Background:

Although thermal imaging can be a valuable technology in the prevention and management of
diabetic foot disease, it is not yet widely used in clinical practice. Technological advancement
in infrared imaging increases its application range. The aim was to explore the first steps in the
applicability of high-resolution infrared thermal imaging for noninvasive automated detection
of signs of diabetic foot disease.

Methods:

The plantar foot surfaces of 15 diabetes patients were imaged with an infrared camera
(resolution, 1.2 mm/pixel): 5 patients had no visible signs of foot complications, 5 patients had
local complications (e.g., abundant callus or neuropathic ulcer), and 5 patients had diffuse
complications (e.g., Charcot foot, infected ulcer, or critical ischemia). Foot temperature was
calculated as mean temperature across pixels for the whole foot and for specified regions of
interest (ROIs).

Results:

No differences in mean temperature >1.5 C between the ipsilateral and the contralateral foot
were found in patients without complications. In patients with local complications, mean
temperatures of the ipsilateral and the contralateral foot were similar, but temperature at the
ROI was >2 C higher compared with the corresponding region in the contralateral foot and to
the mean of the whole ipsilateral foot. In patients with diffuse complications, mean
temperature differences of >3 C between ipsilateral and contralateral foot were found.

Conclusions:

With an algorithm based on parameters that can be captured and analyzed with a high-
resolution infrared camera and a computer, it is possible to detect signs of diabetic foot
disease and to discriminate between no, local, or diffuse diabetic foot complications. As such,
an intelligent telemedicine monitoring system for noninvasive automated detection of signs of
diabetic foot disease is one step closer. Future studies are essential to confirm and extend
these promising early findings.

Discussion

Technological advances in infrared imaging, concerning both the speed of assessment and the
spatial resolution of image pixels, have increased possibilities to quantify thermal patterns and
perform automated analysis on acquired thermal images of patients feet.6 I n t he c urrent
study, we explored t he fi rst steps i n t he applicability of h igh-resolution infrared thermal
imaging for noninvasive automated detection of signs of diabetic foot disease. An algorithm
was developed for detecting signs of diabetic foot disease by measuring the temperature of
the plantar surface of the feet, based solely on parameters that can be captured and analyzed
with an infrared camera and a computer. With this algorithm, a good distinction could be
made between patients having no diabetic foot complications, local complications, or diffuse
complications. Patients without complications showed only small temperature differences
between feet. Patients with local complications such as a noninfected and nonischemic foot
ulcer or abundant callus showed locally increased temperatures of >2 C compared with both
the contralateral foot and the average temperature of the ipsilateral foot. Patients with diffuse
complications such as a foot ulcer with osteomyelitis or a Charcot foot showed an increased
mean temperature of >3 C compared with the contralateral foot. These results indicate that
advanced infrared thermal imaging may be applicable as diagnostic tool for noninvasive
automated detection of signs of diabetic foot disease. This may be applied for early detection
and timely management of diabetic foot complications, which could contribute to the
prevention of further, more devastating consequences.

In the only clinical study known to the authors that measured foot temperature in the diabetic
foot, Lavery and coauthors1214 define a difference of 4 F (or 2.2 C ) between a foot region
and t he corresponding region in the contralateral foot as clinically significant. The
temperature differences measured in the current study confirm this threshold as clinically
relevant. However, from the results of this study, it can be seen that more advanced infrared
cameras allow further specification of temperature differences between feet, where
temperature difference thresholds of 2 C apply to local complications such as neuropathic
ulcers and abundant callus, and temperature difference thresholds of 3 C apply to diffuse
complications such as Charcot foot, ulcers with osteomyelitis, and critical ischemia.

Further testing in larger groups of unselected patients is necessary to confirm the findings of
this pilot study, to refine the classification of complications based on measured temperature
differences, and to calculate diagnostic accuracy with parameters such as sensitivity and
specificity.

The necessity for manual annotation of the boundaries of the foot in the color image was a
limitation in this study.

It is unlikely that manual annotation of foot boundaries has affected the results, as adequate
accuracy could be guaranteed (Figure 2). However, automated analysis is not possible when all
feet require manual annotation. We are currently working on automated image analysis
similar to an already-described method.20 This method includes automated definition of foot
boundaries, calculation of mean temperatures, and comparison of temperatures with
contralateral regions. These developments are needed to achieve our goal of an intelligent
telemedicine monitoring system based on infrared imaging. Inclusion in this study was limited
to patients either with or without existing pathologies. Future studies should follow patients
without existing pathologies over time to investigate the diagnostic accuracy of the system in
detecting diabetic foot complications as early as possible. Another limitation was the rather
rudimentary differentiation in no, local, and diffuse complications. The local
complications neuropathic ulcer and abundant callus have a different clinical significance,
and therefore different referral times, but these signs could not be separated from the thermal
images obtained in this study. Future studies need to explore if further differentiation is
possible between these signs of diabetic foot disease based on thermal images.
Infrared temperature measurements have some limitations when applied for clinical purposes,
which have been described earlier.5,21 T he first i s the detection of local complications that a
re bilaterally present in the same foot region at the same time. This is visible, for example, in
patient 8, who had an ulcer with local temperature increase at the right hallux but also
increased temperature at the left hallux. As a result, the temperature difference between
these regions did not exceed 2 C . As shown in this study, this limitation can be overcome with
further optimization of the algorithms based on the comparison of temperature at the ROI
with the mean ipsilateral foot temperature. The second limitation is the detection of diffuse
complications present in both feet at the same time. Patient 5 had temperatura values in both
feet that are higher compared with the values measured in other patients. Based on the
infrared image only, it is not possible to confirm whether this patient has bilaterally Charcot
feet, bilaterally osteomyelitis, or just a pair of warm feet (e.g., due to the presence of
autonomic neuropathy). By combining infrared imaging with, for example, photographic
imaging, this limitation may be overcome, although it must be noted that the chances of
having such severe complications on both feet at the same time are very low. Finally, the value
of using absolute foot temperatura values for the detection of signs of foot disease is still not
clear. Foot temperatures may vary from person to person as a result of age- and sex-related
differences, presence of autonomic neuropathy or peripheral vascular disease, and
environmental factors such as ambient temperature.5 Properly controlled studies with
advanced infrared imaging in large groups of participants are needed to determine the
(additional) value of absolute foot temperature values for diagnostic purposes. Such studies
preferably conduct measurements over time within patients to determine intraindividual
temperature patterns and changes.

An intelligent telemedicine monitoring system as envisioned in our project is not yet close to
being used in daily clinical practice. Technological issues need to be resolved, including patient
positioning, camera positioning, the need or adding other imaging modalities, and automated
image registration and analysis.6 Feasibility studies are needed to establish the most optimal
requirements for such a system and the most effective application in daily life.

Subsequently, the (cost) effectiveness of using such a system for the prevention of diabetic
foot disease will have to be assessed. Prices of advanced thermal imaging systems are
dropping, but it is not clear whether prices are low enough to implement such a system as a
monitoring tool.
Quantitative Estimation of Temperature Variations in Plantar Angiosomes: A Study Case for
Diabetic Foot

Thermography is a useful tool since it provides information that may help in the diagnostic of
several diseases in a noninvasive and fast way. Particularly, thermography has been applied in
the study of the diabetic foot. However, most of these studies report only qualitative
information making it difficult to measure significant parameters such as temperature
variations.These variations are important in the analysis of the diabetic foot since they could
bring knowledge, for instance, regarding ulceration risks. The early detection of ulceration risks
is considered an important research topic in the medicine field, as its objective is to avoid
major complications that might lead to a limb amputation. The absence of symptoms in the
early phase of the ulceration is conceived as themain disadvantage to provide an opportune
diagnostic in subjects with neuropathy. Since the relation between temperature and ulceration
risks is well established in the literature, a methodology that obtains quantitative temperature
differences in the plantar area of the diabetic foot to detect ulceration risks is proposed in this
work. Such methodology is based on the angiosome concept and image processing.

According to previous studies, the temperature variation between corresponding regions in

both feet usually does not vary beyond 1C; a difference greater than 2.2C is considered as
abnormal [15, 27]. The ranges established by [6] were considered in order to determine
whether the plantar temperature is normal. Moreover, research results presented by Papanas
et al. [7] about plantar temperatura were taken into account.That study concluded that
patients with diabetes type 2 with and without neuropathy have a plantar temperature of 32.2
0.94C and 30.7 1.07C, respectively. The goal in this work is to identify differences
between corresponding areas of the right and left foot in order to obtain quantitative
information about the plantar temperature distribution. Furthermore, it addressed the early
detection of hot spots in an initial phase (small areas) with the aim of preventing their
expansion and future complications.

Both estimations could be useful for the specialist in early risk detection.The thermal images
used in this work correspond to patients diagnosed with diabetes type 2 and neuropathy.

Figure 4 shows one of the obtained images, which is divided into two sections corresponding
to the right and left foot.

These images are processed in order to classify their pixels. All the pixels from each image are
processed and classified in order to make the limits among the temperature regions clear.

As it was described in Table 1, there are eight classes with a representative temperature and
an associated color, and all the pixels are classified according to (1).

The importance of this process lies on the fact that in the original images it is difficult to
determine where a color starts or ends due to the large number of colors (Figures 5(a) and
5(c)). The proposed classification process provides an estimated number of pixels (area) for
every temperatura class. Every time a pixel is classified as belonging to a class, the counter
associated with that class is increased. Since only the foot area is relevant for the analysis, the
remaining temperatures are considered background and they are not taken into account for
the measurement. Thus, the clases 1 to 7 comprise the total foot area (total). In this way,
the foot is off the background, as it can be observed in Figures 5(b) and 5(d), where the
background is uniform and the regions are well defined in their corresponding classes. In this
case, the entire feet were processed by way of example and to provide a better perspective of
the classification process. For practical purposes, in this methodology the pixel classification is
as follows. Once the feet are separated in their corresponding images (Figures 5(a) and 5(c)),
they are again divided into four subsections, one per angiosome (Figure 6). Observing the
original MPA angiosome image of the left foot (Figure 6(a)), it can be supposed that the color
corresponding to the 5 class has the greatest area. However, it is difficult to establish the
limit among the colored regions in a clear way because of the similitude among some colors. A
visual comparison between the MPA angiosome of both feet could be even more complicated;
therefore, it is important to go beyond the visual perception and make a quantitative
comparison of the temperature distribution. Once the pixel classification has been made, it is
possible to clearly observe the area that covers certain temperature associated with a class.

The areas of each angiosome in Figure 6(b) are shown in the plot of Figure 7. The areas are
translated into percentages and based on this data, it determined which classes have the
larger area in the same angiosome of both feet. In the MPA angiosome analysis, it is observed
that the majority of the angiosome area is divided among classes 4, 5, and 6 for the left
foot (Figure 7(a)), that is, a temperature interval of [31, 34)C according to Table 1. For the
right foot, the class 4 is the predominant one with an interval of [31, 32)C. In the LPA plot
(Figure 7(b)), it is observed that the foot temperaturas aremore distributed, being 5 and 6
the largest clases for the left foot, while 4 is the largest for the right foot.Then, the
temperature intervals in the LPA angiosome are [32, 34) and [31, 32)C for the left and right
foot, respectively. For the calcaneal angiosomes (MCA and LCA), the temperatura distribution
has a more drastic variation respect toMPA and LPA. The MCA angiosome temperature in the
left foot with a predominant class 3 is lower ([30, 31)C) than in the right foot with a
predominant class 6 ([33, 34)C), as shown in Figure 7(c). LCA temperature is also very
different, being 2 ([29, 30)C) in the left foot and, 4 ([31, 32)C) the right foot, the
predominant classes (Figure 7(d)).

The importance of including the data of the adjacent classes is because they have values closer
to the class with larger area and they allow a better estimation of the temperatura by (3). For
the MPA in the left foot, = 6 and the adjacent classes are 5 and 7 with classmarks
33.5, 32.5 and 34.5 and areas 28, 27 and 10 (percentages), respectively.

Thus, the estimated temperature in this region is ETleftMPA = 33.2C. Following the same
procedure, theMPA for the right foot with = 4 has an ETrightMPA = 31.8C. With these
data, it is possible to estimate by (4) that between the MPAs regions there is an ETDMPA =
1.4C. Table 2 summarizes the described estimation for all the angiosomes of both feet by (3).
In Figure 6(a), it is noted that the biggest visual difference is in the MCAs and LCAs angiosomes,
and after computing their data it is known that these regions have a difference of temperature
that exceeds the normal difference of 2C. In addition, if the mean temperature of the whole
feet (ET) is estimated based on the temperatures of each angiosome, then both feet have a
temperature of 31.6C and 32.1C, respectively.The difference between this data suggests a
small difference in the temperature condition of the feet.
However, after a deeper analysis, it can be observed that there are important differences in
some regions that may not be distinguished in a qualitative analysis.

Another example is presented in Figure 8, where it is noticed how the pixel classification could
be useful in the improvement of the temperature regions limitation.

Figure 8(a) presents a wide red region in which it is difficult to define the limits of a color due
to smooth transitions. Some regions even seem to be homogeneous at first sight. However,
after the classification process (Figure 8), the regions identification is clearer and their analysis
is easier; it is noted the presence of small hot spots. Table 3 summarizes the information
obtained fromthe temperature classes and areas.

As it is observed, the most representative temperature clases are similar in some angiosomes
in which ETD has lower values. According to the ETD values, there are no abnormal differences
between feet; nevertheless, it is important to analyze the temperature differences inside an
angiosome with the aim of detecting an additional type of abnormalities: hot spots. Although
ETD brings important information, it does not detect abnormalities when the temperatures in
the feet are similar because it is based on the values of larger classes.

As mentioned before, a hot spot is an area whose temperatura is significantly higher respect to
its adjacent areas and the HSE value tries to detect it in an initial phase. Hot spots detection
provides information that help to detect suspicious areas where an ulceration process could be
starting.

In order to detect hot spots, the HSE value is calculated by (6). First, it is necessary to
determine the class associated with higher temperature present in an angiosome ( ). For
instance in theMPAangiosome of Figure 8, corresponds to 6 for the right foot then,HSE
value is 2.0C right foot; once again, the differences inmore than 2Cdegrees are suspicious.

In this angiosome, a small hot spot near the thumb bottom as well as a wider area belonging to
4 can observed. The HSE value suggests that there are abnormal areas but it is the expert
who should decide which ones must be under observation. This is because an angiosome only
shows a segment of the plantar area and one region detected as a hot spot may be part of a
larger region and not a hot area in an initial phase.This is the case for the hot spots detected in
the right MPA. For the left MPA with = 4, the HSE value is 0.2C. In the analysis of the LPA
angiosome (which include the MPA region mentioned above) the HSE does not show abnormal
differences in the right foot. For the left LPA the detector indicates a difference of 2.3C which
is associated with three hot spots near each other. In this case, the hot spots neither belong to
a larger region nor appear in the respective MPA. Another hot spot was found in the left MCA
where

= 6 and the HSE = 2.1 C. Although the angiosome differences (ETD) do not indicate
abnormal temperatures, theHSEestimator indicates the presence of possible hot spots inthe
rightMPA, and in the leftLPAandMCA.The estimator could be suggesting that these angiosomes
contain suspicious areas which should be under a continuous revision.

It is observed that the mean temperatures of the feet in Figure 5 (31.6C and 32.1C) and
Figure 8 (32.2C and 31.3C) exceed the normal plantar temperature (26.8 1.8C) and both
are in the characteristic range classified as neuropathy (32.2 0.94C). It can be noticed by
observing the temperatures of the angiosomes that most of them are also in this last range.
Both thermographs correspond to patients diagnosed with neuropathy as mentioned before.

4. Discussion

While there is an important amount of scientific reports about temperature analysis in the
diabetic foot, most of them rely on qualitative analysis. However, it is not always simple to
estimate the abnormal temperature differences by visual inspection of the thermogram. The
aim of the proposed methodology is to bring precise information about such differences by
facilitating the detection of posible risks regions and their evolution to the medical specialist.

Although not all the regions with abnormal temperature will become an ulcer, their monitoring
is very important as they represent high-risk regions. It is significant to remark that this
methodology is not a diagnostic tool but a tool that brings complementary information to be
evaluated by the medical specialist in order to facilitate early detection of ulceration risks.

5. Conclusions

Thermal imaging and image analysis are useful tools in the medicine field applied to the study
of diseases such as diabetes.

Temperature distribution in the plantar area contains relevant information about the condition
of diabetic foot and ulceration risks. In this work, a methodology was presented aiming at
providing quantitative information about abnormal temperature differences in symmetric
regions between feet and inside of the same foot. The methodology took into account the
temperature differences, their distribution, and area. As a first analysis, differences between
symmetric reas in both feet were studied since it is known that symmetric regions in the body
have similar temperatures. For this, the plantar area was divided into fourmain regions
(angiosomes) and the temperatures inside those regions were grouped in classes according to
a color similitude criterion. An index (ET) based on the relation between the class with larger
rea and its adjacent classes was proposed in order to estimate a representative temperature
for each angiosome.Thus, it was possible to obtain an estimated difference (ETD) between
symmetric regions which obtained an accurate measure to determine whether there is an
abnormal difference or not.

A second analysis was performed to study the temperaturas inside the angiosomes with the
aim of detecting the presence of small abnormal areas (hot spots). For that reason, it proposed
a hot spot estimator (HSE) that relates the representative emperature of the angiosome with
the higher

temperature on it. This estimator was capable to detect the presence of abnormal regions in
initial phase that, for their small area, were not detected by the estimator ETD. In this way, it
was possible to analyze the whole plantar area by bringing quantitative information to
determine the presence of regions in possible risk of ulceration. The results of the temperature
measurement agreed with previous reports about the diabetic foot temperature
characterization. Thus, this study provided an approach to bring reliable information to help
the specialist in the early detection of the ulceration risks of the diabetic foot associated with
high temperatures.

Skin Temperature Monitoring Reduces the Risk for Diabetic Foot Ulceration in High-risk
Patients

Purpose

To evaluate the effectiveness of home temperature monitoring to reduce the incidence of foot
ulcers in high-risk patients with diabetes.

Methods

In this physician-blinded, 18-month randomized controlled trial, 225 subjects with diabetes at
high risk for ulceration were assigned to standard therapy (Standard Therapy Group) or dermal
thermometry (Dermal Thermometry Group) groups. Both groups received therapeutic
footwear, diabetic foot education, regular foot care, and performed a structured foot
inspection daily. Dermal Thermometry Group subjects used an infrared skin thermometer to
measure temperatures on 6 foot sites twice daily. Temperature differences >4F between left
and right corresponding sites triggered patients to contact the study nurse and reduce activity
until temperatures normalized.

Results

A total of 8.4% (n = 19) subjects ulcerated over the study period. Subjects were one third as
likely to ulcerate in the Dermal Thermometry Group compared with the Standard Therapy
Group (12.2% vs 4.7%, odds ratio 3.0, 95% confidence interval, 1.0 to 8.5, P = .038).
Proportional hazards regression analysis suggested that thermometry intervention was
associated with a significantly longer time to ulceration ( P = .04), adjusted for elevated foot
ulcer classification (International Working Group Risk Factor 3), age, and minority status.
Patients that ulcerated had a temperature difference that was 4.8 times greater at the site of
ulceration in the week before ulceration than did a random 7 consecutive-day sample of 50
other subjects that did not ulcerate (3.50 1.0 vs 0.74 0.05, P = .001).

Conclusions

High temperature gradients between feet may predict the onset of neuropathic ulceration and
self-monitoring may reduce the risk of ulceration.

Discussion

There are few effective therapies to help high-risk patients with diabetes prevent foot
ulcerations. Under the best circumstances, high-risk patients will receive a few hours of
education a year concerning the complications of diabetes, regular foot evaluation by their
primary care physician or specialty care by a podiatrist, and protective shoes and insoles.
Unfortunately, standard prevention therapies usually are not provided. And, even when
patients are treated in centers of excellence with aggressive medical and surgical intervention,
between one third and two thirds of subjects with a prior history of diabetic foot ulcers will
reulcerate in 12-24 months. 2 16 18 Lavery and coworkers have reported 2 randomized clinical
trials using a temperature home monitoring device with similar results to this study. 13 19
This study substantively augments those previous studies in a larger, potentially higher-risk
population and is the first, to our knowledge, to suggest that thermometry used by patients
themselves can identify areas of inflammation before skin breakdown.

Self-monitoring is necessary to identify early warning signs to reduce the incidence of diabetic
foot complications and the associated decrements in quality of life and high resource costs.
Sadly, the ability of the most motivated patients with diabetes, their family members, and
even health care professionals to identify early warning signs is limited. The precursors to
ulceration are subtle, especially in a neuropathic foot. Self-evaluation of temperature seems to
offer a mechanism to identify an early sign of injury, when there is still time to avert a wound.
The results of this study suggest that a simple, inexpensive temperature device can be used
effectively by high-risk patients to reduce foot ulceration. This offers a significant advantage
over traditional prevention practices and therapies.

The concept of measuring skin temperature as a marker for tissue inflammation and injury in
the insensate foot has been addressed by several authors. As early as 1971, Goller et al
reported an association between increased local temperature and localized pressure leading to
tissue injury. 7 Sandrow and coworkers subsequently used thermometry as a tool to diagnosis
occult neuropathic fractures in patients with diabetes in 1972. 20 Stess and colleagues
described the use of infrared thermography to assess skin temperatures in patients with
diabetes, patients with diabetes with neuropathic fractures, patients with diabetes with ulcers,
patients with leprosy, and controls. 21 They found that neuropathic foot ulcers frequently
had increased skin temperatures surrounding a central necrotic area, and suggested that
temperature assessment may be a useful technique to identify patients at risk for ulceration.
Similar findings were revealed in a larger study by Armstrong et al using standard
thermometers. 22 The concept is elegant in its simplicity, and even challenging patient
populations appear to have been able to successfully use the device.

Veterans treated in the Department of Veterans Affairs are older, poorer, and have more
comorbidities than general community dwellers. However, they are not appreciably different
from the disproportionately low income and minority population with diabetes in regards to
ulcer outcomes. 23 24 Therefore, the results of this study should be applicable to both
veteran and nonveteran populations alike. In fact, a less challenged population might realize
better preventative outcomes.

In this study we evaluated 2 risk groups: patients with a history of previous diabetic foot
ulceration or amputation (Risk Group 3) and those with sensory neuropathy and loss of
protective sensation with structural foot deformities (Risk Group 2). The majority of subjects
enrolled were in the latter group, and this group had a lower incidence of ulceration than
expected. Our sample size estimate was based on published outcomes using standard
prevention in diabetics with a history of pathology (Risk Group 3). Therefore, we
underestimated the sample size needed to evaluate Diabetic Foot Risk Group 2 subjects, and
we disproportionably enrolled fewer Diabetic Foot Risk Group 3 subjects. Despite these errors,
the smaller patient sample was sufficient to show a significant impact when home
temperature monitoring was used.

Quality of life, functional status, self-efficacy, satisfaction with care, and cost outcomes from
this study will be addressed in future articles. While smaller studies in this area have mirrored
the current project, 13 19 it is unknown if the same outcomes and compliance would be
observed in a multicenter clinical trial over an extended period of evaluation. We look forward
to work that might continue to refine this methods usefulness in a variety of settings.