Non-invasive, multi-parameter methods to estimate core body temperature offer several advantages for monitoring thermal strain, although further work is required to identify the most relevant predictor measures. This study aimed to compare the validity of an existing and two novel multi-parameter rectal temperature prediction models. Thirteen healthy male participants (age 30.9 ± 5.4 years) performed two experimental sessions. The experimental procedure comprised 15 min baseline seated rest (23.2 ± 0.3°C, 24.5 ± 1.6% relative humidity), followed by 15 min seated rest and cycling in a climatic chamber (35.4 ± 0.2°C, 56.5 ± 3.9% relative humidity; to +1.5°C or maximally 38.5°C rectal temperature, duration 20–60 min), with a final 30 min seated rest outside the chamber. In session 1, participants exercised at 75% of their heart rate maximum (HR max) and wore light athletic clothing (t-shirt and shorts), while in session 2, participants exercised at 50% HR max, wearing protective firefighter clothing (jacket and trousers). The first new prediction model, comprising the input of 18 non-invasive measures, i.e., insulated and non-insulated skin temperature, heat flux, and heart rate (“Max-Input Model”, standard error of the estimate [SEE] = 0.28°C, R2 = 0.70), did not exceed the predictive power of a previously reported model which included six measures and no insulated skin temperatures (SEE = 0.28°C, R2 = 0.71). Moreover, a second new prediction model that contained only the two most relevant parameters (heart rate and insulated skin temperature at the scapula) performed similarly (“Min-Input Model”, SEE = 0.29, R2 = 0.68). In conclusion, the “Min-Input Model” provided comparable validity and superior practicality (only two measurement parameters) for estimating rectal temperature versus two other models requiring six or more input measures.

In this paper, we derive a randomized version of the Mirror-Prox method for solving some structured matrix saddle-point problems, such as the maximal eigenvalue minimization problem. Deterministic first-order schemes, such as Nesterov's Smoothing Techniques or standard Mirror-Prox methods, require the exact computation of a matrix exponential at every iteration, limiting the size of the problems they can solve. Our method allows us to use stochastic approximations of matrix exponentials. We prove that our randomized scheme decreases significantly the complexity of its deterministic counterpart for large-scale matrix saddle-point problems. Numerical experiments illustrate and confirm our theoretical results.

Wearable devices for remote and continuous health monitoring in older populations frequently include sensors for body temperature measurements (i.e., skin and core body temperatures). Healthy aging is associated with core body temperatures that are in the lower range of age-related normal values (36.3 ± 0.6°C, oral temperature), while patients with Alzheimer’s disease (AD) exhibit core body temperatures above normal values (up to 0.2°C). However, the relation of body temperature measures with neurocognitive health in older adults remains unknown. This study aimed to explore the association of body temperature with cognitive performance in older adults with and without mild cognitive impairment (MCI). Eighty community-dwelling older adults (≥65 years) participated, of which 54 participants were cognitively healthy and 26 participants met the criteria for MCI. Skin temperatures at the rib cage and the scapula were measured in the laboratory (single-point measurement) and neuropsychological tests were conducted to assess general cognitive performance, episodic memory, verbal fluency, executive function, and processing speed. In a subgroup (n = 15, nine healthy, six MCI), skin and core body temperatures were measured continuously during 12 h of habitual daily activities (long-term measurement). Spearman’s partial correlation analyses, controlled for age, revealed that lower median body temperature and higher peak-to-peak body temperature amplitude was associated with better general cognitive performance and with better performance in specific domains of cognition; [e.g., rib median skin temperature (single-point) vs. processing speed: rs = 0.33, p = 0.002; rib median skin temperature (long-term) vs. executive function: rs = 0.56, p = 0.023; and peak-to-peak core body temperature amplitude (long-term) vs. episodic memory: rs = 0.51, p = 0.032]. Additionally, cognitively healthy older adults showed lower median body temperature and higher peak-to-peak body temperature amplitude compared to older adults with MCI (e.g., rib median skin temperature, single-point: p = 0.035, r = 0.20). We conclude that both skin and core body temperature measures are potential early biomarkers of cognitive decline and preclinical symptoms of MCI/AD. It may therefore be promising to integrate body temperature measures into multi-parameter systems for the remote and continuous monitoring of neurocognitive health in older adults.