Reference Standards for Digital Infrared Thermography Measuring Surface Temperature of the Upper Limbs

(1) Background: although digital infrared thermographic imaging (DITI) is used for diverse medical conditions of the upper limbs, no reference standards have been established. This study aims to establish reference standards by analyzing DITI results of the upper limbs. (2) Methods: we recruited 905 healthy Korean adults and conducted thermography on six regions (dorsal arm, ventral arm, lateral arm, medial arm, dorsal hand, and ventral hand region). We analyzed the data based on the proximity of regions of interest (ROIs), sex, and age. (3) Results: the average temperature (°C) and temperature discrepancy between the right and the left sides (ΔT) of each ROI varied significantly (p < 0.001), ranging from 28.45 ± 5.71 to 29.74 ± 5.14 and from 0.01 ± 0.49 to 0.15 ± 0.62, respectively. The temperature decreased towards the distal ROIs compared to proximal ROIs. The average temperatures of the same ROIs were significantly higher for men than women in all regions (p < 0.001). Across all regions, except the dorsal hand region, average temperatures tended to increase with age, particularly in individuals in their 30s and older (p < 0.001). (4) Conclusions: these data could be used as DITI reference standards to identify skin temperature abnormalities of the upper limbs. However, it is important to consider various confounding factors, and further research is required to validate the accuracy of our results under pathological conditions.


Introduction
Digital infrared thermographic imaging (DITI) has been utilized as an ancillary diagnostic method for various medical conditions related to the upper extremities. For instance, cervical radiculopathy; peripheral nerve entrapment syndrome (e.g., carpal tunnel syndrome); rheumatic disease (e.g., Raynaud's disease); complex regional pain syndrome; tendinopathy; hand arthritis (e.g., psoriatic arthritis); and skin cancer can be confirmed by DITI [1][2][3][4][5][6][7][8][9][10][11][12][13]. The hypo-radiant (hypothermia) or hyper-radiant (hyperthermia) regions can be identified by comparing the temperature between the right and left arms or with empirical normal ranges. However, the significant temperature differences between both sides and deviations from normal ranges have not been established. Furthermore, body surface temperatures are influenced by the measurement environment, ana-tomical area, and subject characteristics, which presents a challenge in deriving reference values from the existing literature [14,15].
Consequently, previous attempts to establish reference standards for DITI using systematic reviews based on meta-analysis or machine learning methods have been limited in providing detailed standard DITI values [10,[16][17][18]. Although a recent study suggested a correct differential diagnosis process for Raynaud's phenomenon in the hand using a deep convolutional neural network, a type of deep learning method, its small data set had limitations [19]. To address this gap and establish scientific reference standards for

Statistical Analysis
SPSS version 27.0 (IBM Corporation, Armonk, NY, USA) was used for statistical analysis. Data were checked for normality using the Kolmogorov-Smirnov test, and the results were expressed as mean ± standard deviation (SD), mean with 95% confidence intervals (CI), or median with range. One-way analysis of variance (ANOVA), paired t-test, or linear regression analysis were applied according to the purpose of analysis. A p value of less than 0.05 was considered statistically significant.

Statistical Analysis
SPSS version 27.0 (IBM Corporation, Armonk, NY, USA) was used for statistical analysis. Data were checked for normality using the Kolmogorov-Smirnov test, and the results were expressed as mean ± standard deviation (SD), mean with 95% confidence intervals (CI), or median with range. One-way analysis of variance (ANOVA), paired t-test, or linear regression analysis were applied according to the purpose of analysis. A p value of less than 0.05 was considered statistically significant.

Subgroup Analysis Based on Sex
The average temperatures of each same ROI were significantly higher in male subjects compared to female subjects across all regions and age groups (p < 0.001, paired ttest). The average temperatures for each region were as follows: 30
Bioengineering 2023, 10, x FOR PEER REVIEW 14 of 18 Figure 6. The average regional temperatures in men and women.

Subgroup Analysis Based on Age Group
The average temperatures of the ROIs significantly increased with age, particularly in individuals in their 30s and older in all regions, except for the dorsal hand region (p < 0.001, ANOVA). However, subjects in their 20s had a higher surface temperature than Figure 6. The average regional temperatures in men and women.

Subgroup Analysis Based on Age Group
The average temperatures of the ROIs significantly increased with age, particularly in individuals in their 30s and older in all regions, except for the dorsal hand region (p < 0.001, ANOVA). However, subjects in their 20s had a higher surface temperature than those in their 30s (p < 0.001, paired t-test). Consequently, surface temperatures were lowest for those in their 30s for all regions, except for the dorsal hand region (p< 0.001, ANOVA post hoc analysis) ( Table 7 and Figure 7). Figure 6. The average regional temperatures in men and women.

Subgroup Analysis Based on Age Group
The average temperatures of the ROIs significantly increased with age, particularly in individuals in their 30s and older in all regions, except for the dorsal hand region (p < 0.001, ANOVA). However, subjects in their 20s had a higher surface temperature than those in their 30s (p < 0.001, paired t-test). Consequently, surface temperatures were lowest for those in their 30s for all regions, except for the dorsal hand region (p< 0.001, ANOVA post hoc analysis) ( Table 7 and Figure 7).

Figure 7.
Relationship between the average regional temperatures and subject age.

Average Temperature and |ΔT| of each ROI
The average surface temperature of the upper limbs varied depending on regions and ROIs, ranging from 27.08 ± 5.14 °C to 31.27 ± 5.53 °C, which is consistent with the results found in our previous study on the lower limbs. However, the average temperatures of the upper limbs were higher than those of the lower limbs (range, 24.60 ± 5.06 °C- Figure 7. Relationship between the average regional temperatures and subject age.

Average Temperature and |∆T| of Each ROI
The average surface temperature of the upper limbs varied depending on regions and ROIs, ranging from 27.08 ± 5.14 • C to 31.27 ± 5.53 • C, which is consistent with the results found in our previous study on the lower limbs. However, the average temperatures of the upper limbs were higher than those of the lower limbs (range, 24.60 ± 5.06 • C-27.75 ± 5.76 • C) [20]. We believe this difference is due to the different distance from the heart [21][22][23].
The average of |∆T| of each ROI also varied by regions and ROIs, ranging from 0.00 to 0.52 • C. However, these values are smaller compared to the reference standards of the lower limbs, where |∆T| reached 0.76 • C [20]. Nevertheless, the practical value of |∆T| of the upper limbs was not within 0.1-0.3 • C, which is considered the normal range, based on the previous consensus [15,24,25].
These data for each ROI provide reference standards for DITI of the upper limbs. Clinically significant cold/hot areas, or specific areas with a significant difference between both sides, can be detected based on comparative analysis between the practical patient's image and these data. Additionally, the detailed ROIs in this study can be modified simply during the processing of DITI capture in actual clinical practice. However, accurate diagnosis requires a comparison of each ROI, not just the averages of whole regions, as normal ranges of surface temperature and ∆T values vary depending on region or ROI.
Various intrinsic factors, such as sex, age, fat percentage, and menstrual cycle stage, and extrinsic factors, such as test environment, testing time, and season, can influence the results [26][27][28][29]. As a result, actual DITI measurements may fall outside the reference standard values. Therefore, further clinical studies are necessary to validate these data by comparing the DITI results of patients with specific diseases with those of healthy subjects.

Subgroup Analysis Based on Proximity of ROIs, Sex, and Age Group
A significant correlation was found between the proximity of ROIs to the body core and surface temperature in all regions, except for the lateral arm region, i.e., surface temperatures decreased from the proximal regions to the distal ends. These findings are consistent with a previous suggestion that the surface temperatures of peripheral regions (e.g., hand or foot) tend to be cooler than central regions (e.g., trunk or proximal arm), due to the greater distances from the main thermal organs, such as the heart, large vessels, or viscera [20][21][22][23]. This tendency was found to be greater for the upper limbs than lower limbs, which can be attributed to the fact that the upper limbs are closer to the heart than the lower limbs, resulting in the proximal region being distinctly warmer than the distal region [20].
However, the |∆T| values showed no specific trend based on proximity of ROIs to the body core, which is different from a previous suggestion that the |∆T| values are higher in the distal regions compared to the proximal regions [2,15]. This was also found in the lower limbs in a previous study [20].
Surface temperatures were found to be higher for men in all regions, which is consistent with suggestions that surface temperatures are lower for women due to the insulating effect of thicker subcutaneous fat [30]. This trend was also observed for the lower limbs [20].
Surface temperatures increased significantly with age from the 30s, except for the dorsal hand region, which is also consistent with previous research on the lower extremities [20]. These trends may be caused by age-related diminished vasoconstriction by the sympathetic nervous system [31][32][33]. The paradoxical higher temperatures for those in their 20s can be explained by a higher basal metabolic rate and less subcutaneous fat [34,35].

Limitations and Significance
Establishing reference standards for DITI is a complex task due to the influence of various confounding factors on skin temperature and DITI measurements [36]. Therefore, it is important to acknowledge several limitations of our study. First, body fat percentage is a significant confounding factor in measuring body surface temperature using DITI [37] However, we did not collect data on body fat percentage or calculate the body mass index based on body weight and height. Secondly, daily biorhythms, such as menstrual cycle status, menopause, sleep patterns, and emotional stress, can also impact DITI results [38,39]. Unfortunately, we did not account for these factors during the examination. Thirdly, various extrinsic factors, including diurnal testing time, season, and ethnicity were not considered [39]. Moreover, even though we recruited healthy adult volunteers based on questionnaire responses, the process does not guarantee the exclusion of specific diseases or conditions that may have affected DITI results.
Nonetheless, the study holds value due to its large sample size and the use of a consistent protocol for measurements. It represents the first attempt to establish reference standards for DITI measurements of the upper limbs.

Conclusions
The findings of this study provide a basis for establishing reference standards for DITI measurements of surface temperatures in the upper extremities. These standards can aid physicians in making objective diagnoses by comparing patient DITI results with the provided data. However, it is crucial to consider the influence of various confounding factors in surface temperature measurements. Moreover, before confirming the results, it is important to consider several parameters, including the specific location of the ROIs, sex, and age.