cohorts of similarly attributed patients.
With the preceding in mind, certain motion-monitoring or assessment aspects of the present approach discussed with respect to
Linear Motion: A first motion aspect for use in conjunction with the present approach is the detection of linear motion. In one implementation, changes in displacement of the patient body parts are optically sensed using a sensing system. Similarly, height profile change may be measured using a ranging aspect of the system (i.e., a ranging system). Body mass placement may be inferred from augmentation of the optical and/or range derived position (using the optical and/or ranging systems noted above) with the incorporation of sensors such as load cells and pressure pads, and so forth. Such measurements in this context, regardless of the sensing mechanism(s) employed, may constitute measures of linear motion.
For example, in an optical sensing (e.g., computer vision) implementation, the exact surfaces of skin and the orientations of limbs and digits of the body are detected with respect to the geometry of the bed and related apparatus. This is in contrast to other approaches, such as pressure pads, load cells, RFID, and centroid detection in computer vision, which all merely approximate relative location of the patient based on some surrogate measure (e.g., a centroid or central mass measurement). In this manner, the actual placement or position of the patient's boundary (e.g., skin surface) may be determined as opposed to a more generalized measure of patient placement, such as a center of mass. In such an implementation, the optical sensing approach enables precise spatial relationships to be determined and evaluated based on the physical boundaries of the patient surfaces.
It can be appreciated that for optical methods that resolve surface location and movement, a patient in bed may have occluded visibility such as from blankets and thus the person's physical boundaries may not be resolvable. This limitation may be overcome by using the shape profiles of what can be observed and momentum vectors, if available, to infer with statistical and machine learning techniques as to what the skin contact locations are as learned on the present and prior humans which have been used to train the system.
Another sensing approach is Doppler-based, which may provide range or elevation type information corresponding to patient surfaces or boundaries. A Doppler-based sensing device may be placed above the patient 100, such as in a light fixture, above a ceiling tile, or mounted on a ceiling or wall or behind a wall. The Doppler-based system may also be placed below the patient or aside the patient. Depending on the Doppler-based sensing device, there may be multiple sensing elements (i.e., sensors) and the sensors may be coupled, as in the computer vision based sensing discussed above, to resolve the precise boundaries of patient surfaces, such as skin, and the exact geometrical location of such surfaces.
In response to the actual physical boundaries of patient surfaces and the monitored motion of such surfaces, and in view of defined pressure ulcer prevention protocol definitions and sequences, one or more instructions or notifications may be automatically issued to control mechanical or computational assets or prompt care providers.
Turning back to
The zones defined by lines 120, 122 are defined virtually and the patient 100 is tracked within these zones and as their determined surfaces traverse different zones, such as for example a hand 124 moving to zone 128 from zone 126. In one implementation, the movements of a patient 100 are detectable by the superimposition of the patient's body parts within these zone areas and between these zones. For the case of optical sensing, the patient body parts visible to the sensing system are used and in the case of range sensing, the heights of the patient's body parts may be used. In addition to the optical and range sensing observable location, the incorporation of other sensing modalities is used if available, such as Doppler, load cells and pressure pads, as discussed in greater detail below.
In an implementation employing an optical sensor to detect linear motion associated with one or more body parts, a spatio-temporal image filter may be applied to the body parts imaged by the sensor to detect the localized extremes of the filtering response pattern. In one such example, the spatio-temporal image filter is biologically inspired (i.e., is based on how biological systems sense and interpret motion) and engineered to resemble how human visual cortex system detects and recognizes human body motions and actions.
A spatio-temporal image filter may be implemented based on a stationary sensor setup, as shown in the image sequence of
Thus, as shown in
Turning to
The matrix of zones are evaluated, for a given observation period, to encode the patient's body parts movement within the matrix (e.g., grid). In an example embodiment, the zones where movement has occurred are recorded and accumulated along the y-axis 202 of
With this particular implementation in mind, it may be appreciated that as data from optical, Doppler, range, pressure pad, and/or load cells are collected, these data collection operations may be performed in continuous time at a prescribed sampling rate, across the finest granularity of sensing area for discrimination. In such an implementation, this raw data is then integrated over an adjustable time period over a post processed zone configuration, i.e., forming a spatio-temporal zone cue for feature data computation generated from the raw data. This may also facilitate pattern discovery and data mining for future set-points of zones and time differentials as well identifying precursor patterns that can be used to define or initiate an alarm condition for future pressure ulcer protocol settings.
Further, as noted above, in certain implementations the bed zones defined by lines 120, 122, an optional height designation or granulation (e.g., elevation 121) and the parts of a patient's body within a particular zone may be used to further characterize or classify which anatomic part(s) of the patient's body are moving. Thus, for example, the upper torso data 206 may be differentiated from the right hip data 208 in the tracked and graphed data 200. The patient's body parts preferably do not remain in the same zones defined by lines 120, 121, 122. Instead, it is expected or desired that a patient body would traverse many zones over time, thus indicating movement. Particular anatomic parts or joints of a patient's body may thus be monitored by specific tracking or by means of pattern recognition to uncover patterns of under-movement (i.e., insufficient movement with respect to a define protocol) by matching against an absolute or comparative cases from other like attributed patient data. Should insufficient movement be identified for a monitored body area (e.g., torso data 206, hip data 208, and so forth) relative to a set point 210 (the set point 210 being defined on an absolute or comparative basis), a clinical workflow 212 (shown in
As shown in
Load Cells: In a second motion-related aspect, data from load cells provided as part of or with the patient bed 102 may be integrated into the measuring system. By way of example, load cells may be provided as part of the bed 102, such as above the wheels, under the wheels, and/or as part of the frame, as shown in
By way of example of a load measuring operation using such structures, and turning to
Positioning across the sensing mechanism uses the measures (e.g., pressure point 270, point load change 279, centroid 248 of mass, and/or motion queues 278, and not the absolute support surface x-y plane. The composite motion queue may be derived by changes in optical image pixilation of body features or Doppler ranges and is appended to by load points 279 from grid-based displacement measuring means and by mass centroid velocity, direction, and momentum, or specific skin surface velocity, direction and momentum.
With this in mind, in certain implementations where load cell data is employed, differential weight changes per unit time are employed to calculate the patient body weight centroid. The weight centroid may then be mapped with respect to the virtual zones (i.e., grid 30) defined by lines 120, 122. Inferences are achieved to relate centroid movement 248 to specific surface movement such as, for example, at a knee 281, using machine learning and deep learning.
In certain approaches the load cells 244 may be used without the benefit of optical or ranging sensing. However, in certain of the present implementations, the load cell information is collected along with the optical and ranging data discussed above to augment or establish movement patterns or trigger alarms, such as based upon the combination of centroid and skin surface data. The coordinate systems for the load cell and optical sensing or ranging data may be spatially aligned (such as by using a shared grid system defined by lines 120, 121, 122) to facilitate the combined use of such datasets. Alternatively, the outputs or derived measurements from the various sensing modalities may be used to jointly make a movement determination, such as by averaging or using a voting (e.g., majority voting) scheme.
By way of example using the computer vision modality, though it can be appreciated that Doppler based range sizing would store these attributes, and turning to
In the depicted example, the data provided by these modalities are augmented by the inclusion of load cells 244, as discussed above, which in the current example reconcile the body mass centroid 248 of a patient 100. Using a virtual grid 280 of the bed which is comprised of adjustable areas, the mass centroid 248 is calculated based on load cell measurements and is superimposed onto the geospatial map of the bed used by the optical and range sensing modalities and is accumulated over time. The integration of the load derived centroid 248 with respect to the explicit anatomic surface locations as derived by one or both of the optical 250 or ranging systems 256 can be accomplished through probabilistic sensor fusion 282 or by other suitable approaches. The principal behind this fusion is that the weight of a person is considered as a function of their body part mass distribution and orientation, which are computable from the optical and/or range sensing modalities. Over time, the integration of the mass derived centroid 248 and the exact optical and/or range sensing placement can be accomplished using the disclosed system. A centroid resolution for momentum estimation and exact surface characterization with range or computer vision are one suitable implementation.
With the preceding in mind, in this example an optical 250 and/or ranging 256 aspect computes placement of a body shape 284 (e.g., a limb or joint) (shown in
In one embodiment, as shown in the graph of
Turning to
Pressure Pad Load Sensing: In a third motion-related aspect, pressure pad load sensing for the bed 102 may be integrated with other data for assessing pressure ulcer protocol compliance. In particular, pressure pads 320 are typically laid on the mattress, under the patient 100. As has been described above, aspects of the present approach may rely on or incorporate differential weight changes per unit time, determined using pressure pad load sensing, and map such changes from the monitored zones of the pad 320 to the zones defined by lines 120, 122 discussed above. In such an approach, load cells may be used to calculate location, velocity and pressure change as opposed to momentum change.
In one such approach the pressure pad information is collected along with one or both of optical 250 or ranging 256 data, in which case any of the monitored sensing modalities could establish movement patterns or trigger alarms. The orientation of the pressure pad's 322 load (pressure) sensing compared to the optical 250 or range sensing 256 location may be physically established by its mechanical orientation on the bed 102. Alternatively, the plurality of sensing modalities may be averaged or use a majority voting scheme to set the final determination.
Motion Queues: In a fourth motion-related aspect, motion queues may be employed to establish sufficiency of patient activity. The optical 250 and ranging 256 sensing modalities use changes of state of a unit of area per unit time to determine motion. This aspect is particularly useful in establishing movement at the intersection of body parts such as knee, hip, elbow, neck joints, torso twisting and head movements. In certain aspects, motion queues may be employed is assessing patient motion and to assess compliance with a pressure ulcer protocol. By way of example, the angular velocity for motion queues may be summed per unit time, per zone and then tabulated and reasoned over as discussed above, by leveraging machine learning approaches to discover the patterns of movements from the training data. The pressure ulcer protocol, requiring a shift of weight per unit time, is thus enabled.
In one particular embodiment, the optical sensor 250 is used to compute the motion queues and their associated motion activity sample data generated from this queue is illustrated in
Pattern Recognition and Machine Learning: In another motion-related aspect, patterns within zones and motion queues may be detected. These patterns may be used to further train pattern detection over time and/or over locations (such as between hospitals) via observed historical patient attributes to recognize the precursors of pressure ulcer formation. This learning aspect may determine how granular to set zones defined by lines 120, 122 based on differing patient characteristics (e.g., age, gender, and so forth), what the time differential is for movement detection and the alarming thresholds, and so forth by training the algorithms with the data collected over multiple hospitals across a variety of patient population.
Technical effects of the invention include a system that detects a movement or lack of movement (or pattern of movements) corresponding to a protocol of care, such as that prescribed by a medical doctor or a therapist and then triggers a response if the protocol is not adhered to. Movements may be detected and quantified either to one or more protocols or on an absolute measured movement basis. In one implementation, there may be multiple (e.g., 3, 4, 5, or 6) motion related aspects, each corresponding to a movement pattern associated with a patient being in a hospital or other bed.
This written description uses examples to disclose the invention, including the best mode, and also to enable any person skilled in the art to practice the invention, including making and using any devices or systems and performing any incorporated methods. The patentable scope of the invention is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insubstantial differences from the literal languages of the claims.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5993400 | Rincoe et al. | Nov 1999 | A |
| 6384728 | Kanor et al. | May 2002 | B1 |
| 6544200 | Smith et al. | Apr 2003 | B1 |
| 6897781 | Cooper et al. | May 2005 | B2 |
| 7030764 | Smith et al. | Apr 2006 | B2 |
| 7378975 | Smith et al. | May 2008 | B1 |
| 8027849 | Johnson et al. | Sep 2011 | B2 |
| 8998830 | Halperin et al. | Apr 2015 | B2 |
| 20040046668 | Smith | Mar 2004 | A1 |
| 20090099480 | Salgo et al. | Apr 2009 | A1 |
| 20110043630 | McClure | Feb 2011 | A1 |
| 20120026308 | Johnson | Feb 2012 | A1 |
| 20120065547 | Hann | Mar 2012 | A1 |
| 20120075464 | Derenne et al. | Mar 2012 | A1 |
| 20120078144 | Sinykin | Mar 2012 | A1 |
| 20120154582 | Johnson | Jun 2012 | A1 |
| 20130091631 | Hayes et al. | Apr 2013 | A1 |
| 20130246088 | Huster et al. | Sep 2013 | A1 |
| 20130317399 | Ribble et al. | Nov 2013 | A1 |
| 20140257057 | Reis Cunha et al. | Sep 2014 | A1 |
| 20150109442 | Derenne | Apr 2015 | A1 |
| 20150141762 | Heinrich | May 2015 | A1 |
| 20150254956 | Shen et al. | Sep 2015 | A1 |
| Number | Date | Country |
|---|---|---|
| 1762856 | Apr 2006 | CN |
| 2 014 892 | Jan 2009 | EP |
| 2 392 304 | Dec 2011 | EP |
| 2 715 695 | Apr 2014 | EP |
| 2009-125812 | Jun 2009 | JP |
| 2007149934 | Dec 2007 | WO |
| Entry |
|---|
| Xu, Chi et al. “Mimo space-delay adaptive processing for wide-area indoor human motion classification.” IEEE Transactions on Aerospace and Electronic Systems 52.3 (2016): 1384-1398. (Year: 2016). |
| “Smart mattress cover prevents bed sores”, The Engineer, Retrieved from Internet URL: http://www.theengineer.co.uk/news/smart-mattress-cover-prevents-bed-sores/1003138, on Nov. 27, 2013, p. 1 (Jun. 24, 2010). |
| Xu, C., et al.,“MIMO Space-Delay Adaptive Processing for Wide-Area Indoor Human Motion Classification”, IEEE Transactions on Aerospace and Electronic Systems, vol. 52, Issue 3, pp. 1384-1398 (Jun. 2016). |
| Invitation to Pay Additional Fees issued in connection with corresponding PCT Application No. PCT/US2017/037598, dated Sep. 21, 2017. |
| International Search Report and Written Opinion issued in connection with corresponding PCT Application No. PCT/US2017/037598, dated Nov. 21, 2017. |
| Kittipanya-Ngam, Panachit, et al.; “Computer vision applications for patients monitoring system”, Information Fusion (FUSION), 2012 15th International Conference on, pp. 2201-2208, Jul. 9-12, 2012, Singapore. |
| Vilas-Boas, Maria C., et al.; “Monitoring of bedridden patients: Development of a fall detection tool”, Engineering in Medicine and Biology Society (EMBC), 2013 35th Annual International Conference of the IEEE, pp. 4742-4745, Jul. 3-7, 2013, Osaka. |
| Number | Date | Country | |
|---|---|---|---|
| 20180000407 A1 | Jan 2018 | US |
