PRELIMINARY REPORTS

USE OF THERMAL IMAGERY FOR ESTIMATION OF CORE BODY TEMPERATURE

DURING PRECOOLING, EXERTION, AND RECOVERY IN WILDLAND FIREFIGHTER
PROTECTIVE CLOTHING
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14

Thirimachos Bourlai, PhD, Riana R. Pryor, MS, Joe Suyama, MD, Steven E. Reis, MD,
David Hostler, PhD

ABSTRACT an average (± standard deviation) of 42.6 (±5.9) minutes and

a distance of 4.2 (±0.6) km on the treadmill. Mean heart rate
Background. Monitoring core body temperature to identify at the end of exercise was 152 (±33) bpm and core body tem-
heat stress in first responders and in individuals participat- perature at the end of exercise was 38.3◦ C (±0.7◦ C). A visual
ing in mass gatherings (e.g., marathons) is difficult. Objec- relationship and a strong correlation between core temper-
tive. This study utilized high-sensitivity thermal imaging ature and thermal imaging of the face were identified in all
technology to predict the core temperature of human subjects subjects, with the closest relationship and best agreement oc-
at a distance while performing simulated field operations
For personal use only.

curring during exercise. The Bland-Altman test of agreement

wearing thermal protective garments. Methods. Six male during exercise revealed the majority of measurement pairs
subjects participating in a study of precooling prior to ex- to be within two standard deviations of the measured tem-
ertion in wildland firefighter thermal protective clothing had perature. Conclusions. High-resolution thermal imaging in
thermal images of the face captured with a high-resolution the middle-wave infrared spectrum (3–5 µm) can be used to
thermal imaging camera concomitant with measures of core accurately estimate core body temperature during exertion
and skin temperature before, during, and after treadmill ex- in a hot room while participants are wearing wildland fire-
ercise in a heated room. Correlations and measures of agree- fighting garments. Although this technology is promising, it
ment between core temperature and thermal imaging–based must be refined. Using alternative measurement sites such
temperature were performed. Results. The subjects walked as the skin over the carotid artery, using multiple measure-
ment sites, or adding pulse detection may improve the esti-
mation of body temperature by thermal imagery. Key words:
heat stress; thermography; thermometry; performance; as-
sessment
Received August 23, 2012, from the Lane Department of Com-
puter Science and Electrical Engineering (TB), West Virginia Univer- PREHOSPITAL EMERGENCY CARE 2012;16:390–399
sity, Morgantown, West Virginia; and the Department of Emergency
Medicine, Emergency Responder Human Performance Laboratory
(RRP, JS, SER, DH), and the Department of Medicine, Division of INTRODUCTION
Cardiology University of Pittsburgh (SER), Pittsburgh, Pennsylva-
nia. Revision received January 6, 2012; accepted for publication Jan- Exertional heat stress is common in many occupational
uary 7, 2012. groups, including soldiers and firefighters.1–4 Heat
The authors thank the participants for their time and dedication to stress that is created while wearing protective equip-
firefighter health and safety. They would also like to thank Ms. Jen- ment is termed uncompensable heat stress (UHS). UHS
nifer Erin, MS, and Priya Khorana, MS, for their assistance with data
occurs from an inability of a person’s body to dissi-
collection. They would also like to thank FLIR Systems, Inc., for their
technical support during this project. pate heat when he or she is performing routine op-
erations while wearing thermal protective clothing
No conflicts of interest, financial or otherwise, are declared by the
authors. and other forms of personal protective equipment. Be-
cause of this increase risk of heat-related illness, early
Address correspondence and reprint requests to: David Hostler, identification and mitigation of heat stress are criti-
PhD, University of Pittsburgh, Department of Emergency Medicine, cal for sustaining these types of operations. However,
3600 Forbes Avenue, Suite 400A, Pittsburgh, PA 15261. e-mail: monitoring of core body temperature is difficult in
[email protected] the field. Measurement of core body temperature us-
doi: 10.3109/10903127.2012.670689 ing rectal thermometers or ingestible capsules during

390
 Bourlai et al. THERMAL IMAGING of BODY TEMPERATURE 391

routine operations is impractical, expensive, and diffi- cate that this technology is capable of predicting the
cult to integrate into daily operations.5,6 Additionally, core body temperature of human subjects at a dis-
during large events (e.g., marathons) and other mass tance. The camera was used to monitor the faces of par-
screening situations, it is too labor-intensive to posi- ticipants performing simulated field operations while
tion workers to deploy any type of direct temperature wearing wildland firefighter protective garments. The
monitoring system for large numbers of individuals. main benefit of performing such a study is to assist
A true surrogate temperature assessment tool or in understanding the level of efficiency that thermal
technique that can accurately estimate the core tem- imaging–based monitoring systems have in detecting
perature during exertion and at rest has been difficult exertional heat stress.
to identify. Oral and axillary thermometers, temporal
artery scanners, and skin temperature devices used in
adults have poor correlation and accuracy in their abil-
METHODS
ity to monitor core body temperature at rest and un- Ethics Approval
der exertional or uncompensable heat stress. One po-
tential solution that may be a surrogate for real-time This study was approved by the University of Pitts-
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14

field monitoring of core body temperature is an imag- burgh Institutional Review Board. Written informed
ing process that takes advantage of the cutaneous vas- consent was obtained from all subjects prior to any in-
culature in human skin. tervention.
Superficial arteries are located immediately under
facial skin. This superficial arterial distribution of the Participants
face branches from the external carotid artery shortly
after leaving the aortic arch. The compliance of the Six healthy men volunteered to participate. These
artery during pulsatile flow results in capillary flow, subjects were a convenience sample from a larger
diffusion effect, and tissue deformation followed by study comparing ingestion of ice slurry vs. temperate
the convection of heat to the skin. As such, it may be water before exercise. Results of the larger study are
For personal use only.

possible to take advantage of the anatomy of the face reported separately. In this report, all subjects were
and facial vasculature to estimate core body tempera- precooled prior to exertion and were followed during
ture using thermal imaging. exertion and recovery. The subjects were in good
An improved approach of using standoff physio- health, 18 years of age or older, and free of diagnosed
logic monitoring through passive (thermal) imaging cardiovascular, metabolic, and respiratory disease.
to measure cardiac pulse was proposed by Bourlai et
al.7,8 In that work, a highly sensitive thermal imaging Visit 1: Assessment of Cardiopulmonary
camera was used to monitor the blood vessels of hu-
Fitness
man faces over time. The camera used was a lower-
end model (lower spatial resolution) than the camera A study physician determined subject eligibility us-
used in this study, both operating at the same spectral ing medical history, physical examination, resting vital
(thermal) band (middle-wave infrared, 3–5 µm). The signs, and a 12-lead electrocardiogram (ECG). The sub-
advantage of middle-wave infrared cameras over cam- jects were asked to refrain from consuming caffeine,
eras operating in the visible or active infrared (0.8–2.5 nicotine, and alcohol and from exercising 12 hours be-
µm) bands is that they are capable of detecting the heat fore the screening. Nude body mass was measured,
radiation of human skin. In addition, middle-wave in- to the nearest 0.005 kg, using a floor weighing scale
frared cameras can be set to detect heat radiation ev- (Kern ITB Version 2.0, Balingen, Germany). Skinfold
ery 1/30 second while achieving high thermal sensitiv- thickness was measured at three sites (chest, abdomen,
ity (i.e., close to 0.01◦ C). Another advantage of thermal and thigh) in duplicate with a skinfold caliper (Beta
middle-wave infrared imaging is that it can operate in Technology Inc., Cambridge, MD). The mean values
either ideal or nonideal conditions; i.e., the accuracy were used to calculate skinfolds and estimate percent-
of the estimation of the facial skin temperature by the age body fat as previously described.9
thermal-based system is invariant to ambient illumina- Subjects performed a graded exercise Bruce tread-
tion. mill test with open-circuit spirometry (Parvomedics
In this work, we studied an application of this tech- Inc., Sandy, UT) to measure maximal oxygen consump-
nology for field temperature and heat stress identifica- tion. Oxygen and carbon dioxide analyzers were cal-
tion under the rest, exertion, and rehabilitation aspects ibrated prior to each test using standard referencing
of first-responder operations. gases. A 12-lead ECG was obtained every 3 minutes
For that purpose, we have employed a middle-wave during and after exercise. A cardiologist interpreted
infrared imaging system that has the highest thermal the test results to identify undiagnosed ischemic re-
sensitivity in the market. Experimental results indi- sponse to exercise.
 392 PREHOSPITAL EMERGENCY CARE JULY/SEPTEMBER 2012 VOLUME 16 / NUMBER 3

Visit 2: Assessment of Core Body

Temperature during Precooling, Exertion,
and Recovery in Wildland Firefighter
Protective Garments
Instrumentation
Core temperature (Tc) was measured with an in-
gestible capsule and radio receiver (CoreTemp, HQ
Inc., Palmetto, FL). Subjects ingested the capsule 8–12
hours before the protocol to avoid the effect of food or
drink in the gastrointestinal tract.10 Upon arrival to the
laboratory, urine specific gravity (USG) was assessed
with a handheld refractometer (SUR-NE, Atago U.S.A.,
Bellevue, WA) to ensure euhydration (USG ≤1.020)
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14

prior to beginning the protocol.

A heart rate (HR) monitor (Model T31, Polar Elec-
tro Oy, Kempele, Finland) was applied along with four
reusable surface thermistors (Physitemp Instruments,
Inc., SST-1 Skin Sensors, Clifton, NJ) placed on the left
pectoralis major, infraspinatus, and the midpoint of the
triceps brachii and quadriceps femoris muscles. Mean
skin temperature (Tsk) was calculated using the fol- FIGURE 1. Sample profile facial images illustrating that subcutaneous
lowing formula: Tsk = chest (0.25) + back (0.25) + information on different subjects varies, and that thermal facial im-
thigh (0.3) + arm (0.2) (1).11 A fifth surface thermis- ages are capable of providing significant information about the tem-
perature of the human body. Note that originally the camera acquires
tor was placed over the left carotid artery identified by
For personal use only.

video frames where each pixel has a temperature value ranging from
palpation midway between the clavicle and the angle ∼24◦ C to 41◦ C. Then these frames are converted to grayscale images
of the mandible. for display purposes.

Protocol HR greater than 10 beats·min−1 over age-predicted

All subjects in this report completed the exercise trial maximum (220 – age); 3) unsteady gait, making it
after consuming ice slurry (0.1◦ C) with added syrup unsafe to continue walking; or 4) subject request.
used for flavor, creating a 3% carbohydrate beverage.
The ice slurries were made with a commercially avail-
Thermal Imaging Camera
able frozen drink machine (Ninja MasterPrep model
QB900 30, Ninja, Boston, MA). During precooling, the The thermal camera used in this study was a high-
subjects sat in wildland firefighting garments (Lion definition middle-wave infrared camera (FLIR Sys-
Apparel, Dayton, OH) while consuming 1.25 g·kg−1 of tems, Inc., Wilsonville, OR). It is capable of acquir-
drink every 5 minutes to ensure a standardized inges- ing thermal imprints of human skin and analyzing the
tion rate, totaling 7.5 g·kg−1 over 30 minutes. thermal distributions and temporal variations (Fig. 1).
The subjects commenced the treadmill protocol The camera generates high-definition thermal images
approximately 5 minutes after finishing the drink and can operate in diverse testing environments. It fea-
ingestion. The subjects wore the wildland firefighting tures a high-resolution 1,024 × 1,024 (i.e., the spatial
garments, a helmet, and a 20.4-kg backpack during ex- resolution of the system at any standoff distance) in-
ercise. The participants walked on a treadmill (Desmo, dium antimonide (InSb) focal plane array achieving
Woodway USA, Inc., Waukesha, WI) for 45 minutes at high megapixel image resolution in a single thermal
4.0 mph in a heated enclosure (38.8◦ C ± 1.0◦ C, 19.2% image. The effective focal length of the camera used is
± 2.2% relative humidity). This protocol mimics the 50 mm. The detector used has 1,024 × 1,024 pixels, with
U.S. Forest Service Pack Test required of wildland a pixel pitch of 18 µm. The standoff distance used was
firefighters.12 The subjects were encouraged to main- approximately 1.5 m, where the target (human face) di-
tain this pace, but whenever necessary they decreased mensions are approximately 0.152 × 0.229 m. In terms
the speed to a level that allowed them to complete the of the output of the camera, the horizontal and vertical
full protocol time interval. Each subject exercised for fields of view are the same (i.e., 0.365 rad) and the spot
45 minutes, in total, unless any one of the following size (1 p × 1 p) is 0.00054 m2 . The spectral range of the
conditions was met: 1) core temperature ≥39.5◦ C; 2) camera is 3–5 µm. The camera has a 14-bit dynamic
 Bourlai et al. THERMAL IMAGING of BODY TEMPERATURE 393

range, and a noise-equivalent temperature difference thermal-based videos of the human face of each subject
of less than 25 mK. The camera was outfitted with a (i.e., 35 minutes of video in total per subject per cam-
50-mm middle-wave infrared lens also provided from era). During the subsequent exercise period, the sub-
FLIR Systems. jects were asked to walk on a treadmill that was placed
The thermal sensitivity for an infrared camera is in a heated room while again being recorded by the
measured in millikelvins (mK). Thermal sensitivity of two cameras. Exercise duration was 45 minutes, dur-
the camera is important and directly correlates with ing which the subjects were recorded every 5 minutes
image quality. Current cameras are able to detect tem- (10 times in total). Finally, during the “recovery step,”
perature differences as small as 0.03◦ C. This equates to the subjects’ faces were recorded every 5 minutes, i.e.,
a thermal sensitivity of 30 mK. Greater sensitivity of 13 times in total. Sample facial images acquired at both
the camera leads to higher quality and more accurate the visible and thermal bands at one instance of each
images. The thermal camera that was used in the study of the four aforementioned steps of the recording pro-
has the highest sensitivity currently available in com- cess (baseline, precooling, exercise, and recovery) are
mercial thermal cameras. illustrated in Figure 2.
The associated camera software (provided by FLIR The proposed method for the estimation of the core
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14

Systems) was used not only to perform data collec- body temperature using thermal imaging is contact-
tion, but also to perform regular calibration (before ac- free; hence, in the absence of good tracking, even the
quiring each set of thermal full frontal face images). slightest movement of the subject being recorded by
Calibration ensures that the camera operates at its the camera will shift the region of interest (subject’s
optimum performance and guarantees measurement face) from its initial selection. Although tracking meth-
accuracy and reliability. The software tool provided by ods could have been used to compensate for motion,
FLIR Systems also performs nonuniformity correction, they were not used in this study. The main reason was
linearity correction, and bad pixel replacement (less to manually select a sequence of good-quality frames
than 0.5% of the pixels are bad/dead) as postprocess- to perform the measurements and prove our concept,
ing operations on the imagery. The software was set avoiding any noisy measurements due to a potential
For personal use only.

to control the temperature scale limits (e.g., setting the inefficiency of the tracking algorithm. Thus, we per-
temperature range from 28◦ C to 40◦ C that is the typical formed an estimation of the maximum facial skin tem-
range of human body temperature) during data collec- perature over a number of frames, first by perform-
tion. ing a manual selection of the region of interest, and
then by manual tracking the subjects’ faces (frame by
frame). For every recording (30 frames per second) that
Data Collection and Analysis
we obtained using the camera during rest and exercise,
The standoff (camera-to-face) distance was set to 1.5 we monitored only the first 50 frames, i.e., <2 seconds
meters. Data were collected over a time period of 30 (frames of nonfrontal faces were disregarded to ex-
days. In the beginning of the session, the subjects were clude noisy estimations from our measurements). Each
briefed about the data-collection process, after which recording with the camera was synchronized with the
they signed a consent document. The six subjects par- temperature estimation using the other sensors (e.g.,
ticipating in this study resulted in a database contain- skin). Manual tracking was also helpful because, dur-
ing 31 thermal and 31 visible videos of human faces ing exertion, most of the subjects were constantly mov-
for each subject, resulting in a total of 366 videos (∼100 ing and, thus, a sequence of good-quality full frontal
GB of data). The visible light camera (Canon 5D EOS facial images (frames) was not possible to be recorded.
Mark II, Canon U.S.A., Lake Success, NY) was used to In such cases, it is very difficult to find areas of the
illustrate the facial appearance of each subject during face that allow a confident estimate of the maximum
each phase of the protocol. Figure 2 illustrates the facial temperature in the facial region (such as the periocular
images of one of the participants acquired by the vis- region). Hence, the manual selection of the region of
ible and the thermal cameras at four selected frames, interest had to be arbitrary (usually an oval or a rect-
representing the four different steps of the monitoring angular region was selected), and the region of interest
process of our study. was always (for all frames used to estimate the tem-
To establish a baseline for our study, the facial area perature) selected to be small enough to fit within the
of each subject was recorded by both the visible and facial region of the subject being recorded.
the thermal cameras while the subject was relaxing Figure 3 outlines the salient stages used to passively
on a chair. Then, during the “precooling step” every estimate core body temperature with thermal imaging.
5 minutes, the subject’s face was recorded while he For each frame (1,024 × 1,024 pixels), we first selected
was sitting on a chair ingesting ice slurry for 3 min- a region of interest within the subject’s face (a vari-
utes in total by both cameras. This process was re- able area, and thus a variable number of pixels). Within
peated seven times, resulting in two sets of visible- and each region of interest, we estimated the maximum
 394 PREHOSPITAL EMERGENCY CARE JULY/SEPTEMBER 2012 VOLUME 16 / NUMBER 3
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14
For personal use only.

FIGURE 2. Sample data collection. Note the region of interest in the thermal faces where the maximum temperature of the subject was measured.
Also shown is the timeline for each step during which the measurement was acquired.

temperature over the first 50 frames of each record- Statistical Analysis

ing (as described above). The median of the maximum
temperatures was then calculated over 50 frames. This We analyzed the data acquired at each step (epoch)
measure was used as an estimate of the core body of the monitoring process, i.e., precooling, exercise,
temperature for each recording. Finally, these thermal and recovery. Within each epoch, we measured the
imaging–based temperature measurements (estimates 1) core body temperature, 2) skin temperature over
of core temperature) were compared with the core the carotid artery, 3) mean skin temperature, and
body temperature measured using the capsule, as well 4) facial skin temperature, estimated using thermal
as the facial and skin temperatures using the skin sen- imaging. Then, we computed the correlations between
sors. thermal imaging–estimated temperature and the con-
Note that, although the standoff distance was fixed ventional core and skin temperature measurements.
to be 1.5 meters, the region of interest (selected area Bland-Altman plots were also generated to test the
within each subject’s face) varied (depending on the agreement of temperatures during exertion.
subject’s facial structure and movement towards and
away from the camera). We estimated the average RESULTS
number of pixels used for the calculation of the maxi-
mum temperature to be about ∼160,000 pixels (i.e., 400
Participant Characteristics
× 400 pixels, if the region of interest is selected to be The subjects were fit men (Table 1). Five subjects com-
rectangular). pleted the entire 45-minute bout of exercise, whereas
 Bourlai et al. THERMAL IMAGING of BODY TEMPERATURE 395
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14
For personal use only.

FIGURE 3. Illustration of the process of computing the median of the maximum temperatures over one (out of the 31 total) video recording using
the thermal camera.

one subject reached volitional fatigue after 30.5 min- The Bland-Altman test of agreement during exer-
utes, resulting in an average (± standard deviation) cise revealed that the majority of measurement pairs
exercise duration of 42.6 (±5.9) minutes and walking were within two standard deviations of the measured
distance of 4.2 (±0.6) km. At the end of exercise, the temperature (Fig. 5). The bias between core body tem-
mean HR was 152 (±33) bpm and the mean core body perature and thermal imaging of the face was –0.07◦ C
temperature was 38.3◦ C (±0.7◦ C). (1.4◦ C). The biases from the temperature sensor placed
on the skin over the carotid artery (–1.9◦ C [1.5◦ C])
and body mean skin temperature (–1.3◦ C [1.4◦ C]) were
Thermal Measurements
both larger than the bias with core body temperature
A visual relationship between core temperature and during exercise.
thermal imaging of the face could be identified in
all subjects, with the closest relationship and best
agreement occurring during exercise (Fig. 4). Correla- DISCUSSION
tions were identified for thermal imaging and all mea- High-resolution thermal imaging of the facial features
sured temperatures with the exception of core body during exertion in wildland firefighting garments cap-
temperature during precooling (Table 2). However, a tured measures of temperature that were highly cor-
negative correlation between thermal imaging temper- related and in close agreement with core body tem-
ature and mean skin temperature was identified dur- perature. While the bias of the agreement was greater
ing precooling and recovery. during precooling and recovery, correlations remained
high during these intervals, indicating that it is possi-
ble to identify a correction factor to estimate core body
TABLE 1. Subject Characteristics
temperature during periods of nonexertion. Potential
VO2max applications for this technology include screening first
Age (yr) Height (cm) Mass (kg) (mL/kg/min) Body Fat (%)
responders (e.g., firefighters and hazardous materials
34.5 ± 9.1 181.9 ± 5.8 89.7 ± 17.7 50.4 ± 8.8 14.6 ± 6.7 technicians) for heat stress and deployment to monitor
Values are expressed as mean ± standard deviation. participants at mass-gathering athletic events such as
VO2max = maximum volume of oxygen consumption. marathons or ultra-endurance races.
 396 PREHOSPITAL EMERGENCY CARE JULY/SEPTEMBER 2012 VOLUME 16 / NUMBER 3
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14
For personal use only.

FIGURE 4. Body core temperature (squares) and thermal imaging temperature (open circles) of the face measured during precooling, exercise,
and recovery in the six subjects.
 Bourlai et al. THERMAL IMAGING of BODY TEMPERATURE 397

TABLE 2. Correlations between Thermal Imaging Temperature and Other Measured Temperatures
Precooling Exercise Recovery

Core Carotid Mean Skin Core Carotid Mean Skin Core Carotid Mean Skin
r 0.086 0.560 −0.421 0.411 0.527 0.473 −0.253 0.575 −0.342
p-value 0.590 < 0.001 0.006 0.001 < 0.001 < 0.001 0.026 < 0.001 0.002

It is not clear why the amount of bias differs between temperature across the face was not uniform. The areas
exertion and resting phases. The protocol required the displaying the maximum temperature were measured
subject to work in a heated room while wearing protec- and typically corresponded with the course of the fa-
tive clothing, so the higher ambient temperature may cial artery. Given the higher HR during exertion and
have contributed to these differences. However, the activation of the thermoregulatory mechanisms, it is
likely that these areas were well perfused with blood,
during which the temperature more closely approxi-
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14

mated core body temperature than might be observed

during rest. Future iterations of the technology may
need to consider HR and ambient temperatures when
estimating core body temperature.
Correlations between core body temperature and
thermal imaging temperature were inconsistent over
the course of the study. Although a positive correla-
tion was seen during exertion, a negative correlation
was seen during recovery and no correlation could be
identified during precooling. This lack of a relationship
For personal use only.

during precooling may be due to local cooling of the fa-

cial structures from the ice slurry. The inverse relation-
ship seen in the recovery phase appears to be the prod-
uct of a static thermal imaging temperature over the
recovery period and a falling core temperature. How-
ever, the correlation between thermal imaging tem-
perature and skin temperature over the carotid artery
was always positive and significant. This might be ex-
pected given that the facial artery is a branch of the ex-
ternal carotid artery. However, the agreement between
the thermal imaging of the face and the skin over the
carotid artery was worse than the agreement with core
body temperature. We hypothesized that this may be
due to the thickness of the skin and physiologic struc-
tures overlying the carotid artery. Furthermore, varia-
tion in body habitus among individuals may limit the
utility of the neck as a measurement site.
Previous work using thermal imagers to estimate
core body temperature have largely examined the
ability to detect fever. A study of 310 individuals
reporting to the emergency department during the
severe acute respiratory syndrome (SARS) epidemic
found that measuring temperature with thermal
imagers in the area of the medial canthus provided
the best estimate of the temperature measure with
a tympanic thermometer.13 In that study, it was
determined that a thermal imager could be set to
signal an alarm when measured temperature exceeded
36.3◦ C, to identify individuals requiring additional
FIGURE 5. Bland-Altman tests of agreement between thermal imaging
temperature of the face and body core temperature (top panel), skin
screening. A follow-up study of 86 febrile and 410
temperature over the carotid artery (middle panel), and body mean normothermic patients reported a sensitivity of 90.7%
skin temperature (bottom panel). and a specificity of 75.8% to detect a fever, defined as
 398 PREHOSPITAL EMERGENCY CARE JULY/SEPTEMBER 2012 VOLUME 16 / NUMBER 3

body temperature >37.7◦ C.14 More recently, a study the room temperature and illumination were stable
employing three different thermal imaging cameras and placed away from the subject. Outdoor condi-
reported sensitivity of 91% and 90% and specificity tions or nearby heat/light sources may confound the
of 86% and 80% to detect fever in the emergency operation.
department via the two best performing cameras.15
The present report expands upon previous work
by examining the utility of a high-resolution thermal CONCLUSIONS
imaging camera to estimate core body temperature
during rest and exertion. We are aware of only one High-resolution thermal imaging in the middle-wave
other study using similar technology to estimate tem- infrared spectrum (3–5 µm) can be used to accurately
perature during exertion.16 That study contrasts with estimate core body temperature during exertion in
the present findings in that the rise of core body tem- a hot room while participants are wearing wildland
perature during exercise was not detected by the ther- firefighting thermal protective clothing. Although this
mal imaging device. However, that study deployed a technology is promising, it must be refined. Using al-
near-infrared camera with an extended range (0.8–2.5 ternative measurement sites such as the skin over the
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14

µm) and lower-resolution (320 × 256) input. In the carotid artery, using multiple measurement sites, or
present report, the thermal imager featured a high- adding pulse detection is expected to improve the esti-
resolution 1,024 × 1,024 InSb focal plane array achiev- mation of body temperature by thermal imagery.
ing megapixel image resolution in a single thermal
image. The high-resolution input and high thermal
sensitivity of the camera allowed for more precise References
measurements. The study by Teunissen and Daanen
(2011) reported difficulty with imaging the medial can- 1. Lieberman HR, Bathalon GP, Falco CM, Kramer FM, Morgan
CA 3rd, Niro P. Severe decrements in cognition function and
thus in all subjects and all conditions.16 The advantage mood induced by sleep loss, heat, dehydration, and undernu-
of our methodology is that it is capable of scanning trition during simulated combat. Biol Psychiatry. 2005;57:422-9.
For personal use only.

the entire face of the subject under study for maximum 2. Horn GP, Gutzmer S, Fahs CA, et al. Physiological recovery
temperatures, providing greater flexibility in obtaining from firefighting activities in rehabilitation and beyond. Pre-
an accurate estimate of temperature. hosp Emerg Care. 2011;15:214-25.
3. Hostler D, Reis SE, Bednez JC, Kerin S, Suyama J. Comparison
Near-term applications for this technology could in- of active cooling devices with passive cooling for rehabilitation
clude screening multiple subjects in short periods of of firefighters performing exercise in thermal protective cloth-
time to identify those with suspected hyperthermia, ing: a report from the Fireground Rehab Evaluation (FIRE) trial.
such as during sporting events, responders entering Prehosp Emerg Care. 2010;14:300-9.
emergency incident rehabilitation, and military pa- 4. Colburn D, Suyama J, Reis SE, et al. A comparison of cooling
techniques in firefighters after a live burn evolution. Prehosp
trols. After correction factors have been identified for Emerg Care. 2011;15:226-32.
nonexertion conditions, the technology could be used 5. Pryor RR, Seitz JR, Morley J, et al. Estimating core temperature
to identify those emergency responders and workers with external devices after exertional heat stress in thermal pro-
who are too hyperthermic to return to work. It can tective clothing. Prehosp Emerg Care. 2012;16:136-41.
also be used for screening large numbers of subjects for 6. Ganio MS, Brown CM, Casa DJ, et al. Validity and reliability of
devices that assess body temperature during indoor exercise in
fever at ports of entry and public areas during periods the heat. J Athl Train. 2009;44:124-35.
of pandemics or the release of biologic agents. 7. Bourlai T, Buddharaju P, Pavlidis I, Bass B. On enhancing
cardiac pulse measurements through thermal imaging. Pre-
sented at: International Conference on Information Technology
and Applications in Biomedicine, Larnaka, Cyprus, November
LIMITATIONS AND FUTURE RESEARCH 2009.
8. Bourlai T, Buddharaju P, Pavlidis I, Bass B. Methodological ad-
The principal limitation of this study is the small num- vances on pulse measurement through functional imaging. In:
ber of subjects who were available for analysis. In Garbey M, Bass BL, Collet C, de Mathelin M, Tran-Son-Tay R
spite of this, strong correlations and agreement dur- (eds). Computational Surgery and Dual Training, Advances in
ing exertion encourage additional studies in this area. Pattern Recognition. New York: Springer, 2010, pp 101-21.
9. Jackson AS, Pollock ML. Prediction accuracy of body density,
These data are further strengthened by employing
lean body weight, and total body volume equations. Med Sci
rigorous data-collection methods and direct measure- Sports. 1977;9:197-201.
ment of core body temperature, instead of surrogates 10. Wilkinson DM, Carter JM, Richmond VL, Blacker SD, Rayson
(e.g., tympanic and oral thermometry) that may be MP. The effect of cool water ingestion on gastrointestinal pill
difficult to accurately use during or following exer- temperature. Med Sci Sports Exerc. 2008;40:523-8.
11. Ayling JH. Regional rates of sweat evaporation during leg and
tional heat stress.5 Additionally, we do not have data
arm cycling. Br J Sports Med. 1986;20:35-7.
for passive heating or febrile conditions, representing 12. Sharkey B. Development and validation of a job related work
additional areas of exploration. Data collection was capacity test for wildland firefighters. Presented at: Interna-
performed in a controlled indoor environment, where tional Association of Wildland Fire Safety Summit, Sydney,
 Bourlai et al. THERMAL IMAGING of BODY TEMPERATURE 399

Australia, 1999. Available at: http://www.iawfonline.org/ 15. Nguyen AV, Cohen NJ, Lipman H, et al. Comparison of 3 in-
proceedings.php. Accessed October 19, 2011. frared thermal detection systems and self-report for mass fever
13. Ng EY, Kaw GJ, Chang WM. Analysis of IR thermal imager for screening. Emerg Infect Dis. 2010;16:1710-7.
mass blind fever screening. Microvasc Res. 2004;68:104-9. 16. Teunissen LP, Daanen HA. Infrared thermal imaging of the in-
14. Ng EY. Is thermal scanner losing its bite in mass screening of ner canthus of the eye as an estimator of body core temperature.
fever due to SARS? Med Phys. 2005;32:93-7. J Med Eng Technol. 2011;35:134-8.
Prehosp Emerg Care Downloaded from informahealthcare.com by McMaster University on 12/03/14
For personal use only.