Thermal element is always associated with a particular physiological state, such as fever, food, sleep, and muscle metabolism. The thermal elements of fever activate the innate and adaptive immune responses. ⁶ The effect of mild hyperthermia treatment on healthy adults enhances innate and adaptive immunity. ⁷ Therefore, to bolster the immune system, mimicking the condition of hyperthermia is achieved externally through heat ingestion from various sources, such as water circulation at 34 ^° C through the perfused suit ⁸ and exposure to a cold environment (ice water bath, cryotherapy), ⁵ which reduces arterial stiffness by increasing the metabolic rate innervated by sympathetic activation. Heat induction in the body involves several mechanisms. ⁹ The metabolic rate depends on brown adipose tissue (BAT) activity, which generates heat in the body. ¹⁰ Infrared thermal (IR) imaging of the skin has been used for several decades to assess spatial skin temperature distributions. ¹¹ The spatiotemporal distributions of temperature are heat generation (thermogenesis) measures as per Penne's bio-heat equation, including heat associated with both metabolic and tissue blood perfusion. ¹² The study of HCRT on spatial change in temperature gradient ( ∇ 2 T ) and slope of temperature change over time is crucial for investigating thermogenesis in general and specifically for the geriatric population.

HRV is widely used for the clinical assessment of the autonomic nervous system (ANS). ¹³ It has been widely used to study the effects of thermogenesis on heat stress. ¹⁴ HRV has various time, frequency, and nonlinear parameters for physiological assessment, ¹⁵ which are altered during thermogenesis. In the frequency domain, the very low-frequency (VLF) band of HRV (0.0039–0.04 Hz) responds to thermoregulation. ¹⁶ The VLF increases significantly in response to core cooling, peripheral vasoconstriction, and shivering, while both very low and low (0.04 - 0.15 Hz) frequency powers increase in response to skin-surface cooling. ¹⁷ However, when exposed to hot environmental conditions at 30 ^° C, the low to high frequency (LFHF) ratio increases. ¹⁸ HRV decreases owing to an increase in the LFHF ratio when the skin is exposed to both cold and hot environments beyond the thermal comfort range (22 – 27 ^° C). ¹⁹ The VLF power increased during core cooling. Thermogenesis by exercise and external hot environments shows a reduction in parasympathetic nervous system (PNS) tone with elevated sympathetic nervous system (SNS) tone. ²⁰ In contrast, the very high-frequency (VHF) band of HRV (0.4-0.9 Hz) responds to cardiovascular autoregulation. ²¹ Autoregulation is improved by thermogenesis for a 0.4 ^° C increase in internal temperature. ¹⁴ This study explored the unique mechanism of thermogenesis by HCRT and its effect on HRV. The heat depends on the rate of change of temperature, and ∇ 2 T . The former was captured from the core (axillary ²² ) temperature slope measured by the LM-35 sensor. ²³ The ∇ 2 T is measured from thermal images captured by IR-camera.

Thermogenesis through exposure to high heat stress conditions improves HRV in T2DM subjects ²⁴ and glucose metabolism. ²⁵ Exergy analysis assesses the efficiency of thermogenesis based on physical activity, ^{26 , 27} altitude change, ²⁸ thermal sensation, ²⁹ and respiration. ³⁰ According to the literature, performing moderate or intense exercise can lead to a decrease in exergy and increase in metabolic heat. In the current study, HCRT intervention was used to evaluate the exergy from the breath to measure the quality of thermogenesis.

Measurement: Exergy depends on the pressure, temperature, and velocity of the exhaust breath. The MEMS pressure sensor, composed of BMP-280, measures both the pressure ³¹ and temperature ³² of the breath. An anemometer was used to measure breath velocity. The standard measurement of dissolved glucose in the blood is the laboratory-estimated venous plasma glucose (VPG) test (glucose oxidase-peroxidase method). However, point-of-care measured capillary blood glucose (CBG) values with a glucometer (Accu-Chek) at a 2-h plasma glucose level can be used for gestational diabetes mellitus (GDM) assessment. ³³

The sympathetic division of the autonomic nervous system (ANS) contributes to the effect of the metabolic heat rate through active and passive heat acclimation. The metabolic rate through non-shivering thermogenesis includes many techniques, of which sitting while clasping hands is very natural. Its effects on thermogenesis and ANS are unknown.

We hypothesized that HCRT causes thermogenesis with a higher HRV. Although higher HRV could be attributed to resting alone, HRV further increased due to HCRT. To test this hypothesis, we measured the spatial gradient and temperature slope with respect to time using thermal images and axillary and breath temperatures before, after, and during HCRT intervention. We also examined the effect of HCRT on the CBG values of subjects with T2DM to validate our hypothesis. Our study shows that, in addition to temperature, the spatiotemporal distribution of temperature is an important parameter that affects HRV.

This study was divided into two parts. 1 ^st part aims to observe the changes in temperature over time and space with 15 healthy participants (height, 160 ± 10 cm; weight, 65 ± 10 kg), and 2 ^nd part focuses on the clinical validation with 12 type-II DM subjects (eight male, four female) for only the CBG test before and after the HCRT intervention. The inclusion criteria for healthy participants with no medications required them to be between the ages of 24 and 60. Female participants were excluded from the first part of the study, as we collected thermal images of their bare bodies to keep them dry and avoid any impact of clothing on heat transfer.

The protocol of this study involving humans was approved by the Institutional Ethical Committee (IEC) of IIT Madras (IEC/2023-03/MM/02/03) on 1 ^st December 2023 for a period from 11/12/2023 – to 10/12/2024. The study conforms to the standards set by the latest revision of the Declaration and was conducted to comply with local legislation and institutional requirements. We collected written consent forms from each participant according to ethical guidelines. The data collection from the participants is from 18 ^th January 2024 to 25 ^th January 2024.

The duration of the experiment was 50 min per subject. The duration consists of three phases: pre-hand-clasping normal (N1, 10 min), hand-clasping (HCRT, 20 min), and post-hand-clasping normal (N2, 20 min). The experiment followed a 10 min time window for the analysis: N1 (0–10 min), HCRT1 (10–20 min), HCRT2 (20–30 min), and N2 (35–45 min). The subjects were asked to breathe normally during the experiment. The study included various sensors to assess the effects of the HCRT intervention. The duration and mode (continuous and discrete) of recording by each sensor during the experiment are listed in Table 1. We collected ECG, transient body temperature, and thermal images of the participants. The ECG was collected by an ECG-Amplifier coupled with an Arduino UNO with a sampling frequency of 182 Hz. ³⁴ The transient body temperature was measured by the LM-35 sensor with a range of −50 ^° C to 150 ^° C and an accuracy of ±0.2 ^° C at 25 ^° C. The ambient temperature was maintained at 30 ^° C. LM-35 records the temperature under the armpit at 1.2 Hz. The sensor records the axillary temperature for 1.5 min as the reading saturates at 35.5 ^° C to 36.04 ^° C ³⁵ after reaching the steady state; it provides the rising transient response of temperature. LM-35 was then kept at ambient temperature for 1.5 minutes to allow it to reach the initial temperature. A KT-150 IR camera captures thermal images of dimension 120 pixels × 120 pixels, with a spectral range of 8–14 μm and a resolution of 0.1 ^° C. The emissivity ( ε) in KT-150 was set to 0.98 for all images resembling the emissivity of the skin. The exergy of the exhaust breath was calculated from its pressure, temperature, and velocity using BMP-280, and the anemometer was recorded for 60 s at the end of the experimental phases, recorded at 1000 Hz, such as N1, HCRT1, HCRT2, and N2. The capillary blood glucose level is measured by a glucometer before and after the HCRT. A schematic of the experimental setup is shown in Figure 1. Thermal images were captured once at every 1.5 min intervals.

The axillary temperature signal and IR images were processed as per the flow chart in Figure 2, described in subsections, to evaluate the spatiotemporal variation of temperature and its impact on HRV.

The normalized axillary temperature, as in equation (1), was 50 Hz from the powerline and high-frequency noise, which were filtered by a notch and low-pass filtered at the cutoff frequency ( f _c = 0.05 Hz). The filtered signal is convoluted with the impulse train results to obtain the fitting line on the temperature variations with an average mean square error of 0.004, as shown in Figure 3a. The differential signal from the average thresholding operation provides a rising slope in the temperature signal, as shown in Figure 3b. T normalised = T c − T amb T steady − T amb (1) where T c is the core (axillary) temperature, T amb the ambient temperature, T steady the steady-state axillary temperature.

The maximum slope of the temperature increase was of higher significance. It was obtained from the exponential fitting operation to the positively sloped region of the body temperature, as shown in Figure 4c. Image noise in the form of shadows is removed by applying median filtering ³⁶ to the double-precision image of the raw IR thermal image. The emissivity was set to that of the skin ( ε = 0.98). ³⁷ Regions of interest were obtained through image processing techniques of background subtraction that differentiate between the subject and the background, ³⁸ having only thoracic and facial areas, making the region of interest (ROI) pixel dimensions 120 × 120. The dominant colors in the image are orange and red; therefore, the thermal image is converted to a hue- saturation-value (HSV) image to obtain the intensity of the pure colors in the IR image. This helps to clearly distinguish between the hot and cold pixels. The average thresholding with a threshold of 0.3 on the HSV image provides grayscale images. The hue and saturation are masked to zero, while the value of the HSV images is masked to one to obtain the indexed image. The indexed image was then scaled between 30 ^° C to 35 ^° C to obtain the temperature image. This range is based on the skin and the core temperature difference, which lies between 2 ^° C and 4 ^° C, ³⁹ with an ideal core temperature of 37 ^° C. The temperature field obtained during postprocessing is shown in Figure 4d.

We quantified the effect of thermogenesis through indirect calorimetry from the temporal and spatial temperature changes using Penne's bioheat equation, ⁴⁰ given below in equation (2): q m = ρ C p ∂ T c ∂ t max − k [ ∂ 2 T ∂ x 2 + ∂ 2 T ∂ y 2 ] − q b (2) where q m is the metabolic heat generation per m ³, x and y are the positions of the pixels in the thermal image, and

( q b = W b C b ( T c − T ( x , y , t ) ) , q b is the tissue blood perfusion heat per m ³, and ρ , C p , k are the density, specific heat, and thermal conductivity of the skin, respectively. T is the skin temperature and T c is the core (axillary) temperature. C b is the specific heat of the blood.

Therefore, predominantly q m depends on the spatiotemporal changes in temperature. In this study, the maximum slope of axillary temperature, and ∇ 2 T are analyzed to observe the thermogenesis during HCRT.

We used the finite difference method to find the spatial change of the temperature gradient, as given below in equation (3): ∂ 2 T ∂ x 2 = T n + 1 , m − 2 T n , m + T n − 1 , m Δ x 2 , ∂ 2 T ∂ y 2 = T m + 1 , n − 2 T n , m + T m − 1 , n Δ y 2 (3) where T is the skin temperature, m is the row index along y, n is the column index along x, and Δ x , Δ y are the pixel distances along the columns and rows, respectively. The magnitude of Δ x , Δ y in terms of the physical dimension is 4 mm, as measured from the facial width of 136 mm of a typical participant from the diameter of the head containing 34 pixels.

The thermal conductivity of the dermis of the skin, which is just below the epidermis and has a thermal conductivity ( k) of 0.47 W m ⁻¹ K ⁻¹, density ( ρ ) of 1085 kg m ⁻³ and specific heat ( C _P ) of 3680 J kg ⁻¹ K ⁻¹. ⁴¹ We used the max − min value of ∇ 2 T to differentiate between the hot and cold spatial regions of the skin. Each pixel in these regions is called a hotspot (HS) or cold spot (CS). The distribution of these spots follows a kernel probability distribution function (PDF). The distance between the peaks of the distribution is referred to as spot-width (SW). We estimate the variability measures in thermogenesis by the maximum range, the standard deviation of change in the temperature gradient, the Frobenius norm, and the maximum eigenvalue of ∇ 2 T which are given below in equation (4): T = ∑ m = 1 H ∑ n = 1 W T ( m , n ) N pixels , Range = max ( HS ) − min ( CS ) , σ HS , CS = HS , CS − μ HS , CS N HS , CS | | ∇ 2 T | | F = ∑ m = 1 H ∑ n = 1 W | ∇ 2 T ( m , n ) | 2 (4) where H and W are the height and width of the thermal image, N pixels is the total number of pixels in the image; N HS , CS is the number of pixels in the hot and cold spots, max (HS) (K.mm ⁻²) is the maximum value of the Hot Spot, and min (CS) (K.mm ⁻²) is the minimum value of the Cold Spot. μ HS , CS (K.mm ⁻²), σ HS , CS (K.mm ⁻²) are the mean and standard deviation of temperature in hot and cold spots, respectively. ∇ 2 T is the spatial change of temperature gradient, and|| ∇ 2 T || _F is the Frobenius norm (K.mm ⁻²). We used singular value decomposition to find the eigenvalues of ∇ 2 T along the x- and y-axes.

HRV requires RR intervals from the filtered ECG. Raw ECG have power line interference and baseline movements due to respiration and hand motion. The power line interference in the ECG was removed using a 50 Hz notch filter. A symbol of order four is used as the mother wavelet to remove baseline wandering in ECG, and the approximate ECG is reconstructed by summing the wavelet coefficients from 3 ^rd -5 ^th level of the decomposed ECG ⁴² to obtain an enhanced QRS-region. The peak detection algorithm was used to obtain an undistorted RR interval for evaluating HRV. HRV is a measure of oscillation in heart rate over a minimum period of five minutes. ¹⁶ It has various time, frequency, geometric, and nonlinear features ¹⁶ to describe its physiological significance. In this study, HRV analysis constituted the frequency domain metrics alone. This focuses on one metric for a specific understanding of the power under frequency bands and their physiological significance solely due to HCRT. The primary analysis was to observe the VLF power and sympathovagal response due to HCRT. In the subsequent analysis, sub-band-spectrum analysis was carried out to find the frequency zone in the VLF responsible for HCRT, as the VLF power components respond to thermoregulation. ¹⁷ The sub-band of VLF-power is obtained by dividing the VLF range (0.0033 Hz to 0.04 Hz) equally into five different bands: 0.0033–0.0106, 0.0106–0.018, 0.018–0.0253, 0.0253–0.0327, and 0.0327–0.04 Hz.

The sensor setup had six BMP-280 MEMS sensors as shown in Figure 5 with a sampling frequency of 156 Hz, as shown in Figure 5. The BMP sensor module (pressure resolution: 1 Pa and temperature resolution: 0.01 ^° C from BOSCH) was used with an ST microcontroller (STM32F401CCU6 black-pill board) to record breath pressure and temperature at 1000 Hz. It runs in Real-Time-Operating-System (RTOS) mode and SPI proto- col to communicate with six pressure sensors. However, the inward sensors (3,4 in Figure 5) near the nostril, there- fore are used for measurement. Additionally, an off-the- shelf anemometer (range: 0.1 m s ⁻¹ to 30 m s ⁻¹, resolution: 0.1 m s ⁻¹, and accuracy: ±5 %) was used to obtain the velocity of the breath. The planar axis configuration for measuring breath parameters is inspired by. ⁴³ Exergy analysis was carried out considering the following assumptions: •

The standard fluid in and out of the nostril is air

Exergy constitutes enthalpy, kinetic energy, and entropy generation per \SI{}{kg} for an open system, as given in Equation (5). Exergy = C p ( T 1 − T 0 ) + 1 2 ( V max − V 0 ) 2 − T 0 ( C p ln T 1 T 0 − R ln P 1 P 0 ) (5) where T 1 and P 1 are the temperature and pressure of exiting the nostril, respectively. Vmax is the maximum exit velocity obtained from anemometer readings. The P 0 and T 0 are the pressure (base pressure recorded by the sensor) and temperature (29.5°C = 302.65 K) of the experimental environment. C p and R are the specific heat and universal gas constant of the air, respectively.

Thermogenesis has two components: temporal and spatial variations in temperature. We have taken a 10 % change in the median value of the parameters as a significant difference, ⁴⁴ as thermoregulation is a slow process. The ∇ 2 T is shown in Figure 6a, which has two regions such as hot and cold depending on the maximum value of the minimum value of ∇ 2 T tensor max (min( ∇ 2 T )). The hot and cold regions are defined with ∇ 2 T greater, and lesser than max (min( ∇ 2 T )) respectively. The histogram of ∇ 2 T at both zones provides the kernel distribution pattern as shown in Figure 6c. The spot width (SW) in Figure 6d is the distance between the peak values in the distribution, which depicts the effect of thermogenesis. An increase in thermogenesis causes a reduction in SW and vice versa.

The parameters of ∇ 2 T − matrix are tabulated in Table 2, and are plotted in the spider plot as shown in Figure 6b. The magnitudes of the maximum, mean, standard deviation, and peak of the PDF are increased during HCRT1 and sustained during HCRT2. Homogenization due to heat addition can be observed from the reduction of magnitudes in SW, range, and the Frobenius norm of ( ∇ 2 T ). The post-normal phase N2 has a higher spread in the spider plot showing an overall increase in ∇ 2 T parameters than pre-normal phase N1. The plots show a clear activation of thermo-regulation by heat addition during HCRT, reflected on the ∇T than only the temperature. There is an increase in thermogenesis, evident from an increase in the temporal slope of temperature and ∇ 2 T . The average value of ∇ 2 T at hot and cold regions rises by 3.42 %, 2.78 %, and 2.53 %, 1.84 % during HCRT1 and HCRT2 respectively compared to N1. The maximum value of ∇ 2 T has relatively increased during the HCRT phase than N1. The standard deviations of the ∇ 2 T at hot and cold spots increased by 1.47 % and 1.84 %, respectively, during HCRT2 than HCRT1 owing to the entropy generation by heat addition to the system. It eventually propagated to the post-normal phase N2, with an increase of 4.55 % at the hot spots. The peak value of kernel distribution at spots increased by 9.26 % during HCRT1, signifying the heat addition to the system during 1st phase of HCRT. This is evident as the axillary temperature slope increased by 11.54 % during HCRT1, as shown in Table 2. The field parameters during HCRT2 sustained the thermogenic effect of HCRT1. The skin temperature almost remains constant in the range of 32.72 ^°C to 32.76 ^°C, as shown in the Table 2. The median temperature of breath in both left and right nostrils is equalized by HCRT as the difference between breath temperature of left and right reduced by 88.6 % during HCRT1 and 23.03 % during HCRT2. The breath temperature rose significantly for both left (p < 0.001) and right (p = 0.0064) nostrils during HCRT1, dropped further during HCRT2 due to thermoregulatory effects, and increased further for both left (p < 0.001) and right (p = 0.0060) nostrils during N2 compared to HCRT2.

HRV improved during HCRT with a non-significant increase in HF power and a significant reduction (p < 0.05) in VLF power during HCRT2. This is evident from the VLFHF ratio for HCRT2 being significantly lower (p < 0.05) than N2 by 52.62 %, as shown in Figure 7a, with minimal changes in LF power, as shown in Figure 6d. The percentage of power in the sub-VLF-HRV band of 0.0327 Hz to 0.04 Hz during HCRT2 is significantly (p < 0.05) higher by 41.7 %. Therefore, shows the precise sub-band of VLF contributing to thermoregulation by HCRT, as shown in Figure 7b, which is a unique feature of this study. The other sub-bands did not show any significant differences during the HCRT. Moreover, there was a significant increase (p < 0.05) in the VHF power of HRV by 54.2 % and 66.54 % during HCRT2 in comparison to N1 and N2, respectively.

The boxplot in Figure 7c indicates a non-significant increase in the average exergy in both nostrils during the HCRT. However, there was a significant difference in the standard deviation of the exergy between HCRT2 and N2 (p < 0.05), as presented in Figure 7d. During HCRT1, σexergy is high for the right nostril owing to the larger box width, and subsequently decreases in HCRT2 and N2. The σexergy in the left nostril was reduced non-significantly by HCRT. Thermogenesis by HCRT is mediated by the right nostril exergy transfer, which may be attributed to the upright position of the right-hand thumb.

The CBG was measured from the prick test by the glucometer before and after the HCRT intervention of twenty minutes. The bar plot in Figure 7e shows the effect of HCRT on CBG for with twelve II DM. There was a significant reduction (p < 0.001) in CBG value by 24.14 %, as shown in Table 2. The results showed a definitive thermogenesis process during HCRT, which facilitates glycolysis. Hand clasping creates resistance to blood flow to the fingers because of the partial compression of artery branches on the lateral side of the fingers. It accelerates the heat exchange from the muscle to the blood, which may be responsible for glycolysis. The glucose utilization rate of the body is regulated by the secretion rate of insulin from the pancreas. Usually, the amount of glucose that can diffuse into cells is restricted, except for the liver and brain cells. Insulin significantly increases this diffusion. The effect of HCRT on insulin production requires further study.