Early exertional heat stroke detection system and method for real time situational awareness and illness risk assessment

Abstract

A system having a combination of several sensors that measure the following physiological measurements: heart rate (HR), skin temperature (Tsk), and movement, and a method receiving the system data and using the interplay of these data to reveal individual's signature(s) for use in a multi-factor assessment to provide contextual and clinically relevant information to determine the individual's Exertional Heat Stroke (EHS) risk. When the risk is above a threshold alerting the monitored individual or somebody else for the monitored individual's benefit to take action to lower the EHS risk and thereby lessen the associated consequences of EHS.

Description

Early Exertional Heat Stroke Detection System and Method for Real Time Situational Awareness and Illness Risk Assessment

[0001] This patent application claims the benefit of U.S. Patent App. No. 63 / 539,629, filed on September 21 , 2023, which is hereby incorporated by reference.I. FIELD OF THE INVENTION

[0002] In at least one embodiment, a system and / or a method utilizes a combination of several sensors that measure the following physiological measurements: heart rate (HR), skin temperature (Tsk), and movement (e.g., steps), and using the interplay of these signals to reveal the individual’s signatures for comparison to Exertional Heat Stroke (EHS) signatures as part of a multi-factor assessment, and to provide contextual and clinically relevant information to determine EHS risk to allow for an intervention to be taken by the monitored individual or by another individual for the monitored individual’s benefit to lower the EHS risk and thereby lessen the associated consequences of EHS.II. BACKGROUND OF THE INVENTION

[0003] As global temperatures continue to rise, both the risk for, and the severity of, exertional heat illness (EHI) is likely to increase simultaneously. This threat is particularly great for populations who regularly put themselves in high-risk situations, such as outdoor workers, athletes and military personnel, who must carry out physically demanding tasks during prolonged exposure to hot environments. As one example, over the years there have been several football players that have suffered from an EHI case and even have succumbed to death from outdoor preseason training. It is a regular occurrence to see athletes in outdoor events to suffer leg cramps during games in hot and humid conditions. Military training and operations in hot and / or humid conditions must continue to occur to prepare active-duty military personnel for deployment-relevant environments with similar conditions. Therefore, there is an indispensable need to develop and implement effective heat mitigation action plans to protect individuals against the heightened risk level for heat illness. There is a need for earlier warning mechanisms to allow for earlier interventions than current approaches.

[0004] EHI occurs along a continuum, varying in severity and symptoms, that are a result of physical exertion in hot environments. Milder forms of EHI include heat cramps and heat exhaustion and can progress into more severe illness, such as EHS, which can result in severe and / or long-term organ damage and even death. Unfortunately, in the U.S. multiple individuals die from EHS each year. The risk of death can be greatly reduced if not eliminated with prompt recognition and aggressive cooling. While increased focus on these issues in recent years has produced an array of guidelines and procedures that have shown some success in preventing heat illness, there are gaps that remain unresolved. One of the most pressing issues is finding the appropriate balance between conducting essential training to develop the operational capabilities needed for hot environments, while protecting military personnel and / or civilians (e.g., outdoor workers and athletes) from heat illness. Activities that pose a high-risk for heatillness are routinely performed in congruence with risk mitigation tools, such as work-to-rest guidelines based on combinations of environmental conditions (heat, humidity) and exercise intensity. However, even accurate measures of environment and activity (which are not always available) do not consider other important factors that can alter an individual’s ability to effectively dissipate heat, such as clothing, load carriage, and type of activity.

[0005] An additional challenge is the extent of inter- and intra-individual variability in risk factors for EH I. While some reports indicate certain risk factors were positively correlated with occurrence of heat illness, other studies have found no correlation of the same risk factors with EHI. The growing body of evidence demonstrating complex relationships among risk factors and physiological responses to heat further supports the argument that a better understanding of pathophysiological and inter-individual risk factors would better allow the development of more comprehensive heat mitigation plans that would not inhibit operational capability and performance. One proposed strategy would be to use wearable devices, such as physiologic status monitors to provide real-time physiological information regarding thermal strain. The emergence of physiologic status monitors in recent years has shown promise to monitor thermal strain during high-risk events. However, using real-time thermal strain measures to determine physiological patterns that are both present in all cases of EHS and are clinically significant has been difficult due to the lack of data prior to treatment and collapse from EHS.

[0006] EHS is a life-threatening illness typically characterized by a rapid and abnormal increase in core body temperature (Ter) (often higher than 40.5°C), central nervous system (CNS) dysfunction, and organ and tissue damage. Bouchama A. et al., Heat stroke, New England Journal of Medicine 2002, 346(25):1978-1988; Leon LR et al., Heat stroke, Comprehensive Physiology 2015, 5(2):61 1-647; Shibolet Set al., Heat stroke: a review, A viation, Space, and Environmental Medicine 1976, 47(3):280-301 . Recent studies have demonstrated that the interaction between different components, such as exercise type, environmental conditions, clothing, and other individual factors can increase both the incidence and severity of EHS. Buller, M.J. et al. (2018), “Wearable physiological monitoring for human thermal-work strain optimization”, Journal of Applied Physiology, 124(2), 432-441 . An approach to reduce the incidence and severity of EHS is to utilize on-body sensors to identify individuals with prodromal signs of emerging illness and refer them for medical care before the illness can further progress further. Buller MJ et al., “Real-time core body temperature estimation from heart rate for first responders wearing different levels of personal protective equipment”, Ergonomics 2015, 58:1830-1841 , doi:10.1080 / 00140139.2015.1036792; Muniz-Pardos et al. (2019), “The Use of Technology to Protect the Health of Athletes during Sporting Competitions in the Heat, Frontiers in Sports and Active Living", PERSPECTIVE 2019, doi:10.3389 / fspor.2019.00038.

[0007] Despite all the research, there is still no uniformly recognized way to consistently project when EHS is about to happen and thus preventing earlier interventions. The lack of a tool is becoming even more critical as there is a general warming trend around the globe withhigher than normal temperatures regularly happening with individuals still doing a variety of tasks outdoors from walking, hiking, climbing, jogging, running, marching, cycling, skateboarding, enjoying the outdoors, athletic training, playing sports, construction work (e.g., road work and building construction), landscaping, search and rescue missions, and a variety of other tasks. In addition, EHS can happen in indoor environments, for example firefighting and repair work in open garages or underground environments. Also, EHS can arise when the individual is performing tasks in a protective suit such as a HAZMAT or personal protective equipment.III. SUMMARY OF THE INVENTION

[0008] EHS remains a stubborn challenge for military communities around the world. Severe EHS can lead to long-term organ damage and even death without rapid treatment. EHS often occurs during events with high metabolic demands coupled with a high level of individual motivation. Physiological monitoring may provide early warning that an individual is at risk of EHS. To optimize physiological monitoring, how the physiological response of individuals who suffer an EHS differs from those who did not needs to be understood. The purpose of the underlying work that led to one or more embodiments of the invention was to compare the physiological profiles of individuals who experienced EHS with those of others completing the same training event, who did not experience EHS. These are discussed later in connection with seven EHS and six EHS cases.

[0009] But such a system requires physiological signatures of heat stroke coupled with contextual information. According to different embodiments, a combination of sensors, and physiological signatures of exertional heat stroke together provide a real-time assessment of exertional heat stroke risk before the individual exhibits symptoms and thus allow for an earlier life-saving intervention. The need for earlier and quicker interventions necessitates the need to use a computer to analyze the sensor data in real time.

[0010] In at least one embodiment, the system and the method utilize a combination of several sensors that measure the following physiological measurements: heart rate (HR), skin temperature (Tsk), and movement and / or accelerometry, and using the interplay of the sensor outputs to obtain signatures to be compared to known EHS signatures to provide contextual and clinically relevant information to determine EHS risk. When the EHS risk is above a threshold, alerting the individual or someone else of the risk to allow for an intervention to take place.

[0011] In at least one embodiment, three signatures are used as a warning preceding collapse from EHS that build the confidence of the EHS determination:• differential response of heart rate and skin temperature of EHS individuals compared to controls during exercise,• the conflicting response of the slowing of movement but an increase in heart rate minutes prior to collapse, and• the conflicting response of the slowing of movement but a decrease in skin temperature minutes prior to collapse.These signatures are combined into a comprehensive likelihood model to predict a risk of a heat stroke for the monitored individual. In at least one embodiment, these models may be used individually or with one of the other two models.

[0012] The invention is as defined in the independent claims. Further advantageous are set out in the dependent claims and may be as described hereinbelow.

[0013] In at least one embodiment, a method for detecting exertional heat stroke (EHS) signatures for an individual, the method including: monitoring or receiving a heart rate for the individual from a heart rate sensor; optionally storing the heart rate over time in a memory; monitoring or receiving a skin temperature for the individual from a skin temperature sensor; optionally storing the skin temperature over time in the memory; comparing the heart rates and the skin temperatures to an expected set of heart rates and skin temperatures that optionally are based on a similar activity completed by the individual previously; a published model that predicts responses to activity, environment, and / or clothing; and / or other members of a group including the individual that is performing the activity to determine a probability that a rate of change for the heart rate for the individual and / or a rate of change for the skin temperature for the individual falls outside the expected set of heart rates and skin temperatures; monitoring or receiving an activity level for the individual from an activity monitor; optionally storing the activity level over time in the memory; comparing a trend of the heart rates to a trend of the activity levels to determine a probability that the two trends are divergent from each other; comparing a trend of the skin temperatures to the trend of the activity levels to determine a probability that the two trends are divergent from each other; combining the three probabilities together optionally using predetermined weights to obtain an overall probability for EHS for the individual; alerting the individual or another individual when the overall probability is above a risk threshold, trending towards a projected combined probability representative of EHS, or a rate of change for the combined probability exceeds a change threshold; and optionally performing an intervention for the individual in response to the alert.

[0014] In an alternative embodiment, a method for detecting signatures for an individual based on physiological data for comparison to EHS signatures, the method including: receiving a plurality of hearts rates, skin temperatures, and activity levels for the individual; inputting the heart rates, skin temperatures, and activity levels for a monitoring window into a model having sub-models for determining a signature of the individual based on an existence of a differential response of 1 ) the heart rate and the skin temperature, 2) the heart rate and a temperature gradient between skin temperature and estimated core temperature, or 3) temperature gradient between skin temperature and estimated core temperature of EHS individuals compared to controls during exercise or activity, a conflicting response of the slowing of movement but an increase in the heart rate minutes prior to collapse, and a conflicting response of the slowing ofmovement but a decrease in the skin temperature minutes prior to collapse; combining the outputs of the sub-models to determine an overall probability that the individual will have EHS based on a comparison between the signatures and the EHS signatures; and optionally performing an intervention for the individual in response to the alert.

[0015] In another embodiment, a method for alerting an individual of a risk of an EHS, the method including: receiving heart rate data, skin temperature data, and activity data for the individual; creating a signature for the individual based on the received data; providing the signature to a model configured to determine the risk of an exertional heat illness for the individual, wherein the model includes up to four sub-models each having a signature based on previously received heart rate data, skin temperature data, and activity data for the individual or similarly situated individuals heart rate data, skin temperature data, and activity data representative of individuals that have suffered EHS or are a control group; determining the risk of the EHS based on a comparison of signatures; outputting the risk and recommending an intervention when the risk is above a threshold, and wherein similarly situated individuals have a similar physiological and / or fitness makeup to the individual being monitored.

[0016] Further to the prior method embodiments, comparing the heart rates and the skin temperature uses a probability model according to p(zt) = W(zt|Azt:t-ls, r) and p(yt) - N(yt|Byt:t-15, A) where zt-and B are the regression weighting coefficients and T and A are the covariance matrices, HR is heart rate, and Tsk is the skin temperature. In a further embodiment,dt = W, dHR + MA and dt = W1d7’sfc+ Wowhere Woand Wi are linear regression weights for offset and a coefficient in change of heart rate. Further to the previous two embodiments, data for the regression weighting coefficients and the covariance matrices are based on data for a group performing a given task or activity where optionally the group is representative of the individual being monitored in terms of physiology characteristics. Optionally, F and A represent a mean response distribution that may be based on an activity being performed, a type of clothing being worn, and an environment.

[0017] Further to the prior method embodiments, comparing the trend of the heart rates to the trend of the activity levels includes using a Gaussian likelihood model which optionally may have an offset over time between the activity level and the associated heart rate for a particular time t. In a further embodiment, Gaussian model is p(mt|nt) = A / (mt|Cnt.t-15, E) where mt is heart rate, n is a step count, C is a regression weighting coefficient, and E is a covariance matrix. The Gaussian model optionally is a prediction of the likelihood of observing a particular heart rate at time t given a certain step count at time t. Optionally, the regression weighting coefficient and the covariance matrix are based on data for a group performing a given task or activity being performed by the individual where optionally the group is representative of the individual being monitored in terms of physiology characteristics.

[0018] Further to the prior method embodiments, comparing the trend of the skin temperatures to the trend of the activity levels includes using a Gaussian likelihood model which optionally may have an offset over time between the activity level and the associated skin temperature for a particular time t. In a further embodiment, where the Gaussian model is p(ot|nt) = N(ot|Dnt;t-15, Z) where Ot is skin temperature, n is step count, D are the regression weighting coefficients, and Z is the covariance matrix. Optionally, the step count n is over a subwindow of a larger time window, optionally sub-windows of 5 seconds, 30 seconds, or 1 minute in a larger window of 5 minutes, 10 minutes, or 15 minutes, respectively. Optionally, the Gaussian likelihood model provides a prediction of the likelihood of observing a particular skin temperature at time t given a certain step count at time t. Optionally, the regression weighting coefficient and the covariance matrix are based on data for a group performing the given task or activity being performed by the individual where optionally the group is representative of the individual being monitored in terms of physiology characteristics.

[0019] Further to the prior method embodiments, the overall probability is determined using the following model p(EHS+10|zt,yt,mt, nt, ot) = w^l - p(zt)) • w2(l - p(yt)) • w3(l - p(mt|nt)) ■ w4(l - p(ot|nt)) where wi, w2, w3, and w4are weights, p(zt) and p(yt) are models for heart rates and skin temperatures over time, p(mt|nt) is a model for heart rate-activity trends, and p(ot|nt) is a model for skin temperature-activity trends. Optionally, the weights are equal weights for each probability, one or more weights are heavier than one or more other weights, or the weights that vary over time to emphasize some probabilities over other probabilities depending on the length of the activity being performed by the individual. Alternatively, the weights are variable such that during an earlier part of the activity, the weights for the first two probabilities are weighted more heavily to reflect that the factors of these probabilities being more likely to provide earlier warnings than the other probabilities; and as the activity progresses, the weights are adjusted to increase the weight of the last two probabilities more heavily as these probabilities are better indicators of EHS closer to when EHS may occur.

[0020] In at least one embodiment, a system including: a heart rate monitor or sensor; a skin temperature sensor; a movement sensor; a housing with the heart rate monitor or sensor, the skin temperature sensor, and the movement sensor; an optional communications component; and a processor configured to perform any of the prior methods except performing an intervention. In a further embodiment, the system having a chest strap attached to the housing. Optionally, the housing includes the communications component. Optionally, the movement sensor is a tri-axial accelerometer. Optionally, the system including a central server in communication with a plurality of sensors on different individuals and configured to receive sensor data for aggregation into signatures for determining the risk of EHS.IV. BRIEF DESCRIPTION OF DRAWINGS

[0021] FIGs. 1A and 1 B illustrate block diagrams of systems according to different embodiments of the invention.

[0022] FIG. 2 illustrates a flowchart according to at least one embodiment of the invention.

[0023] FIG. 3 illustrates the heart rate (beats per minute (BPM)) for the EHS cases and the control cases by segment of the activity being performed.

[0024] FIG. 4 illustrates the step count for the EHS cases and the control cases by segment of the activity being performed.

[0025] FIG. 5 illustrates the core temperature for the EHS cases and the control cases by segment of the activity being performed.

[0026] FIG. 6 illustrates the skin temperature for the EHS cases and the control cases by segment of the activity being performed.

[0027] FIG. 7 illustrates the temperature gradient for the EHS cases and the control cases by segment of the activity being performed.

[0028] FIG. 8 illustrates the heart rate and chest skin temperature responses of EHS case versus controls completing loaded ruck marches.

[0029] FIG. 9 illustrates an example EHS case from Iteration 31 , conducting a loaded 12- mile ruck march showing, HR, Tsk, ECTemp, and step count (steps / min).

[0030] FIG. 10 Illustrates an EHS case from Iteration 12, conducting a loaded 12-mile ruck march showing, HR, Tsk, ECTemp, and step count (steps / min).V. DETAILED DESCRIPTION OF THE INVENTION

[0031] In at least one embodiment, the use of the context of movement speed and the physiological responses of the group are used to identify the EHS cases as increasingly likely as their physiology responds contrary to the norms allowing for detection of EHS signatures.

[0032] In at least one embodiment as illustrated in FIGs. 1A and 1 B, a system includes a heart sensor 110, a skin temperature sensor 120, and a movement sensor 130 such as a tri- axial accelerometer that are connected to a processor 140, which may be separate from the system housing the sensors but in communication with the sensors, for example through a communications module 150. An example of a system that may be worn is a chest (or mid-torso) strap to improve the skin temperature readings and step counting as oppose to a wrist worn watch or a phone residing in a pocket or being carried. In at least one embodiment, the heart sensor 110, the skin temperature sensor 120, the movement sensor 130 and the optional communications module 150 are in a housing 100. The processor 140 uses the data provided by these sensors to assess at least three signatures (or factors) predicting collapse from EHS in sufficient time to allow for an intervention to occur to lessen the risk that EHS may occur and thus to potentially avoid the consequences of EHS. Examples of interventions include reducing the pace of activity or stopping the activity, adjusting the clothing being worn by the monitored individual, providing fluids to and / or pouring fluids over the monitored individual, providing ice orother cooling material or equipment to the monitored individual, placing the monitored individual in an ice bath or cooler temperature water and / or environment, and seek medical care. The cooling approaches may be thought of as examples of aggressive cooling.

[0033] In at least one embodiment as illustrated in FIG. 2, a method includes monitoring an individual (i.e., the monitored individual) using a system like that of the previous paragraph to collect (or receive) heart rate, skin temperature, and activity data over time, 205, to create at least one signature for comparison to EHS signatures, 210. In an alternative embodiment, the acidity data is converted into step data, if not provided in that format, based in part on the individual’s height. Using up to three feature (or factor) models, determine a probability of onset of EHS, 215, and thus to potentially avoid the consequences of EHS by recommending an intervention (e.g., providing an alert), 220. The models may operate in parallel to each other with models providing their respective outputs for use in the probability determination. The alert being provided to someone other than the person being monitored may be beneficial because the EHS individual may have an altered mental status from the situation and not able to act.

[0034] The first factor model provides a probability that the change in heart rate and skin temperature over time differs from non-EHS individuals, each of which are an example of a factor component. In an alternative embodiment, the first factor model is split into two separate factor models. The second factor model provides a probability based on changes in the heart rate as compared to the change in the level of activity (e.g., the number of steps) such that the probability increases when the heart rate increases while the level of activity slows, and this probability is an example of another factor component. The third factor model provides a probability based on changes in the skin temperature as compared to the change in the level of activity (e.g., the number of steps) such that the probability increases when the skin temperature decreases while the level of activity slows, and this probability is an example of another factor component. These three factor models are combined together with predetermined weights to provide an overall probability that the monitored individual is at risk of EHS before EHS occurs. Examples of the weights include, but are not limited to, equal weights for each factor, a heavier weight for one or more factors over the other factors, and weights that vary over time to emphasize some factors over other factors depending on the length of the activity. For example, during the earlier part of the activity, the output(s) of the first factor model may be weighted more heavily to reflect that the factors of this model are more likely to provide earlier warnings than the other factors; and as the activity progresses, the weights may adjust to weight the outputs of the second and third factor models more heavily as these are better indicators of EHS closer to when EHS may occur. In an alternative embodiment, the weights are omitted as separate components and are instead incorporated into the individual factor models.

[0035] Data were collected from 918 (913 men and 5 women) U.S. soldiers and marines during one of seven high-risk military training activities which included a pass / fail component. These activities, all of which had at least one individual who suffered EHS during the event,included one 7-mile ruck (n=157); two 12-mile rucks (n=175); the Crucible event, which consisted of one 3-mile ruck and one 9-mile ruck (n=483); and one 10-mile ruck (n=103). The 7 and 12-mile ruck marches were part of the Ranger Assessment Selection Program (RASP) and were required to be completed in under 2 hours for 7-miles and under 3 hours for 12-miles. The 10-mile ruck was a training requirement for the Marines Infantry Officer Course (IOC). Finally, the 3-mile ruck and 9-mile ruck were a part of the Crucible event, which is the final training requirement for Marines in basic training. Demographic information from each volunteer was collected as follows: age, gender, place of birth, height and weight, most recent physical fitness test run time (2-mile run for soldiers, 3-mile run for marines), history of heat illness, fluid and food intake within the last 12-hours, whether they were feeling unwell or had been ill within the last 60 days, if they had received any immunizations within the last 30 days, and any medications and dietary supplements they were currently taking.

[0036] Real-time measurements of heart rate (HR), skin temperature (Tsk) and accelerometry (128 HZ) were collected using a torso-worn physiologic monitoring system (Heat Illness Prevention System (HIPS); Odic®, Littleton, MA). Three additional variables were calculated post-hoc using the system-measured variables. Estimated Core Temperature (ECTemp) was determined, and examples of how it may be determined include the approaches described in U.S. Pat. No. 10,702,165 and U.S. Pat. App. Pub. No. 2022 / 0125388 A1 , and in an alternative embodiment, the core temperature could have been measured directly by use of a temperature pill. Core-to-skin temperature gradient (Tect-Tsk) was calculated by subtracting Tsk from ECTemp. Step count was included as a measure to assess if the individuals who suffered EHS were exerting themselves more, as step count can be used as a correlate for exertion level and movement speed. Step count was calculated by multiplying the magnitude of the accelerometry data by estimated stride length (0.415 x subject height (m)). Group means for each of these variables were obtained by averaging the data from all individuals for their respective groups together to obtain a single data set representing the entire group.

[0037] To analyze how group means differed from each other, each event was divided into equidistant segments based on the total mileage of the event. For the 7-, 10- and 12-mile ruck, the event was divided into four equal segments of lengths 1.75, 2.5 and 3 miles, respectively. For the 3- and 9-mile rucks, the event was divided into three equal segments of 1 and 3 miles, respectively. In the events with four segments, all EHS cases occurred during the third segment. Therefore, only the first three segments were used for the analysis, which was consistent with the other events with three segments.

[0038] For each of the six events where EHS occurred, three individuals who did not experience EHS while completing the same event, were assigned as control cases to the EHS subject in that same event. These individuals were matched to the EHS subject by age (year), height (m), weight (kg) and their 2-mile (Army) or 3-mile (Marines) run time (minutes), which can provide an indication of fitness level. These criteria are examples of how to determine whetherindividuals are similarly situated. A two-factor (segment number vs group) repeated measures ANOVA (Analysis of Variance) with post-hoc least significant differences (IBM SPSS Statistics v.28) was used to examine differences in HR, ECTemp, Tsk, and Tc-Tsk where Tc is core temperature. Any measurements with a p-value<0.05 were considered significant. Results are presented in the form of mean±standard deviation (SD).

[0039] Turning to the first factor and its basis on a differential response of heart rate and skin temperature of EHS individuals compared to control individuals during exercise. For this feature, the physiological responses of six EHS cases were examined and compared with three matched controls who participated in the same event but did not experience an EHS. The exercise was a loaded ruck march event for a Ranger 12 Miler, a Marine Infantry Officer Course10 Miler, or a Marine Crucible. A second group of seven individuals (all male) that suffered from EHS during one of the six observed training events: one during a 7-mile ruck, two during a 12- mile ruck, three during a 9-mile ruck and one during a 10-mile ruck were also used. Table 1 provides demographic characteristics for each group across all the observed training events. Table 2 provides EHS event and subject characteristics.N Age (yr) Height (m) Weight (kg)Heat Cases 7 22.1 ± 3.8 1.81 0.1 86.1 1 11.3Control Cases 21 21.3 12.6 1.810.1 85.41 10.8Table 1 : Group CharacteristicsEvent WBGT Relative Event Load Age Height Weight Trectal at PreDescription (°C) Humidity Start Carried (year) (m) (kg) Collapse event(%) Time (kg) (°C) Risk(hh:mm) Factors7-mile ruck 19.511.4 75.015.4 5:30 18.1 22 1.91 102.1 41.3 07 August 2018 12-mile 14.210.2 90.514.9 5:00 27.2 19 1.70 72.6 41.9 1 ruck 11 March 2019 12-mile 22.510.4 94.112.5 4:10 27.2 27 1.78 88.5 42.2 0 ruck 17 July 2019 3-mile 82.718.4 60.7115.9 19:30 21 1.73 78.9 42.1 1 movement 13 August 20219-mile ruck 80.413.9 69.6116.0 3:30 20.4 19 1.83 83.9 41.0 014 August 202110-mile 87.0±4.6 68.2+8.3 20:30 28 1.88 99.8 40.7 1 ruck 21 July 2022 9-mile ruck 85.2±4.2 76.7+11.8 3:30 20,4 19 1.75 77.1 41.8 030 July 2022Table 2: Event and EHS Subject Characteristics; individual reported recent immunization; ^individual reported recent immunization and felt ill after receiving immunization; findividual reported taking vitamins

[0040] For the events where EHS occurred, there was no significant difference in overall HR between the two groups (p=0.062), as well as within groups between each segment (p=0.29). However, there was a significant group by segment interaction (p=0.042), indicating that the pattern of change in HR was different. HR was higher in EHS individuals than in control individuals for segment two (160±22bpm vs 138±23bpm; p=0.037) as well as segment three (161 ±24 vs 139±21 bpm; p=0.029) as illustrated in FIG. 3.

[0041] There was no significant difference in step count both within (p=0.2) and between groups (p=0.73). As illustrated in FIG. 4, there was (1) no significant increase in movement speed or activity level for both groups during the event and (2) no difference in movement speed or activity level between the two groups.

[0042] As illustrated in FIG. 5, there was a significant change in Tect within each group over time. Tect in EHS individuals increased from segment one to segment two (38.1±0.3°C to 38.9±0.7°C; p<0.001), from segment two to segment three (38.9±0.7°C to 39.1±0.9°C; p=0.012) and from segment one to segment three (38.1±0.3°C to 39.1±0.9°C; p<0.001). Tect in control individuals increased from segment one to segment two (37.9±0.3°C to 38.4±0.6°C; p<0.001), from segment two to segment three (38.4±0.6°C to 38.5±0.7°C; p=0.016) and from segment one to segment three (37.9±0.3°C to 38.5±0.7°C; p<0.001 ).

[0043] Tsk was significantly higher overall in EHS individuals than in control individuals (37.8±0.4°C vs. 35.7±0.2°C respectively; p<0.001) as illustrated in FIG. 6. There was an opposite response in the overall pattern of Tsk between the two groups. Tsk in EHS individuals increased from segment one to segment two (37.1±1 ,5°C to 38.0±1.5°C; p=0.03), from segment two to segment three (38.0±1.5°C to 38.5±1.3°C; p=0.02), and from segment one to segment three (37.1±1.5°C to 38.5±1.3°C; p<0.001). Tsk in control individuals decreased significantly between segment two and segment three (35.9±1.1°C to 35.5±1.1 °C; p=0.007), between segment one and segment three (35.9±0.8°C to 35.5±1.1°C; p=0.023), but Tsk remained constant between segment one and segment two (35.9±0.8°C to 35.8±1.1 °C; p=0.756). There was a significant difference in Tsk between EHS and control individuals for segment one (37.1±1.5°C vs. 36.0±0.9°C; p<0.03), segment two (38.0±1.5°C vs. 35.9±1.1°C; p<0.001), as well as segment three (38.5±1.3°C vs. 35.5±1.1°C; p<0.001).

[0044] The core to skin thermal gradient (which is represented by the difference in °C between the two measurements; Tect-Tsk) was significantly different (p=0.012) both within groups over time and between groups as reflected in FIG. 7. Tect-Tsk in the control individuals increased significantly from segment one to segment two (1.9±0.8°C to 2.4±1.3°C; p=0.011), from segment two to segment three (2.4±1 ,3°C to 2.9±1 ,5°C; p<0.001), and from segment one to segment three (2.0±0.7°C to 2.9±1 ,5°C). There was no significant difference between any of the segments in EHS individuals. There was a significant difference in Tect-Tsk between EHS and control individuals for segment two (1 ,3±1 ,0°C vs. 2.4±1 ,3°C; p=0.021) and segment three (1.2±1.2°C vs. 2.9±1.5°C; p=0.01), but not for segment one (1.6±1.1°C vs. 1.9±0.8°C; p=0.45). Tsk remained close to Tect for the entire event for EHS individuals, but Tsk diverged from Tect for control individuals.

[0045] Turning to a subset of six individuals who experienced an EHS (age=22.7±3.6yr., ht=1.8±0.1 m., wt=87.6±10.6kg., Tr=41 ,5±0.5°C at collapse) during one of three loaded ruck march events and the first factor. Similar data were collected on three control individuals for each of the EHS individuals, with the controls being matched for age, height, weight, and two- mile run time (N=18, age=21.5±2.5yr., ht=1 ,8±0.06m., wt=86.9±10.7kg.). Each ruck was divided into 4 equidistant quarters. Mean HR, Tsk, and Tect were computed for each quarter. The ruck march physiology profiles were compared using a 2-way ANOVA.

[0046] The results had a variety of differences between the individuals who succumbed to EHS and those individuals who did not. Tect rose over time (p<0.001) with higher Tect in EHS individuals than the control individuals (38.1 ±0.4 to 39.2±0.9°C vs. 37.9±0.3 to 38.6±0.3°C; p<0.001). HR was higher for EHS individuals (p<0.001) and showed a different pattern overtime (p<0.03). The control individuals’ HR remained constant at ~142±22 bpm but increased progressively in EHS individuals (154±25 to 164±25 bpm). Tsk in EHS individuals was higher (p<0.001) and showed an inverse pattern overtime (p=0.13) compared to the control individuals. EHS individuals’ Tsk rose from 37.1±1.5 to 38.2±2.1°C, where for the control individuals’ Tsk decreased from 36.0±0.8 to 35.4±1 .1 °C; p < 0.05 for both.

[0047] FIG. 8 illustrates the significant differential response of both HR and Tsk for the EHS individuals (top two lines) versus control individuals (bottom two lines).

[0048] It can be difficult to distinguish EHS cases from control individuals with a higher HR and core temperature. The differential response of HR and Tsk over time provides an approach to identify individuals who are at risk for EHS as compared to other individuals.

[0049] Turning to the second factor and its basis on the conflicting response of the slowing of movement but an increase in heart rate minutes prior to collapse. After conducting further analysis on the six EHS cases, a second pattern emerges from the data. Just prior to collapse, HR appears to increase to near maximal levels while movement speed decreases (inferred from a decrease in minute step count). FIG. 9 illustrates this feature for one of the EHS individuals. At around minute 170 minutes (about 10 minutes prior to collapse), HR begins to increase whilestep count begins to decrease. Prior to this point in time, step count (which is proportional to speed of movement) is highly correlated with HR. This disassociation between speed of movement and HR appears to indicate a redirecting of cardiac output away from the metabolic demands of movement (fuel and oxygen to the muscles) to another area for the maintenance of critical functions (e.g., blood pressure). FIG. 9 is an example of an EHS individual from Iteration 31 , conducting a loaded 12-mile ruck march showing, HR, Tsk, ECTemp, and step count (steps / min). This pattern is also reflected for the EHS individual’s data shown in FIG. 10.

[0050] Turning to the third factor and its basis on the conflicting response of the slowing of movement but a decrease in skin temperature minutes prior to collapse when viewing the data for the six EHS individuals. Conversely, the relationship between speed of movement and Tsk appears to be negatively correlated demonstrated by lower skin temperatures at higher speeds of movement compared to higher Tsk with lower speeds of movement. Similar to the previous feature, minutes before collapse, skin temperature appears to respond differently to this correlation. FIGs. 9 and 10 illustrate this effect, where movement speed reduces along with Tsk just minutes prior to collapse. In FIG. 10 at minute 125, HR jumps to near maximal levels. Similarly, at this time, Tsk begins to decrease. Prior to this point in time, the step count (which is proportional to speed of movement) is negatively correlated with Tsk. This disassociation between speed of movement and Tsk appears to again indicate a redirecting of cardiac output away from the skin and thermoregulation due to the demands of other critical functions.

[0051] To develop an overall risk indicator of EHS risk, the three relationships discussed above were combined using a mixture of likelihood models (e.g., linear regression models) for each of the three features. The approach requires learning an individual’s response to the stressful event and determine their deviation from that response. The resulting model for each factor is discussed below.

[0052] Over a sufficient time window (e.g., 15 minutes), the first feature model computes a regression equation for the group that computes the most likely rate of change for both HR and Tsk. An example iswhere w0and Wi are linear regression weights for offset and a coefficient in change of heart rate.

[0053] These can be put into a likelihood model and compute the probability of observing any individuals rate of change in HR and Tsk in equations 1 and 2:where zt=yt= p A and B are regression weighting coefficients and F and A are covariance matrices. These equations are used as a prediction of the likelihood of seeing the rate of change in either HR or Tsk of an individual given the distribution of the rate of change of the group. In at least one embodiment, the data for the regression weighting coefficients and the covariance matrices are based on data for a group performing a given task or activity. The goal is to have the group be representative of the individual being monitored in terms of physiology characteristics. In a further embodiment, F and A represent a mean response distribution that may be based on the activity being performed, the clothing being worn, and the environment in which the activity is being performed. The environment may include environmental characteristics but also indoors vs. outdoors, light levels, etc.

[0054] In an alternative embodiment, the skin temperature model is replaced by a temperature gradient model for the difference in temperature between the skin and the estimated core temperature. In a further alternative embodiment, the first feature model using heart rate and skin temperature is replaced by the temperature gradient model. Although duplicative, the temperature gradient model may be added as a fourth feature model.

[0055] The second feature model uses a regression equation over a sufficient moving time window (e.g., 15 minutes) that learns the relationship of HR and step count. Since step count will lead HR slightly the regression equation will need to be offset slightly in time, but again can be represented by a Gaussian likelihood model in equation 3: p(mt|nt) = (mt|Cnt:t-15, E) (3) where mt=HR, nt=step count, C is a regression weighting coefficient, and E is a covariance matrix. This equation provides a prediction of the likelihood of observing HR at time t given a certain step count at time t. In at least one embodiment, the data for the regression weighting coefficient and the covariance matrix are based on data for a group performing a given task or activity. The goal is to have the group be representative of the individual being monitored in terms of physiology characteristics.

[0056] The third feature model uses a regression equation over a sufficient moving time window (e.g., 15 minutes) that learns the relationship of Tsk and step count. Since step count will lead Tsk slightly, the regression equation will need to be offset slightly in time, but again can be represented by Gaussian likelihood model in equation 4: p(ot|nt) = N(ot|Dnt;t-15,Z) (4) where ot=Tsk, nt=step count, D are the regression weighting coefficients, and Z is the covariance matrix. In at least one embodiment, the step count n is over a sub-window of a larger time window, for example sub-windows of 5 seconds, 30 seconds, or 1 minute in a larger window of 5 minutes, 10 minutes, or 15 minutes. This equation provides a prediction of the likelihood of observing Tsk at time t given a certain step count at time t. In at least one embodiment, the data for the regression weighting coefficient and the covariance matrix are based on data for a groupperforming a given task or activity. The goal is to have the group be representative of the individual being monitored in terms of physiology characteristics.

[0057] In all these cases / features, the model is looking for outliers or the opposite of what is occurring with the group. The model uses the outputs of the feature models to determine a probability of EHS. Thus, to get an overall likelihood or risk of EHS in at least one embodiment, the four equations are combined with appropriate weights (e.g., w w2, w3, and w4) in equation 5:In an alternative embodiment, the outputs of the feature models are summed or added together to provide the overall probability in the model where the weights would be one.

[0058] In at least one embodiment, the method includes the monitoring or receiving of the heart rate and the skin temperature for a monitored individual, for example from a heart rate sensor and a skin temperature sensor, which in at least one embodiment the sensors may be present on the same device with or without a processor configured to use the above-described models. In at least one embodiment, the heart rate and the skin temperature are measured, for example, at 30 second or 1 minute intervals over a 15 minute window. This data is used as inputs into three feature models, which in at least one embodiment may operate substantially at the same time or any order.

[0059] The first feature model compares monitored individuals data to a set of expected heart rates and skin temperatures which may be determined using (1) a similar activity completed by the individual prior; (2) from published models that predict responses to activity, environment, and clothing; or (3) across the group in which the monitored individual is a member to determine a probability that the rate of change for their heart rate and skin temperature falls outside the predetermined set or the group, for example using equations 1 and 2. In the group embodiment, the system would communicate with other systems worn by other individuals and / or a central system (e.g., central server) that receives the data from all of the individuals being monitored to provide the distribution of data for the changes in HR and Tsk, and in a further embodiment, this occurs for the other two models. In at least one embodiment, the central system may be cloud-based or server-based.

[0060] The second feature model uses the heart rate and the level of activity to determine whether they are divergence from each other. In at least one embodiment, the level of activity is represented by a step count over the sampling interval, which step count may be derived from accelerometer data or an activity monitor such as a smart watch, a phone, or a chest strap. In a step count embodiment, equation 3 above may be used to determine this probability or a similar equation adjusted for the type of activity being monitored.

[0061] The third feature model uses skin temperature and the level of activity as inputs to determine whether there is a divergence between the two. As with the second feature model,the activity level may be a step count. In the step count embodiment, equation 4 above may be used to determine this probability or a similar equation adjusted for the type of activity being monitored.

[0062] In at least one embodiment for the second and third models, a regression weighting coefficient and a covariance matrix is used to determine the likelihood that the observed vital sign should occur for that combination of time t and level of activity. In at least one embodiment, the activity level used by the second and third models are for different activities and / or selected based on the activity being performed by the monitored individual.

[0063] In at least one embodiment, the method then combines these probabilities together using predetermined weights, for example using equation 5. The system then provides the resulting probability as an output to allow for an intervention to occur when there is an increased probability of EHS that exceeds a predetermined risk threshold. In an alternative embodiment, the intervention may occur when the probability trend projects out to a higher risk of EHS and / or the rate of change exceeds a predetermined change threshold.

[0064] Based on this disclosure, it should be appreciated that the feature models are based on data, and during use may use historical data and / or real time group data that is gathered from multiple individuals during their current activity to facilitate identifying any outlier that may be about to suffer an EHS event. The data provides the signatures used in the models to compare an individual’s current physiological data to and facilitate the ability of the system to alert the individual or someone else if the risk of an EHS event is present.

[0065] The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, circuit, segment, or portion of code, which includes one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

[0066] The terminology used herein is for the purpose of describing embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the root terms “include” and / or “have”, when used in this specification, specify the presence of stated features, integers, steps, operations, elements,and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. As used, “A, B and / or C” or “at least one of A, B, and C” means just A; just B; just C; A and B; A and C; B and C; or A, B and C.

[0067] The corresponding structures, materials, acts, and equivalents of all means plus function elements in the claims below are intended to include any structure, or material, for performing the function in combination with other claimed elements as specifically claimed. The description of the present invention has been presented for purposes of illustration and description, but is not intended to be exhaustive or limited to the invention in the form disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the invention.

[0068] Although the present invention has been described in terms of example embodiments, it is not limited to those embodiments. The embodiments, examples, and modifications which would still be encompassed by the invention may be made by those skilled in the art, particularly in light of the foregoing teachings.

[0069] As used above “substantially,” “generally,” and other words of degree are relative modifiers intended to indicate permissible variation from the characteristic so modified. It is not intended to be limited to the absolute value or characteristic which it modifies but rather possessing more of the physical or functional characteristic than its opposite, and preferably, approaching or approximating such a physical or functional characteristic.

[0070] Those skilled in the art will appreciate that various adaptations and modifications of the embodiments described above can be configured without departing from the scope and spirit of the invention. Therefore, it is to be understood that, within the scope of the appended claims, the invention may be practiced other than as specifically described herein including dependent claims depending from more than 2 claims may depend from all other non-conflicting dependent claims depending from the same independent claim.VI. INDUSTRIAL APPLICABILITY

[0071] The disclosed embodiments provide an improved way to predict an EHS event based upon an individual’s heart rate, skin temperature, and activity level. In some embodiments, the systems and methods provide a mechanism to alert the individual of the potential for an EHS event with the prediction allowing for an intervention to occur for the monitored individual. The above-described models with use of physiological monitors may provide real-time early detection of impending EHS prior to collapse for individuals at risk, therefore reducing the time to treatment, decreasing injury severity, and improving prognosis for recovery.

Claims

Claims1 . A method for detecting exertional heat stroke (EHS) signatures for an individual, the method comprising: monitoring or receiving a heart rate for the individual from a heart rate sensor; optionally storing the heart rate over time in a memory; monitoring or receiving a skin temperature for the individual from a skin temperature sensor; optionally storing the skin temperature over time in the memory; comparing the heart rates and the skin temperatures to an expected set of heart rates and skin temperatures that optionally are based on a similar activity completed by the individual previously; a published model that predicts responses to activity, environment, and / or clothing; and / or other members of a group including the individual that is performing the activity to determine a probability that a rate of change for the heart rate for the individual and / or a rate of change for the skin temperature for the individual falls outside the expected set of heart rates and skin temperatures; monitoring or receiving an activity level for the individual from an activity monitor; optionally storing the activity level over time in the memory; comparing a trend of the heart rates to a trend of the activity levels to determine a probability that the two trends are divergent from each other; comparing a trend of the skin temperatures to the trend of the activity levels to determine a probability that the two trends are divergent from each other; combining the three probabilities together optionally using predetermined weights to obtain an overall probability for EHS for the individual; alerting the individual or another individual when the overall probability is above a risk threshold, trending towards a projected combined probability representative of EHS, or a rate of change for the combined probability exceeds a change threshold; and optionally performing an intervention for the individual in response to the alert.

2. The method according to claim 1 , wherein comparing the heart rates and the skin temperature uses a probability model according to p(zt) = N(zt|Azt;t-15,r) p(yt) = / V(yt|Byt;t_i5,A) where zt=A and B are the regression weighting coefficients and F and A are the covariance matrices, HR is heart rate, and Tsk is the skin temperature.

3. The method according to claim 2, whereinCli ck= W1dTsk+ at where w0and Wi are linear regression weights for offset and a coefficient in change of heart rate.

4. The method according to claim 2 or 3, wherein data for the regression weighting coefficients and the covariance matrices are based on data for a group performing a given task or activity.

5. The method according to claim 4, wherein the group is representative of the individual being monitored in terms of physiology characteristics.

6. The method according to claim 2, wherein F and A represent a mean response distribution that may be based on an activity being performed, a type of clothing being worn, and an environment in which the activity is being performed.

7. The method according to claim 1 , wherein comparing the trend of the heart rates to the trend of the activity levels includes using a Gaussian likelihood model which optionally may have an offset over time between the activity level and the associated heart rate for a particular time t.

8. The method according to claim 7, wherein the Gaussian model is p(mt|nt) = JV(mt|Cnt;t-15, E) where mt is heart rate, n is a step count, C is a regression weighting coefficient, and E is a covariance matrix.

9. The method according to claim 7 or 8, wherein the Gaussian model is a prediction of the likelihood of observing a particular heart rate at time t given a certain step count at time t.

10. The method according to claim 8, wherein the regression weighting coefficient and the covariance matrix are based on data for a group performing a given task or activity being performed by the individual.1 1. The method according to claim 10, wherein the group is representative of the individual being monitored in terms of physiology characteristics.

12. The method according to claim 1 , wherein comparing the trend of the skin temperatures to the trend of the activity levels includes using a Gaussian likelihood model which optionally may have an offset over time between the activity level and the associated skin temperature for a particular time t.

13. The method according to claim 12, wherein the Gaussian model isP(°tlnt) = JV(ot|Dnt.t-15,Z) where otis skin temperature, n is step count, D are the regression weighting coefficients, and Z is the covariance matrix.

14. The method according to claim 12, wherein the step count n is over a sub-window of a larger time window, optionally sub-windows of 5 seconds, 30 seconds, or 1 minute in a larger window of 5 minutes, 10 minutes, or 15 minutes, respectively.

15. The method according to any one of claims 12-14, wherein the Gaussian likelihood model provides a prediction of the likelihood of observing a particular skin temperature at time t given a certain step count at time t.

16. The method according to 13 or 14, wherein the regression weighting coefficient and the covariance matrix are based on data for a group performing the given task or activity being performed by the individual.

17. The method according to claim 16, wherein the group is representative of the individual being monitored in terms of physiology characteristics.

18. The method according to claim 1 , wherein the overall probability is determined using the following modelwhere wi, w2, w3, and w4are weights, p(zt) and p(yt) are models for heart rates and skin temperatures overtime, p(mt|nt) is a model for heart rate-activity trends, and p(ot| nt) is a model for skin temperature-activity trends.

19. The method according to claim 18, wherein the weights are equal weights for each probability, one or more weights are heavier than one or more other weights, or the weights that vary over time to emphasize some probabilities over other probabilities depending on the length of the activity being performed by the individual.

20. The method according to claim 18, wherein the weights are variable such that during an earlier part of the activity, the weights for the first two probabilities are weighted more heavily to reflect that the factors of these probabilities being more likely to provide earlier warnings than the other probabilities; and as the activity progresses, the weights are adjusted to increase the weight of the last two probabilities more heavily as these probabilities are better indicators of EHS closer to when EHS may occur.21 . The method according to claim 1 , 2, or 18-20, wherein the intervention is selected from a group consisting of reducing the pace of activity or stopping the activity, adjusting the clothing being worn by the monitored individual, providing fluids to and / or pouring fluids over the monitored individual, providing ice or other cooling material or equipment to the monitored individual, placing the monitored individual in an ice bath or cooler temperature water and / or environment, and seeking medical care.

22. A method for detecting signatures for an individual based on physiological data for comparison to exertional heat stroke (EHS) signatures, the method comprising: receiving a plurality of hearts rates, skin temperatures, and activity levels for the individual;inputting the heart rates, skin temperatures, and activity levels for a monitoring window into a model having sub-models for determining a signature of the individual based on an existence of a differential response of 1) the heart rate and the skin temperature, 2) the heart rate and a temperature gradient between skin temperature and estimated core temperature, or 3) temperature gradient between skin temperature and estimated core temperature of EHS individuals compared to controls during exercise or activity, a conflicting response of the slowing of movement but an increase in the heart rate minutes prior to collapse, and a conflicting response of the slowing of movement but a decrease in the skin temperature minutes prior to collapse; combining the outputs of the sub-models to determine an overall probability that the individual will have EHS based on a comparison between the signatures and the EHS signatures; and optionally performing an intervention for the individual in response to the alert.

23. A system comprising: a heart rate monitor or sensor; a skin temperature sensor; a movement sensor; a housing with said heart rate monitor or sensor, said skin temperature sensor, and said movement sensor; an optional communications component; and a processor configured to perform any of the methods according to any one of claims 1 , 2, 18-20, or 22 other than the optional intervention.

24. The system according to claim 23, further comprising a chest strap attached to said housing.

25. The system according to claim 23, wherein said housing includes said communications component.

26. The system according to claim 23, wherein the movement sensor is a tri-axial accelerometer.

27. The system according to claim 23, further comprising a central server in communication with a plurality of sensors on different individuals and configured to receive sensor data for aggregation into signatures for determining the risk of exertional heat stroke (EHS).

28. A method for alerting an individual of a risk of an exertional heat stroke (EHS), the method comprising: receiving heart rate data, skin temperature data, and activity data for the individual; creating a signature for the individual based on the received data;providing the signature to a model configured to determine the risk of an exertional heat illness for the individual, wherein the model includes up to four sub-models each having a signature based on previously received heart rate data, skin temperature data, and activity data for the individual or similarly situated individuals heart rate data, skin temperature data, and activity data representative of individuals that have suffered EHS or are a control group; determining the risk of the EHS based on a comparison of signatures; outputting the risk and recommending an intervention when the risk is above a threshold, and wherein similarly situated individuals have a similar physiological and / or fitness makeup to the individual being monitored.

29. The method according to claim 28, wherein one sub-model compares the heart rate data and the skin temperature data in a probability model according to p(zt) = / V(zt|Azt;t-15, r)P(yt) = / V(yt|Byt:t-15,A) where zt=A and B are the regression weighting coefficients and F and A are the covariance matrices, HR is heart rate, and Tsk is the skin temperature.

30. The method according to claim 29, whereinwhere w0and Wi are linear regression weights for offset and a coefficient in change of heart rate.31 . The method according to claim 28, wherein one sub-model compares the trend of the heart data to the trend of the activity data using a Gaussian likelihood model which optionally may have an offset over time between the activity level and the associated heart rate for a particular time t.

32. The method according to claim 31 , wherein the Gaussian model is p(mt|nt) = JV(mt|Cnt:t-15, E) where mt is heart rate, n is a step count, C is a regression weighting coefficient, and E is a covariance matrix.

33. The method according to claim 31 or 32, wherein the Gaussian likelihood model is a prediction of the likelihood of observing a particular heart rate at time t given a certain step count at time t.

34. The method according to claim 28, wherein one sub-model compares a trend of the skin temperature data to a trend of the activity data using a Gaussian likelihood model which optionally may have an offset over time between the activity level and the associated skin temperature for a particular time t.

35. The method according to claim 34, wherein the Gaussian model is p(ot|nt) = JV(ot|Dnt;t-15,Z) where ot is skin temperature, n is step count, D are the regression weighting coefficients, and Z is the covariance matrix.

36. The method according to claim 35, wherein the step count n is over a sub-window of a larger time window, optionally sub-windows of 5 seconds, 30 seconds, or 1 minute in a larger window of 5 minutes, 10 minutes, or 15 minutes.

37. The method according to any one of claims 34-36, wherein the Gaussian likelihood model provides a prediction of the likelihood of observing a particular skin temperature at time t given a certain step count at time t.

Images