Infrared Thermography and Vascular Disorders in Diabetic Feet

Abstract

Aim:

Diabetes mellitus (DM) and related foot complications constitute a growing healthcare burden. Diabetes mellitus is associated with lower-limb amputation, but diabetic foot assessment is challenging. Here, we evaluated a novel noninvasive diagnostic method—infrared thermography (IRT) —assessing its diagnostic potential compared to conventional noninvasive measurements.

Methods:

This study included patients with DM (n = 118) and healthy controls (n = 93). All participants underwent ankle brachial index and toe pressure (TP) measurements, and IRT using a standardized protocol with temperature measurement at five foot areas.

Results:

Compared to controls, patients with DM generally had warmer feet and exhibited a significantly greater temperature difference between feet (P < .001). Mean temperatures were highest in patients with DM with neuroischemia, followed by neuropathy. Patients with DM with angiopathy showed the lowest mean temperature—similar to controls and noncomplicated diabetics. Mean temperatures at all measurement sites were significantly higher with abnormal TP (<50 mmHg) than normal TP (≥50 mmHg) (P < .001). Infrared thermography revealed differences between angiosome areas, subclinical infections, and plantar high-pressure areas.

Conclusion:

Infrared thermography revealed local temperature differences in high-risk diabetic feet. Normal skin surface temperature varies between individuals, but in combination with other tools, IRT might be useful in clinical screening.

ClinicalTrials ID:

14212016

Introduction

Diabetes mellitus (DM) and its complications constitute a growing healthcare burden. According to the International Diabetes Federation (IDF), the worldwide prevalence of diabetes was around 8.8% in 2017, meaning that 425 million people between the ages of 20 and 79 years were living with the disease. Moreover, diabetes accounts for up to 12% of global health expenditure.¹

Although greater attention is focused on the eye, renal, and cardiovascular complications of DM, foot problems are the leading cause of hospitalization among patients with diabetes. Diabetes mellitus is associated with an increased risk of lower-limb amputation. While amputation rates have declined in the general population, rates among patients with DM remain unchanged and are over 20 times higher than in the nondiabetic population.^2,3 Diabetes mellitus is strongly associated with peripheral arterial disease (PAD), with a PAD rate of 11% among people with diabetes compared to 4% among those without diabetes,^4,5 and PAD is strongly related to the risk of major amputation.⁶ In the Seattle diabetic foot study (2018), arterial disease and neuropathy emerged as the limb-specific risk factors for amputation.⁷ Importantly, the five-year survival rate after major limb amputation is only 42%,^8,9 and major amputations are associated with significant disability and impaired quality of life.¹⁰

There are many challenges associated with investigating the diabetic foot. A substantial proportion of patients with DM show symmetrical distal sensory neuropathy, which may mask symptoms of intermittent claudication and ischemic rest pain. The ankle brachial index (ABI) and the measurement of toe pressure (TP) are relevant indicators of PAD, but they have certain limitations. The high prevalence of medial sclerosis, which can render arteries incompressible upon cuff inflation, may lead to falsely elevated ABI values in approximately one-third of patients with DM.^5,11,12 Since digital arteries are less frequently calcified, TP may provide a more accurate estimation of circulation; however, TP cannot be measured in patients missing a toe or part of a foot. Duplex ultrasound and tissue oxygen pressure measurement are the other noninvasive methods currently used to determine disease severity and the affected vasculature, but these examinations are time-consuming and can only practically be applied in a limited group of patients. Moreover, these methods do not account for the angiosomes or achieve the accuracy of invasive angiographic imaging.

Infrared thermography (IRT) is a novel potential diagnostic method that does not require physical contact. It is a noninvasive, safe, and reliable technique that allows quick evaluation of radiating energy related to skin temperature.^13-16 A healthy person has a nearly constant human core temperature (within ±0.6°C), which can be used for routine physiological monitoring. However, skin temperature can change substantially depending on the need and extent of thermoregulation. Vasodilatation and vasoconstriction of skin blood vessels can effectively regulate temperature, because skin acts as both an effective heat radiator and an insulator, depending on the amount of blood flowing through the organ.¹⁷ The most common factors known to affect skin temperature and indicate abnormalities are inflammation in the subcutaneous/underlying tissues and changes in blood flow.¹³ The literature includes multiple suggestions that skin temperature variations may be useful for identifying pathological situations in the diabetic foot.^18-20 Infrared thermography is being increasingly appreciated as a modality that may contribute to early detection of incipient tissue damage predisposing to diabetic foot ulceration.^21,22 Understanding the temperature variations that occur in the diabetic foot improves the potential of IRT for detecting foot complications.²³

Foot screening in patients with diabetes is important for preventing amputations, but the diagnostic potential of IRT remains unknown. In this study, we created a simple clinical setting and used it to investigate the potential use of IRT as a screening tool for vascular disorders in a diabetic foot or as a rapid diagnostic tool in combination with the other noninvasive methods.

Methods

Subjects

Study participants were recruited by a vascular surgeon at the outpatient clinic of Oulu University Hospital between November 2016 and December 2017. The investigated population was heterogeneous group of patients who were prediscrabed to vascular surgeons examination by a clinician because of suspicion of circulatory disorders. Peripheral vascular status was evaluated during a routine clinical investigation that included medical history, comorbidities, symptoms, and signs. Bilateral brachial blood pressures, ABI, and TP were measured in all patients. Examination also included routine duplex ultrasound flow measurement. The extent of neuropathy was checked mostly at the diabetic clinic. When clinical examination indicated a circulatory disorder, patients underwent appropriate investigation and treatment. This study protocol was approved by the local Ethics Committee.

Participants were divided into two groups. Individuals with normal circulatory findings and with no earlier diagnosis of DM or PAD were assigned to the healthy control group (n = 98). All patients with DM, with or without peripheral circulatory disturbance, were assigned to the DM group (n = 118). Patients with DM were divided into three subgroups based on ABI classification: normal (ABI 0.9-1.4), mild (ABI 0.5-0.89), and severe (ABI < 0.5). Patients with DM were stratified by TP analysis using a cutoff value of 50 mmHg. For the analysis of angiosomes, we prospectively collected data regarding atherosclerotic lesions in lower-limb arteries by magnetic resonance angiography (MRA) or digital subtraction angiography (DSA). Patients were excluded from the study if they exhibited venous insufficiency requiring treatment, an active infection, or arthritis. Patients with Raynaud’s phenomenon (n = 5) were excluded from the healthy control group.

Thermographic Analysis

Infrared thermography results can be influenced by several environmental, individual, and technical factors that affect the human skin.^17,24 In this study, we prepared the vascular measurements according to the existing protocol at the outpatient clinic with constant room temperature (21°C-24°C). Before IRT, the patients were placed in a supine position, without shoes or socks. This position was maintained for 15 to 20 minutes during the routine vascular examinations. The camera was set about one meter from the foot. Infrared thermography was performed on both the plantar and dorsal sides of the foot at five circular sites with a nine-pixel diameter (Figure 1). The measurement sites roughly represented the angiosomal areas. All IRT recordings were acquired using a digital IR camera (FLIR A325sc) with a spatial resolution of 320 × 240 pixels and thermal resolution of 0.05°C. The IRT data were analyzed using the Thermidas Imager and FLIR ResearchIR programs.

Figure 1.

Examples of infrared thermography in healthy and disease conditions. The “rainbow” palette indicates cold (blue) to hot (red) temperatures. (a) The five areas measured are shown (circles) on the (left) plantar and (right) dorsal sides of foot (patient in the healthy control group). (b) Patient with diabetes with bilateral neuropathy and mottled colorations, showing critically ischemia in the first toe on the right side (white arrows). (c) Patient with diabetes with acute inflammation of Charcot neuroarthropathy on the right side (black arrows). (d) Patient with diabetes with bilateral peripheral angiopathy, which is worse on the left side (red arrows) (ankle brachial index: 0.93 right/0.79 left; toe pressure: 89 mmHg right/28 mmHg left).

Statistical Analyses

Continuous variables are expressed as mean ± SD, unless otherwise stated. Simple between-group comparisons were made using the t-test or Wilcoxon’s test for continuous variables, and using Pearson’s chi-squared test for categorical variables. The linear mixed model (LMM) was used for subgroups of ABI analysis. In LMM, the patients were set as random effects and the covariance pattern was chosen according to Akaike’s information criteria. Two-tailed P-values are reported. Statistical analyses were performed using SPSS 22.0.

Results

The DM group included 13 patients (11%) with type 1 diabetes and 105 (89%) with type 2 diabetes. In general, patients in the DM group were older and had more comorbidities compared to the healthy control group. Table 1 presents the basic demographics of the study patients. In the healthy control group, 11 participants had undergone the previous vascular reconstruction for abdominal aneurysms or vascular injuries, and one had undergone a previous minor amputation due to infection. In the DM group, five patients had previously undergone major amputations, and 12 had undergone minor amputations due to atherosclerosis, infections, or both.

Table 1.

Baseline Characteristics of Patients Without Diabetes or Peripheral Arterial Disease (Healthy Controls) and Those With Diabetes.

Characteristic	Healthy control group n = 93	Diabetes group n = 118	P value
Age in years, mean (SD)	61 (18)	74 (10)	<.001
Men, n (%)	55 (59)	81 (69)
BMI in kg/m², mean (SD)	28 (6)	29 (5)
Ever smoked, n (%)	11 (12)	37 (31)
Hypertension, n (%)	36 (39)	84 (71)	<.001
Ischemic heart disease, n (%)	23 (25)	74 (63)	<.001
Cerebrovascular disease, n (%)	10 (10)	30 (25)	.007
eGFR in mL/min, mean (SD)	63 (13)	54 (17)	.001
Normal/mild eGFR (≥59), n (%)	51 (83)	75 (64)
Moderate eGFR (30-58), n (%)	10 (16)	30 (25)
Severe eGRF (<29), n (%) Dialysis	1 (1) 0 (0)	13 (11) 5 (4)	.04
Pulmonary disease, n (%)	16 (17)	26 (22)
Atrial fibrillation, n (%)	9 (9)	38 (32)	<.001
Coagulopathy, n (%)	1 (1)	3 (3)
Previous vascular reconstruction, n (%)	11 (11)	63 (53)	<.001
Previous amputation, n (%)	1 (1)	17 (14)	.001
Wound, n (%)	4 (5)	61 (52)	<.001
Pain at rest, n (%)	5 (6)	26 (22)	.001
Claudication, n (%)	22 (24)	52 (44)	<.001
Pain at <50 m, n (%)	6 (7)	21 (18)
Pain at >50 m, n (%)	16 (19)	31 (27)

Abbreviations: BMI, body mass index; eGFR, estimated glomerular filtration rate.

For every measurement, the temperatures spanned a greater range in the DM group than in the healthy control group, with a range of 21.5°C to 38.6°C in the DM group and of 22.5°C to 34.3°C in the control group (Figure 2). On the plantar side, the mean temperatures were significantly higher in the DM group than in the control group (P < .001). Considering both feet as one bloc, the mean temperatures on the plantar and dorsal sides, respectively, were 27.7°C ± 2.0°C and 29.9°C ± 1.7°C among controls, and 28.7°C ± 2.5°C (P < .001) and 29.7°C ± 1.8°C (P = ns) in the DM group.

Figure 2.

Mean temperatures measured in both feet at five sites in healthy control patients (blue) and patients with diabetes (red). Boxplots show the mean (colored squares), median (horizontal line), interquartile range (size of box), minimum and maximum values (whiskers), and outliers (diamonds).

Dist: distal; Prox: proximal; Lat: lateral side; Med: medial side, *P < .001

Side-to-side comparisons of temperatures revealed significant differences between feet (P < .001) at all measurement sites (Table 2). Analysis of both feet as one bloc revealed significantly greater differences between feet in the DM group compared to controls, on both the plantar sides (control: 0.53°C ± 0.5°C vs DM: 1.46°C ± 1.4°C, P < .001) and the dorsal sides (control: 0.53°C ± 0.4°C vs DM: 1.30°C ± 1.3°C, P < .001).

Table 2.

Absolute Values of the Between-Foot Temperature Differences in Healthy Control Patients and Patients With Diabetes.

	Site of measurement	Healthy control group n = 93	Diabetes group n = 110	P value
		Difference between feet in °C mean (SD)	Difference between feet in °C mean (SD)
Plantar	Distal site on lateral side	0.73 (0.6)	1.77 (1.7)	<.001
	Distal site on medial side	0.70 (0.6)	1.79 (1.8)	<.001
	Middle	0.59 (0.4)	1.13 (1.1)	<.001
	Proximal site on lateral side	0.79 (0.7)	1.53 (1.5)	<.001
	Proximal site on medial side	0.83 (0.7)	1.70 (1.6)	<.001
Dorsal	Distal site on lateral side	0.67 (0.5)	1.56 (1.6)	<.001
	Distal site on medial side	0.73 (0.7)	1.76 (1.7)	<.001
	Middle	0.82 (0.8)	1.37 (1.4)	<.001
	Proximal site on lateral side	0.73 (0.6)	1.30 (1.2)	<.001
	Proximal site on medial side	0.64 (0.5)	1.20 (1.1)	<.001

Infrared thermography was an effective means of detecting local temperature differences (Figure 1). In general, feet seemed to be warmer in the DM group than in healthy controls. Patients with DM with neuropathy (27%) had a higher mean temperature compared to the healthy controls (plantar side: 29.4°C ± 2.3°C vs 27.7°C ± 2.0°C, P = .001; dorsal side: 30.1°C ± 1.7°C vs 29.9°C ± 1.7°C, P = ns). Patients with DM with neuroischemic feet (33%) showed an even higher mean temperature (plantar side: 29.7°C ± 2.4°C; dorsal side: 30.3°C ± 1.7°C), which was significantly higher compared to the healthy controls on the plantar side (P < .001) (dorsal side: P = ns). The side-to-side difference between feet was highest in the feet with neuropathy (plantar side: 1.9°C ± 1.4°C; dorsal side: 1.6°C ± 1.7°C) and in neuroischemic feet (plantar side: 1.9°C ± 1.7°C; dorsal side: 1.5°C ± 1.3°C), and was lowest in the feet of patients with angiopathy (18%) (plantar side: 0.8°C ± 0.6°C; dorsal side: 0.9°C ± 0.8°C). Patients with angiopathy had lower mean temperature compared to the healthy controls (plantar side: 26.9°C ± 2.0°C vs 27.7°C ± 2.0°C, P = ns; dorsal side: 28.3°C ± 1.6°C vs 29.9°C ± 1.7°C, P = .001) and compared to patients with the noncomplicated diabetes without angiopathy or neuropathy (22%) (plantar side: 26.9°C ± 2.0°C vs 27.4°C ± 2.2°C, P < .001; dorsal side: 28.3°C ± 1.6°C vs 28.9°C ± 1.8°C, P < .001).

To perform a more detailed analysis of patients in the DM group, we divided the DM group into three subgroups based on ABI classification: normal (n = 56 ft, 24%), mild (n = 93 ft, 40%), and severe (n = 26 ft, 11%). In 56 cases (24%), the ABI values were unreliable due to mediasclerosis or because the measurement was below the detection limit in one or both legs. We found no statistical differences between any sites of measurements. For a separate analysis, we subdivided patients with DM into two subgroups based on TP: normal/mild (TP ≥ 50 mmHg, n = 130 ft, 56%) or severe (TP < 50 mmHg, n = 80 ft, 35%). Data were missing from this analysis, mainly due to missing toes or to immeasurable TP (n = 26 ft, 11%). The mean temperature was higher in feet with clearly abnormal TP (<50 mmHg) compared to in feet with TP over 50 mmHg, with significant differences at all measurement sites (Table 3).

Table 3.

Mean Temperature (°C) of the Plantar and Dorsal Sides of the Foot in Patients with Diabetes Stratified by Toe Pressure.

	Site of measurement	Diabetes subgroups		P value
		Normal/mild, TP ≥ 50 mmHg n = 130 ft mean (SD)	Severe, TP < 50 mmHg n = 80 ft mean (SD)
Plantar	Distal site on lateral side	27.3 (2.7)	29.3 (2.9)	<.001
	Distal site on medial side.	27.6 (2.8)	29.2 (2.8)	<.001
	Middle	29.2 (1.9)	30.2 (1.9)	<.001
	Proximal site on lateral side	28.0 (2.3)	29.7 (2.5)	<.001
	Proximal site on medial side	28.2 (2.5)	30.0 (2.6)	<.001
Dorsal	Distal site on lateral side	28.6 (2.1)	29.7 (2.3)	.001
	Distal site on medial side	28.1 (2.5)	29.3 (2.3)	.001
	Middle	30.3 (1.8)	30.9 (1.6)	.05
	Proximal site on lateral side	30.1 (1.6)	30.7 (1.6)	.008
	Proximal site on medial side	29.8 (1.6)	30.5 (1.8)	.001

Abbreviation: TP, toe pressure.

Magnetic resonance angiography and/or DSA images of the lower-limb arteries were available from 67 patients (120 ft) in the DM group. For some patients, only DSA for one leg was available. Patients having an open anterior tibial artery along with stenosis or occlusion of the other crural arteries (dorsalis pedis angiosome) (n = 29) exhibited a higher temperature on the dorsal side of feet (30.2°C ± 1.9°C) compared to patients without any stenotic or occluded crural arteries (n = 39, 29.1°C ± 1.7°C, P = .02). Patients with an open posterior tibial artery (plantar angiosomes, n = 27) exhibited a higher temperature on the plantar side of feet (open plantar angiosomes: 28.3°C ± 2.1°C vs open crural arteries: 27.4°C ± 2.2°C, P = .11). In 17 patients, only the peroneal artery was open (peroneal angiosomes), but the IRT was not targeted to this area. There were eight patients showing occlusion of all three crural arteries.

Discussion

The present IRT study yielded three major findings. First, the side-to-side differences between feet were greater in the DM group than among healthy controls, where the temperatures were more constant between feet. Second, mean temperatures significantly differed between the subgroups with patients with noncomplicated diabetes and DM patients with angiopathy, neuropathy, and neuroischemia. Notably, neuropathic feet were warmest and neuroischemia increased the temperature in diabetic feet. Third, IRT was able to reveal differences between angiosome areas.

In this investigation, we aimed to clarify the diagnostic potential of IRT for vascular disorders in a diabetic foot. Infrared thermography effectively revealed differences in temperature between different skin areas. Skin temperature normally shows symmetrical bilateral distribution; thus, strong asymmetry indicates an abnormality.^25-27 Thermography has been used to monitor healing of Charcot’s arthropathy in diabetic feet^28,29 and IRT can characterize temperature gradients in regions affected with vascular disorders. Skin temperature variations of over 2.0°C may be useful for identifying pathological situations in the diabetic foot.^18,30,31 Our present results showed that the mean side-to-side difference between feet was higher in the DM group (range, 0.0°C-10.2°C) than in the healthy control group, although the healthy control group still showed a wide range (0.0°C-3.1°C). Our results support the previous findings, with some apparent congruency regarding abnormal variations in skin temperature. These data may be helpful, for example, in creating IRT software solutions for automatic recognition of skin temperature abnormalities. However, there remains a need for further investigations of the variation limits and tendencies in skin temperature maps, and how they correspond with patient symptoms, conditions, and disease stages.

Important risk factors for amputation include neuropathy (sensory, motor, or autonomic) and arterial disease with or without foot ulcer.^7,32,33 In our present study, the presence of neuropathy raised foot temperature, and neuroischemic and ischemic feet were also warmer than healthy feet. These findings can be explained by the thermoregulatory mechanisms of the feet. Local ischemia may disrupt sympathetically mediated noradrenergic vasoconstriction, promoting increased flow to the cutaneous vessels rather than through the deeper nutritive vessels. Diabetes mellitus is associated with several functional abnormalities of the microvasculature, including increased arteriovenous shunting. Arteriovenous anastomoses are low-resistance conduits that allow high flow rates directly from arterioles to venules, and that are richly innervated by sympathetic vasoconstrictor nerves. These changes reportedly lead to capillary hypoperfusion, with likely impairment of wound healing.^34-36

In a large prospective longitudinal study, Armstrong et al report that baseline measurement of skin temperatures was not effective as a one-time screening tool for future events among patients with diabetes.³⁷ However, if neuropathy indicates a high risk and raises the temperature of a diabetic foot, this information should be of some clinical value. An IRT image provides information about broad regions of interest rather than only the local temperature of a small area. Sometimes, IRT can even provide surprising information, as seen in Figure 1. Without IRT, it may not have been noticed that the examined foot had Charcot’s arthropathy due to the lack of symptoms or signs.

The presence of PAD is defined as a nonpalpable pulse and an ABI of less than 0.9, which are widely used criteria for the screening and diagnosis of PAD among patients with diabetes. However, the presence of pulses does not always correlate with the absence of ischemia, and the absence of foot pulses does not indicate the degree of the tissue perfusion deficit.^5,38 Moreover, medial sclerosis that renders arteries incompressible by cuff inflation may lead to falsely elevated values in approximately one-third of patients with diabetes.¹² Among patients with diabetes, there is a U-shaped relationship between ABI categories and cardiovascular death; patients with low (<0.9) and high (>1.4). ABI values are both at higher risk. Toe pressure or toe brachial index is the recommended measurement for PAD assessment in people suspected to have arterial calcification.^39,40 Toe pressure is a linear predictor of the risk of cardiovascular death, with low TP categories consistently predicting high risk. The TP provides more reliable information regarding small artery health in a clinical setting, and there seems to be a relationship between TP and the risk of foot complications, including ulceration and amputation.^41-44

In our present study, one-fourth of the patients with diabetes had an unmeasurable ABI, and 30 (54%) of those patients had critical ischemia according to TP. We were able to perform IRT in all patients. Infrared thermography imaging reveals tendencies that can be related to ABI and TP findings. In indefinite cases, IRT might sway the decision making toward a definite diagnosis and further invasive investigations. In high-risk patients, the early detection of subclinical infections, abrasions, or ulcers can be vital for saving the limb. Noninvasive IRT provides this kind of information.

The six angiosomes of the foot and ankle originate from the three main arteries to the foot and ankle. The plantar side of the foot receives artery flow mainly from the posterior tibial artery (the calcaneal, medial plantar, and lateral plantar arteries) and the dorsal side from the anterior tibial artery. The peroneal artery flows to the lateral side of the calf and calcaneus (the calcaneal branch and anterior perforating branches).^45,46 Nagase et al provide a detailed description of plantar thermographic patterns, showing wider variations in patients with diabetes than in normal subjects, which might represent a screening option for daily foot care and surgical intervention.⁴⁷ DM patients endovascular procedures should be targeted according to the angiosome concept, to achieve better wound healing and limb salvage in the treatment of critical limb ischemia.^48,49 Our study demonstrated that IRT could reveal angiosomal disorders in feet (Figure 3). In particular, distal angiopathy and distal PAD, which are common among patients with diabetes, could be evaluated at the outpatient clinic upon first contact before invasive investigations. This might enable straightforward approach in certain cases. In addition, IRT can clearly reveal local hotspots in patients with diabetes. In our clinic, IRT has been used to reveal subclinical infections and plantar high-pressure areas in cases where early detection is crucial. Our present results suggest that IRT may be clinically useful for estimating effective vascularization; however, a longer follow-up study is needed to improve our understanding of temperature variations.

Figure 3.

Infrared thermography and magnetic resonance angiography in the feet of a patient with diabetes according to angiosomes. The anterior tibial artery angiosome is warmer on the dorsal side of the foot, and the anterior tibial artery is open on the right foot (red arrows). The posterior tibial artery angiosome is warmer on the plantar side of foot, and the posterior tibial artery is open on the left foot (white arrows).

Conclusion

Due to the lack of treatments that target underlying nerve damage, prevention is a key component of diabetes care. Infrared thermography reveals local temperature differences, mottled coloration, and higher mean temperatures in high-risk diabetic feet with neuropathy and neuroischemia. Infrared thermography could also reveal angiosomal disorders in feet; in particular, distal angiopathy and distal PAD, which are common among patients with diabetes. Our present findings indicate that IRT may be useful as an additional method for screening diabetic feet at an early stage. Further studies are needed to elucidate the limitations and potential of IRT and to standardize the IRT findings and their relationships with certain conditions and disease stages.