`
`EXHIBIT 2001
`
`
`
`Earbud-Based Sensor for the Assessment
`of Energy Expenditure, HR, and VO2max
`
`STEVEN FRANCIS LEBOEUF', MICHAEL E. AUMER', WILLIAM E. KRAUS2, JOHANNA L. JOHNSONZ,
`and BRIAN DUSCHA2
`
`’ Valencell, Inc., Raleigh, NC; and 2Division ofCardiology, Department ofMedicine, Duke University
`School ofMedicine, Durham, NC
`
`ABSTRACT
`
`LEBOEUF, S. F., M. E. AUMER, W. E. KRAUS, J. L. JOHNSON, and B. DUSCHA. Farbud-Based Sensor for the Amessment of
`Energy Expenditure, HR, and V02,,.,,. Med. Sci. Sports Exen-., Vol. 46, No. 5, pp. 1046 1052, 2014. lntroductionllhrrposc: The goal
`of this program was to determine the fmsibility ofa novel noninvasive, highly miniaturized optornechanical earbud sensor for accurately
`estimating total energy expenditure (TEE) and maximum oxygen consurnptim (V()2,,_,,). The optomechanical sensor module, small
`enough to tit inside commercial audio earbuds, was previously developed to provide a seamless way to measure blood flow informa-
`tion durirrg daily life activities. The sensor module was configured to continuously measure physiological information via photo-
`plethysmography and physical activity infomratim via accelerometry. 'Ihis information was digitized and sent to a microprocessor
`where digital signal—processing algorithms extract physiological metrics in real time. These metrics were streamed wirelessly from
`the earbud to a computer. Methods: In this study, 23 subjects of multiple physical habitus were divided into a training group of 14
`subjects and a validaion group of 9 subjects. Each subject underwent the same exercise measurement protocol consisting of treadmill-
`based cardiopulmonary exercise testing to reach \70;»_....,. Benchmark sensors included a 12-lead IKZG sensor for measuring HR, a
`calibrated treadmill for measuring distance and speed, and a gas-exchange analysis instrument for measuring TEE and \-102...... 'Ihe
`earbud sensor was the device under test Benchmark and device under test data collected from the 14-person training data set study
`were integrated into a preconceived statistical model for correlating benchmark data with earbud sensor data. Coefficients were op-
`timized, and the optimized model was validated in the 9-person validation data set. Results: It was observed that the earbud sensor
`estimated TEE and V0-,._,,_, with mean 1 SD percent estimation errors of 0.7 1 7.4% and
`3.2 L 7.3%, respectively. Conclusion: The
`earbud semorcan accurately estimae TEE and V()2,,_,, during cardiopulmomry exercise testing. Key Words: FAR, ACCH..EROMEI'ER,
`PHOTOPLETHYSMOGRAPHY, PULSE
`
`odifiable health risk fitctors, such as high stress,
`poor diet, and sedentary lifestyle, account for 25%
`of all medical expenses and millions of deaths
`per year worldwide (2). The U.S. population is becoming
`increasingly overweight and unhealthy, with an estimated
`66% of adults categorized as obese or overweight by the
`CDC (26). Nonetheless, more than half of American adults
`exercise on a regular basis (1 1), spending more than $55
`billion in weight loss programs and more than $17 billion
`on fitness products (31). The disconnect between dollars
`spent on weight loss and obesity levels may be explained by
`recent findings that traditional diets do not work (24) alone
`
`Adttas fa‘ carespondence: Steven Francis Lfioeut; Ph.D., Valencell, lnc.,
`28(1)-154 Sumner Bhrd., Raldgh, NC 27616; E-mail: LeBoeu@vak:rrceI.corrL
`Submitted for publication March 2013.
`Accepted for publication October 2013.
`0195-9131/I4/4605-1046/0
`MEDICINE & SCIENCE IN SPORTS & EXERCISE,
`Copyright© 2014 by the American College of Sports Medicine
`D01: 10.1249/MSS.(l)00000000000l83
`
`to prevent weight gain and to promote weight loss. Dietary
`measure must be combined with energy expenditure to ac-
`complish long-terrn weight loss and maintenance.
`Weight loss programs aimed at promoting fitness through
`direct measurement of physical activity (PA) via pedometer
`feedback have shown promise. In particular, incorporating
`a pedometer in daily life activities has been shown to result
`in a significant reduction in body mass index (BMI) and
`blood pressure (7). Furthermore, combining engaging feed-
`back with an online user experience correlates with im-
`proved maintenance of weight loss in long-term diet/weight
`management studies (10). These observations indicate that
`even better outcomes may result from a more direct feed-
`back about energy expenditure and aerobic fitness level,
`such as V02, calories burned, and \'/02.“,
`Indeed, there is a clear opportunity to encourage a broader
`population to embrace active lifestyles by integrating mobile
`fitness monitoring devices with compelling user experiences.
`However, compelling user experiences must be meaningfirl,
`and to be meaningful, the fitness monitoring gadgets must
`provide information that is sufliciently accurate to be action-
`able. This goal is challenged by the fact that commercial pe-
`dorneters are inaccurate by greater than i 20% in reporting
`calories burned (8,29).
`
`1046
`
`Copyright 9 2014 by the American Colege of Sports Medicine. Unauthorized reproduction of this article is prohibited.
`
`VALENCELL_0001
`
`<
`
`Ua
`
`n DL
`
`L!
`:1
`CL
`CL
`
`an
`LL!
`
`UZ L
`
`L’
`
`
`
`Wmm
`
`Cm Um Qm Z(
`
`)>
`‘U
`‘U
`
`
`
`FIGURE I The components and size of the device under test (DUT).
`Shown are the earbud and the medallion containing the majority of
`the computational components. Shown for scale is a U.S. quarter. Note
`the pmition of the sensor module at the bottom of the antitragus. The
`sensor module is configured to lit between the concha and the antitragus
`of the ear.
`
`Recent energy expenditure studies, using a wearable Acti-
`Health chest strap monitor for measuring both PA and HR,
`have demonstrated greater accuracy (5,6). These researchers
`achieved such predictive accuracy through branched equation
`modeling, using HR information and accelerornetry infor-
`mation as independent variables. Although these findings are
`quite encouraging, researchers using the ActiHealth monitor
`point out several shortcomings. First, despite the relatively
`high precision achievable through branched equation model-
`ing, poorer accuracy is observed if individual calibrations are
`not used (5,6). This means that the wearable monitors must
`be calibrated for each user, in a process that is both time con-
`suming and burdensome. Furthermore, as audio earbuds are
`packaged with srnartphones and digital media players that
`are sold in volumes of hundreds of millions of units a year
`(16), the audio earbud form factor provides the opportunity
`to reach a larger consumer audience than that of an HR
`chest strap, which is sold in volumes of less than 10 million
`pa year.
`The goal of this study was to determine the feasibility of
`a highly miniaturized, noninvasive optomechanical earbud
`sensor technology for accurately monitoring physiological
`metrics such as HR, total energy expenditure (TEE), and
`
`maximum oxygen consumption 02",“), and this study is
`reported herein.
`
`METHODS
`
`To overcome these reported limitations, an earbud sensor
`module—as opposed to an ActiGraph wrist-, arm-, or leg-
`wom sensor—was selected in this study (Fig. 1). Details
`of the mechanism of operation are described elsewhere
`(17-21), but in summary, the earbud comprised a highly
`integrated sensor module capable of measuring subtle blood
`flow changes via reflective photoplethysrnography (PPG)
`and changes in body motion through a three-axis acceler-
`ometer. This sensor module was designed 1) to capture and
`digitize the optical PPG signal and 2) to send the digitized
`infomration to a digital signal processor (DSP) for remov-
`ing motion artifacts and environmental noise fi'om the PPG
`signal and to continuously generate estimates of HR and
`V02 metrics in real time based on a statistical model com-
`prising PPG and accelerometry information. The DSP was
`in electrical communication with a Bluetooth chipset so that
`the real—tirne metrics could be called upon by a client de-
`vice (such as a laptop or smartphone). A preliminary feasi-
`bility study of this PerformTek® earbud sensor module had
`previously demonstrated accurate HR measurements during
`exercise, thus potentially eliminating the need for an elec-
`tmcardiographic chest strap in many use cases. This was a
`critical finding for the issue of user compliance, as 58% of
`U.S. headphone owners listen to headphones while exercis-
`ing and 34% wear headphones during evayday life activities
`(such as doing work around the house) (13), 10 times greata
`than the number ofAmericans who exercise with chest straps.
`Subjects. In this study, 23 subjects of good physical
`health were divided into a training group of 14 subjects and
`a validation group of 9 subjects. This sample size is justified
`by the high “effect size” observed for calibrated comela-
`tions of V702 and HR (22) and is further supported by the
`va-y high R2 coefficient observed (23) when comparing
`the earbud-deterrnined HR to 12-lead ECG-measured HR
`
`during exercise. The training group (Table 1a) comprised
`12 men and 2 women: age = 39 i 11.8 yr, weight = 73.5 i
`12.2 kg, height = 69 i 2.9 cm, BMI = 23.6 : 2.1 kg-m_2.
`The validation group (Table 1b) comprised five men and
`four women: age = 36 i 6.9 yr, weight = 67.6 i 15.7 kg,
`height = 173 i 7.4 cm, BMI = 22.3 i 4.0 kg-m'2. Each
`
`TABLE 1. Descriptive characteristics (mean + SD) of (a) trairing group and (1)) Validation group.
`(a) Training Group
`
`(Ir) Validation Grnup
`
`Paramehr
`Sex
`Age
`weight
`Height
`Distance
`Energy expendiirre
`Ilaximum V02
`IMI
`
`Value (Mean + SD)
`2 females, 12 males
`30 + 11.8 yr
`735 + 12.2 kg
`175 + 7.4 cm
`2.95 + 0.5 km
`213 + 47.8 km!
`55.9 r 6.5 mLAkg ‘-min ‘
`23.6 + 2.1 kg»m 2
`
`Pararnohr
`Sex
`Age
`Weight
`Height
`Distance
`Energy expenditure
`Maximum V02
`BMI
`
`Value (Mean + SD)
`4 females, 5 males
`36 r ('19 yr
`67.6 + 15.7 kg
`173 + 7.4 cm
`230 + 0.3 km
`178 + 51.5 lrcal
`55.1 + 5.5 nl.-kg ‘min ‘
`223 + 4.0 kgm 2
`
`EARBUD-BASED SEHSOR FOR PHYSICAL ASSESSMENTS
`
`Mediche & Science in Sports & Exercise.-,
`
`1047
`
`Copyright 9 2014 by the American Colege of Sports Medicine. Unauthorized reproduction of this article is prohibited.
`
`VALENCELL_0002
`
`
`
`subject underwent the same exercise measurement proto-
`col, including a treadrnill-based cardiopulmonary exercise
`(CPX) test, at 0° incline, to reach \'/02...“. The achievernernt
`of VO2,,,,,, was determined by reaching at least two of the
`three following criteria: plateau in V02 over the last nninute
`of exercise, achievement of at least 1.10 RER, arnd achieve-
`mernt ofat least 17 in perceived exertion on the Borg scale. The
`mean i SD \'/02.,“ values of the training group and the vali-
`dation group wee 55.9 : 6.5 and 55.1 i 5.5 mL-kg”-min‘ ',
`respectively. Benchmark sernsors included a 12-lead ECG for
`measuring HR, a calibrated treadmill for measuring distance
`traveled, arnd a gas-exchange analysis instrument for mea-
`suring TEE and VO;h,,,,,. The earbud sensor served as the
`device under test. All subjects provided infonned consent
`as approved by the investigational review board of the Duke
`University School of Medicine.
`CPX testing. Subjects began the study by first being
`prepped for wearing the benchmark sensors. A Quintonl2-
`lead ECG system was used as a bernchmark for HR, and a
`TnreMax 2400 Pan/oMedies (ParvoMedics, Sandy, UT) gas-
`exchange analysis mouthpiece was used as a benchmark for
`ernergy expenditure and continuous measures of V02. The
`banclnmark sensors were calibrated according to the stan-
`dard rnainternance guidelines of the manufacturers. The sub-
`jects were then fitted with an earbud sansor (Fig. 1) powered
`by the aforementioned PerforrnTek physiological monitor-
`ing technology. Participants were then asked to sit at rest in
`a supine position in a reclining chair for a few minutes while
`wearing the benchmark equipment and earbud sansor. Afia
`the resting period, subjects were instructed to move fiorn the
`chair to the calibrated treadmill and execute the CPX testing
`with graded irnternsity ranging from 0 to 9.1 mph speeds. The
`protocol used cornsisted of 2-min stages, increasing the work-
`load by approximately one metabolic equivalent per stage.
`Measurements fiom the bernchmark sensors and earbud sen-
`
`sor were collected continuously throughout the treadmill run.
`Participants were asked to continue running during each in-
`creasing speed until they were completely exhausted. The
`last 40 s of bernchnnark gas-exchange analysis data were av-
`eraged to determine measured peak V02.
`Earbud sensor. The novel noninvasive earbud sen-
`
`sor (Fig. 1) used in this study was designed by Valencell,
`lnc. (Raleigh, NC). The earbud sensor comprised a sensor
`module, a microprocessor, and a wireless Bluetooth® chipset.
`The optomechanical sensor module, comprising the sensor
`elements, was embedded within the right audio earbud, as
`shown in Figure 1, such that the sensor module would rest
`between the concha and the arntitragus of each subject upon
`earbud placement. The right arnd the left earbuds were de-
`signed to be pluggable to a wireless Bluetooth “medallion”
`via a detachable connector (as shown in Fig. 1). The medal-
`lion housed the nnicroprocessor and the Bluetooth clnipset.
`At the heart of this noninvasive earbud sensor is a highly
`miniaturized optomechanical module (l7—21,23) that mea-
`sures optical and mechanical inforrrnation from the area of a
`user’s ear between the antitragus and the concha. This novel
`
`sensor module comprises an infi'ared light-ernitting diode,
`a photodetector element, a tluee-axis accelerometer, and an
`optomecharnical housing. Designed to fit flush with the body
`of a standard audio earbud, the earbud essentially maintains
`the form factor of a typical audio earbud and does not re-
`quire an ear clip or an in-ear-canal sensor to firnction.
`The optical and mechanical infonnation collected fiom
`the car are sampled via methods akin to reflective PPG and
`three-axis accelerometry, and this sampled infomnation is
`then processed by novel algorithms (l7,l8) coded on finn-
`ware within the microprocessor for extracting weak blood
`flow signals from excessive motion noise. It is well known
`that motion artifacts are the greatest limiting factor to accu-
`rate vital sigrns monitoring via PPG (3,l4,27). However,
`Valencell’s PerformTek biometric algorithms actively pro-
`cess noisy body sigrnals and extract accurate biometrics even
`during aggressive nnnning and PA (23). These biometric sig-
`nals are then combined with contextual accelerometry infor-
`mation within a statistical model to gernerate assessments of
`HR zorne, calories burned, aerobic capacity (VO2,.,,,,,), and
`otha parameters (17-21). A phorne, computer, or otlner mo-
`bile device can communicate directly with the nnicroproces-
`sor via a Bluetooth link. In this particular study, the earbud
`sensor data wae streanned direcfly to a laptop via Bluetooth.
`Statistical methodology. A multiple linear regression
`model had been developed previously by Valencell to pro-
`vide a linear relation between estimated TEE, as predicted
`by the earbud sensor measurements, and the measured TEE,
`as recorded by the benchmark gas-exchange analysis de-
`vice. This linear model comprised fixed and tinne-varying
`terms. The fixed terms included weight (W), age (A), and
`sex (G) having a binary value of 0/1 for womenlmen,
`respectively. The tirne-varying terms included the earbud-
`estimated TEE GEB TEE) and the linear operations of real-
`time PPG and accelerometry (ACC). Although the details
`of the linear model are outside the scope of this article,
`the formalism of the resulting linear equation may be de-
`scribed by EB TEE = f(g(PPG), h(ACC), W, A, G), where
`g and h are functions of PPG and ACC, respectively. It is
`important to note that this lirnear model was directed to-
`ward estimating TEE, arnd not the individual elements of
`resting energy expenditure (REE) or activity-related ernergy
`expenditure (AEE), as TEE is what is measured by the gas-
`exchange arnalysis.
`A separate model had beern previously developed by
`Valencell
`to estimate V‘/O2,,,,,,, based on the HR and
`accelerometry data collected during several prior rounds of
`CPX testing. The methodology behind this \'/02...,“ esti-
`mation is described elsewhere (18), and the equation
`follows the formalism of EB \'/02.,” = j(Max HR,
`Min HR, k(ACC)), where EB \'/02,.” is the earbud-
`derived VO2.,,,,,, Max HR is the maximum reliable HR
`measured by the earbud sensor, Min HR is the minimum
`reliable HR measured by the earbud sensor, and k is a
`fimction of the accelerometer readings measured through-
`out the CPX testing.
`
`1048 Otficl Journal of the American College of Sports Medicine
`
`http://www.acsm-msse.org
`
`Copyright 9 2014 by the American Colege of Sports Medicine. Unauthorized reproduction of this article is prohibited.
`
`VALENCELL_0003
`
`<12
`
`n QL
`
`L!
`:1
`L
`
`QC
`
`an
`LL!
`
`UZ L
`
`L’
`
`Ua
`
`
`
`--ECGHR
`--Earbud (DUI) HR
`—-Activitv
`
`200
`
`180
`
`150
`140
`
`120
`
`3‘
`a
`53
`0»
`5 100
`a:
`ad
`'0
`
`80
`
`30:»
`
`so
`40
`20
`
`2
`
`L8 >
`0
`1-5 E
`1.4
`‘-'1
`3
`1.2 m
`'5.’
`1 <
`0
`0.8 8
`3
`
`0.6 3
`0.4
`9’
`F
`
`0.2 "' O
`
`0
`
`200
`
`400
`
`600
`
`800
`
`1000
`
`1200
`
`1400
`
`Time (5)
`
`FIGURE 2 CPX testing output from a characteristic test. In this
`characteristic CPX test. the time of the progressive exercise test con-
`ducted on a standard treadmill is shown on the abscissa; the PA in-
`tensity using accelerometer counts in arbitrary units (a.u.; green line) is
`shown on the rightward ordinate; the HR in beats per minute from
`either the l’erformTek earbud device (red) or from the simultaneously
`measured ECG (blue) is shown on the leftward ordinate. The mrbud-
`determined HR and the ECC-measured HR show complete alignment
`in this exemplary characteristic test.
`
`Alter the 14-person training data study, the best-fitting
`coefficients for the TEE and V02“, models were deter-
`mined, and the resulting optimized equations were used in
`the nine-person validation data study to estimate TEE and
`V02“, in real time. The resulting earbud-estimated values
`GEB TEE and EB V03“) were then compared with
`barchmark-measured values in accordance with the Bland-
`
`Altrnan plot (1,4).
`
`RESULTS
`
`HR. As previously described, the earbud measurements
`of HR and PA are part of the foundational formulas for
`EB TEE and EB v02...,,.. Therefore, it is important that
`these measurements are accurate. An exemplary character-
`istic plot of real-tirne ECG, PerformTek HR, and h(ACC) for
`a subject undergoing a CPX test is shown in Figure 2. Note
`
`
`
`
`
`200
`
`R-squared = 0.98
`
`SD = 4.4%
`
`50
`
`.
`100
`
`.
`150
`
`0
`
`200
`
`E g
`
`§ 150
`
`EQ
`
`100
`3 so
`-E(0
`Lu
`
`0
`
`n
`
`‘W
`
`mL
`
`U Cm Um Dm Z(
`
`>1
`
`::
`
`that for this test, the benchmark ECG and the earbud HR
`are nearly identical throughout the urn, such that they com-
`pletely overlap each other. Although complete overlap was
`not always observed, complete overlap was typically ob-
`served. Only rarely did the earbud or the ECG diverge to a
`substantial degree, as exemplified by the fight conelation
`shown in Figure 3. Also, on the rare occasions when diver-
`gence was observed, it was ofien attributable to either the
`earbud moving out of the ear or the ECG leads decoupl-
`ing from the subject’s skin. For the sake of objectivity in
`this study, all HR data points measured by the earbud and
`ECG sensors are shown in Figure 3, even for the case
`where earbud or ECG failures are subjectively believed to
`have occurred.
`
`A Bland—Altman plot comparing earbud—estirnated HR
`(EB HR) versus the benchmark 12-lead ECG measured from
`the l4—pason training group is presented in Figure 3. This
`figure illustrates the excellent agreement between EB HR
`and ECG throughout a full range of activity fiom rest to
`>200 bpm; the mean difference (bias) was -0.2%, the SD
`was i4.4%, and the coefficient of determination (R2) was
`0.98. In contrast with other reported optical HR measure-
`ment devices reported in literature (3,l4,27), the EB HR
`measurement is quite robust throughout a firll range of ac-
`tivity because the PerforrnTek biometric signal extraction
`algoritluns are capable of charactaizing motion noise during
`numerous activities and attenuating motion artifacts fiom the
`optical signal in real time.
`In contrast with hip and pocket-wom pedometer-based
`approaches for calculating distance (8,l5,29), the PA level
`measured by the earbud prototype provides a good reference
`for body displacement during walking, jogging, and nrnning
`without requiring knowledge of the user's sex, height, age,
`weight, or fitness. Furthermore, neither a calibration re-
`gimen nor a GPS is required to tune parameters to the
`wearers’ gait. The earbud prototype distance measurement
`was highly accurate, with a bias of 0.3%, an SD of 4.2%,
`and an R2 of 0.93. This distance mezsurement was ob-
`
`tained through a novel transformation of three-axis acceler-
`ometer data, and its accuracy is aided by the sensor location
`at the ear.
`
`2°
`
`.
`
`3
`2: ‘°.
`I
`C
`8
`0 ‘Q,
`5
`°
`E’ 10
`'
`>6 -
`
`'
`o.‘
`J,
`
`
`HR
`—mean
`do
`.
`V
`V —+/-1.96'sD
`53' , .9
`‘ ‘
`X‘
`..
`.
`..
`,
`'
`.3
`’. "" ".“r°:
`C
`=
`.3.’..‘4
`‘O
`o
`. .'
`,
`e
`0‘ “ZN;
`
`_20
`
`50
`
`.
`'
`100
`
`2
`
`O
`
`‘ .
`150
`
`v
`
`.
`
`200
`
`ECG heart rate (bpm)
`
`average HR (bpm)
`
`FIGURE 3 HR using the earbud (device under test |DUT|) and the simultaneously mmsured ECG benchmark. A. Regression relation comparing
`estimated (earbud) versus measured (ECG) HR for all data points collected for each participant. B. The Bland Altman plot of same. All the data
`points were taken from the training data collected during the Duke CPX test. The mean dilference (bias) was 0.2% and the SD was 4.4%. The mean
`is shown by the green line, and the 1.96 SD (95% limits of agreement) boundaries are shown by the red lines.
`
`EARBUD—BASED SBISOR FOR PHYSICAL ASSESSMENTS
`
`Mediche & Science in Sports & Exaciseg
`
`1049
`
`Copyright 9 2014 by the American Colege of Sports Medicine. Unauthorized reproduction of this article is prohibited.
`
`VALENCELL_0004
`
`
`
`able to provide understandable, actionable, and motivational
`feedback to the user, and 4) autonomous and user-fiiendly.
`Today, a variety of commercially available products offer
`step counting and estimated calorie monitoring. Many of
`these solutions have provided value to researchers study-
`ing energy balance and to fitness professionals and clini-
`cians monitoring the progress of exercise and diet plans.
`However, none of these products satisfy all the previously
`mentioned criteria, limiting the effectiveness and scope of
`l) long-term clinical research on energy balance research
`and 2) health and fitness solutions for end users. In con-
`trast, newly developed earbud sensor technology offers the
`promise of meeting these needs, enabling a truly seamless
`energy balance—monitoring platform for use in clinical re-
`search, consumer fitness, clinical assessment of energy bal-
`ance, and mobile health management.
`The feasibility has been established for the highly minia-
`turized, noninvasive optical earbud sensor technology for
`accurately monitoring physiological metrics such as HR,
`TEE, and maximum oxygen consumption (\'/Ohm) through
`the ear. The earbud sensor accurately predicted HR through-
`out all activity levels investigated, fiom rest to peak perfor-
`mance, with a mean difference (bias) of -0.2% and an SD
`of i4.4% when compared with an ECG benchrmrk device.
`Furthermore, real-time algorithms within the earbud sensor
`accurately predicted (a) TEE with a bias of -0.7% and an
`SD of 417.4% and (b) V02“, with a bias of -3.2% and an
`SD of 417.3%. This particular evaluation did not address
`user comfort or battery life but a commercially available
`Bluetooth audio headset, the iriverONT'“, incorporating the
`evaluated Pa'formTek® biometric sensor teclmology adver-
`tises several hours of measurement time while also support-
`ing music.
`The excellent performance of the earbud sensor for ac-
`curately measuring HR throughout extreme PA is especially
`noteworthy. Motion artifacts have been the greatest limita-
`tion to the continuous monitoring of vital signs during ac-
`tivity (3,l4,27), and the ability to accurately monitor vital
`signs with a consumer-priced audio headset is particularly
`irnpactful to public health.
`
`00O
`
`E 20
`3
`
`+ ‘calories
`mean
`—+/-1.96 ~ so
`
`g 10
`
`0
`
`;:_> -10
`‘E
`
`-20
`100
`
`.
`.
`.
`.
`300
`250
`200
`150
`average calories (kcal)
`
`i
`350
`
`FIGURE 4 Bland Altman plots simultaneously comparing the refer-
`ence CPX data and the earbud (device under test |DUT|) benchmark
`data collective. The Bland Altman plot of the difference between the
`energy expenditure (I-IE; kcal) using the reference CPX test and the
`DUT benchmark EE measurements taken from the 14-person CPX
`training group. The mean is shown by the green line. and the L96 SD
`(95% limits of agreement) boundaries are shown by the red lines.
`
`TEE. The EB TEE closely estimated the benchmark TEE
`for the training group data set, with a bias of -0.7% and
`an SD of i7.4% (Fig. 4). The correlation between the
`EB TEE and the benchmark TEE for the validation group
`data set was identical with that of the training data set, with
`a bias of -0.7%, an SD of i7.4%, and an R2 coefficient of
`0.86 (Fig. 4).
`\'Io,,..,,,. The EB vo,...,, closely estimated the bench-
`mark V02,m, for the training group data set, with a bias
`of -0.1%, an so of :8.7%, and an R2 coefficient of 0.36
`(Fig. 5). The correlation between the EB V02,” and the
`barchmark TEE for the validation group data set was simi-
`lar to that of the training data set, with a bias of -3.2% and
`an SD of i7.3%.
`
`DISCUSSION
`
`To satisfy commercial, clinical, and research oriented
`markets for personal energy balance monitoring, a wear-
`able sensor module must satisfy four key criteria. The sensor
`module must be 1) seamless with daily living (comfortable,
`convenient, and socially acceptable), 2) sufficientiy accu-
`rate for multiple life activities (indoors and outdoors), 3)
`
` 7;
`
`O 8
`
`C9
`
`A
`3
`"V 10.
`0-‘
`g
`%
`'6
`Q)
`
`0-
`
`§_,_1o.
`°
`
`,
`
`,
`~ distance
`. —mean
`.
`—+/- 1.96 SD
`
`
`
`.
`
`‘
`
`t
`
`..
`1 .5
`
`4
`
`2
`
`,
`
`2.5
`
`,
`B
`3 30’
`5 20,
`"
`E
`8 10*
`> 0
`Q)
`'
`
`§_,-1o-
`'3-20:
`40
`
`,
`o V02max
`
`—mean
`.
`—+/- 1.96 SD
`
`60
`
`.
`80
`
`.4
`100
`
`,
`C
`? 3°’
`‘-5
`«n 20'
`31:2
`g 10-
`8
`r
`
`.
`
`§-1o-
`U-20:.
`100
`
`
`
`
`,
`,
`
`« calories
`—mean
`.
`—+/- 1.96 SD
`
`
`
`..
`
`150
`
`200
`
`250
`
`average calories (kcal)
`
`
`300
`
`average distance (miles)
`
`average VO2max (ml/kg/min)
`
`FIGURE 5 Bland Altman plots simultaneously comparing reference CPX data and earbud (device under test |DUT|) benchmark data collective. A.
`Distance traveled. B. V0,.” (mL-kg '-min '); EI-I (kcal). Nine subjects were studied in th CPX validation group study. The mean is shown by the
`green line. and the 1.96 SD (95% limits of agreement) boundaries are shown by the red lines.
`
`1050 Otficl Journal of the American College of Sports Mediche
`
`http://www.acsm-msse.org
`
`Copyright 9 2014 by the American Colege of Sports Medicine. Unauthorized reproduction of this article is prohibited.
`
`VALENCELL_0005
`
`<11
`
`UZ L
`
`L’
`
`Ua
`
`n QL
`
`L!
`:1
`L
`
`QC
`
`an
`LL!
`
`
`
`mm
`
`U Cm Um Qm Z(
`
`>1
`
`::
`
`‘W
`
`When compared with the approaches in HR chest strap
`for estimating energy expenditure (5,6), the earbud sensor
`algorithms for estimating TEE and V/02m, are also note-
`worthy. It is likely that the fixed location of the earbud
`with respect to the spine increases the accuracy of activity
`measurements, which feed the TEE and the V02...” models.
`However, some important limitations of these algorithms
`are of note. First, these algorithms have demonstrated sub-
`stantial efficacy in estimating TEE and \702,,,,,,, under
`walking, jogging, and mmring conditions, conditions com-
`mon for clinical CPX evaluations. However, it is yet to be
`determined how accurate these algorithms will be at esti-
`mating these parameters during everyday life activities and
`other exercising regimens, such as weight lifiing, swimming,
`contact sports, daily household activities, and the like. There
`are several studies emphasizing the importance of caution
`when applying V02 estimation models to universal PA
`(9,12,22,25,28,30,32,33). Moreover, it is not clear how well
`these algorithms will predict REE or energy expended during
`sedentary activity, where caloric expenditure is dominated
`by the metabolic rate of the individual as opposed to PA.
`Second,
`the V02,“ range of participants in the valida-
`tion study was relatively small: fiorn approximately 50 to
`65 mL-kg"-min" (Fig. 5). Indeed, the relatively low R2
`coefficient for estimated V02“, and the ostensible non-
`linear bias of Figure 5B together suggest that the accuracy
`of the V02“, model cannot be affirmed with the current
`data set. Finally, to be clinically relevant, the predictions for
`TE and V02“, would ideally be even more accurate, with
`the goal of an SD of less than 5%.
`To address these areas for improvement, future work
`should evaluate the efiicacy ofthe earbud sensor at estimating
`
`REFERENCES
`
`l. Altman DG, Bland JM. Mersurement in medicine: the analysk of
`method comparison studies. Statistician. l983;32(3):307 l7.
`2. Anderson DR, Whitmer RW, Goetzel RZ, et al. The relationship
`between modifiable health risks and group level health care ex
`penditures. Am J Health Promot. 2000;l5( l):45 52.
`3. Asada HH, Shaltis P, Reisner A, Sokwoo R, Hutchinson RC.
`Mobile monitoring with wearable photoplethysmogaphic bio
`sensors. IEEE Eng Med Biol Mag. 2003;22(3):28 40.
`4. Bland JM, Altman DG. Statistical methotk firr essing agree
`ment between two methods of clinical mezsurement. lancet.
`l986;l(8476):307 10.
`5. Brage S, Brage N, Franks PW, et al. Branched equation modeling
`of simultaneous accelerornetry and heart rate monitoring improves
`estimate of directly measured physical activity energy expenditure.
`J Appl Physiol. 2004;96(l):343 51.
`6. Brage S, Brage N, Franks PW, et al. Hierarchy of individual cali
`bration levels for heart rate and accelerometry to measure physical
`activity. J Appl Physiol. 2007:l03:682 92.
`7. Bravata DM, Smith Spangler C, Sundaram V, et al. Using pe
`dometers to increase physical activity and improve health. JAMA.
`2007;298(l9):2296 304.
`8. Crouter SE, Schneider PL, Karabulut M, Basett DR Jr. Validity of
`10 electronic pedometers for measuring steps, distance, and energy
`cost. Mad Sa‘ Sports Exerc. 2003;35(8): 1455 60.
`
`TEE and VO2,,,,,, in a larger cohort group having a broader
`range of aerobic capacity, ranging from approximately 35
`to 70 mL-kg‘ '-min_'. Furthermore,
`the earbud sensor
`should be put to the test of estimating TEE during a broader
`set of activities than simple treadmill testing, using the
`standard doubly labeled water (DLW) methodology as a
`benchmark. In addition, the ability of the earbud sensor to
`estimate the resting metabolic rate (REE) of subjects should
`also be assessed.
`
`lrnproving the accuracy of these assessments will rely on
`1) optimizing algorithms based on a larger study sample of
`subjects exercising in more diverse enviromnents (such as
`daily life activities) and 2) adding additional biometrics to
`the predictive algorithms for energy expenditure and
`\702,,,,,,. Valencell’s PerfcrmTek earbud sensor is comprised
`mostly of novel optomechanics and signal extraction algo-
`rithms. The accuracy of the HR algorithms is so high,
`approaching machine error, such that it is unlikely that ad-
`ditional improvements can be made in the optomechanical
`sensor module for accuracy. Rather, advancements are likely
`to arise from the development of an optimized statistical
`model that incorporates personalized REE estimations into
`the model. The algorithms for estimating personalized REE
`can be developed by evaluating the PPG profile of subjects
`at rest with a gas-exchange analysis benchmark (REE test-
`ing) and by identifying new blood flow profile features that
`correlate with the gas-exchange analysis data.
`
`The validation testing it this research was funded it part by the
`National Institutes of Health via Phase I SBIR 1R43DKo83141 —01A1.
`There are no conflicts of interest.
`The results of the present study do not constitute endorsement by
`the American College of Sports Mediche.
`
`9. Daniels .IT. A physio1ogist‘s View of running economy. Med Sci
`Sports Exenc. l985;l7(3):332 8.
`10. Frmk KL, Stevens VJ, Appel l.J. Axociations of Internet website
`use with weight change in a long term weight loss maintenance
`program. J Med Intenra Rm. 20l0;l2(3):e29.
`11. Gallup Healthways Well Being Index. 2012. Available from:
`http://www.gallup.oom/poll/lS7505/americans exercising sliyrtly
`20l2.aspx
`12. Harris C, Debeliso M, Adams K]. The effects of rtmning speed
`on the metabolic and mechanical energy costs of nrnning. J Exerr:
`Physiol. 2003;6(3):28 37.
`13. Headphones: Ownership & Application Study, 2012. The NPD
`Group. August 2012. p. 4.
`14. Jiang HH, Asada HH, Gibbs P. Active noise cancellation using
`MEMS accelerometers for motion tolerant wearable biosensors.
`
`ln: Cotfaarce Proceedings ofthe IEEE Engineering in Medicine
`and Biology Soeiay. 2004. pp. 2157 60.
`15. Kong YC, Ming S. lrnproving energy expend