The aim of our study was to determine the feasibility and suitability of IRT as a noncontact, bedside, clinical tool for the detection of stress and discomfort in neonates through stress-induced facial thermal variations. Hunger was chosen as a common, mild non-pain associated stressor. Temperatures were evaluated in six ROIs: nose tip, periorbital and supraorbital region, forehead, perioral region and chin.

Our study provides evidence that (a) it is possible to obtain suitable images in which all ROIs are visible at the same time with noncooperative neonatal patients; (b) hunger as a mild, non-pain associated stressor led to measurable and significant facial temperature variations; (c) facial action has no significant effect on facial temperature and showed no significant interaction effect with the stressor (hunger).

Autonomic arousal, defined by changes in the activity of sympathetic and parasympathetic branches of the autonomic nervous system, represents an important tool in stress research. The autonomic nervous system exerts a regulatory function, helping the body adapt to internal and environmental demands and, thus, is an important regulator of emotional and stress-related physiological response.39 Various measures (e.g., cardiovascular and electrodermal activity) can be used to examine these changes in activity as a response to stressful stimuli.40,41,42,43 Thermal variation and, more specifically, facial temperature have also been shown to be a measure of arousal variations.26,42,44 Furthermore, it has been determined that, compared to other established physiological stress markers, the face temperature correlates with stress-induced mood changes, finding correlations between thermal imprints and stress-induced psychological responses. Engert et al., for instance, suggested that the observable thermal imprints reflect the general arousal that underlies a stress experience.20 Nonetheless, as previously discussed, the question of what kind of thermal response is expected to specifically reflect stress-related autonomic arousal has different answers throughout the literature. Acute stress (e.g., acute pain and acute psychologically stressing situations) is mentioned to lead to a decrease in facial temperature as a reflection of sympathetic activation.13,14,15,20 However, stress and the arousal-related increase of facial temperature has also been observed, mostly in relation to psychological stressors and shifts in emotion.16,17,18,19 Research on infants has shown an increase in temperature in response to mild social stressors, such as the “Still-Face Paradigm,” where the mother suddenly ceases all interaction with the child, resulting in a negative effect on the child. A decrease in temperature has been observed in response to positive emotional states, such as when infants laughed.23,34 We could show in our study that hunger, as a mild, non-pain associated stressor, leads to a significant increase in facial temperatures in premature and term neonates. This could indicate the activation of the parasympathetic component of the autonomic nervous system, as an increase in parasympathetic activity leads to increased peripheral vasodilatation. Arousal mediated by the parasympathetic nervous system has been described in children.23,40 The results of our study, therefore, support the theory established by Engert et al.20 that thermal images reflect the unspecific arousal that underlies a stress experience. This inevitably poses the question of the suitability of this method to specifically detect strongly distressing and painful stimuli in neonates. It would, therefore, be desirable to further study the facial thermal response patterns of neonates to strongly distressing and painful stimuli where a stronger sympathetic activity can be expected and how these differ from the unspecific arousal response that was observed in this study. Whether our results can be transferrable into painful stress has yet to be studied.

When it comes to measuring facial thermal response, the nose tip has repeatedly been identified as the most stress-sensitive region of the face.15,20,33,34,35,37 The results of our study confirm these observations. In our study, the nose tip was the ROI with the most important hunger-related rise in temperature, which was shown to be highly significant.

The corrugator and periorbital region have also received great interest in stress research.15,16,18,19,20,26,35,36 They play, similar to the nose tip, a central role in behavioral clinical scores taking into account facial action in these regions. In contrast to the nasal region, however, results in the literature differ. While some authors report no thermal response to stress in this region,20 others found the opposite as they observed the experience of stress being associated with an instantaneous increase in blood flow to the eye and corrugator region, leading to an increase in temperature most notably in the anastomosis region of the dorsal nasal artery and the angular artery.18,19,35,45 In our study, the periorbital and supraorbital corrugator region depicted significant hunger-related thermal response.

The forehead region is often described as rather insensitive in the literature.20,23,34,35,37 In our study, the forehead region showed the smallest absolute thermal variance and the temperature differences measured were not significant.

The perioral and chin region are not very often described in facial IRT-based research, although they have shown great thermal sensitivity and response to emotional and distressing stimuli along with arousal.20,26,37 In our study, the perioral and chin region had highly significant hunger-related thermal variations and the overall strongest thermal response after the nose tip.

No significant effect of facial action on the temperatures measured was observed in our study and an interaction effect of facial action was ruled out. Although the evaluation of the facial expression is an important element in various behavioral and clinical scales for the detection of stress-induced changes, our results suggest that facial action is not a relevant factor when analyzing stress-induced thermal responses through IRT. The thermal variations observed reflect the autonomous stress response and arousal and, therefore, possibly allow a more objective, contact-free assessment of the premature and term neonate’s state, independent of behavioral indicators. The IRT could, thus, be an answer to the challenge of assessing stress and pain in neonates, taking into account the issue of different gestational ages having widely varied behavioral reactions. The smallest and sickest neonates having the least amount of behavioral responses to pain and the neonates with lower gestational age expressing less behavioral pain than more mature neonates.2,3,6,7,9,11 Moreover, studying the thermal response patterns to stressful and painful stimuli in neonates of different levels of maturity would, therefore, be of great interest to understand more clearly the possible role gestational age plays in facial thermal response patterns.

Additionally, the possibility of combining facial imaging data with other established stress markers to improve the predictive modeling could be explored in future studies.

The primary methodological limitation of this study is its small sample size. The observations and conclusions made in this pilot study are, therefore, to be regarded as preliminary until reproduced in a larger sample and are not adequate to evaluate clinical practice change decisions. Additionally, it would be desirable to study the normal fluctuation in facial thermal activity over time, particularly in terms of circadian rhythm, as well as studying the possible effects of gestational and chronological (since birth) ages. There is also a need to repeat the study on the patterns of facial thermal response in larger populations and future studies in diverse populations, including participants with ancestry from different geographic regions of the world. The thermal patterns observed may only occur in this standardized protocol setting based on the feeding rhythms of the study population and may not be generalized towards other stressors, for example, painful stimuli, or other mild stressors, such as full diapers. Therefore, additional studies in different standardized settings are desirable.

Changes in the homeostasis and cutaneous adaptation to the environment are limiting factors regarding the use of IRT imaging. No special measures were taken to avoid thermoregulatory or acclimatization processes in this study; it did not take place in the controlled temperature and humidity setting of a stress lab or incubator. However, the ambient temperature in the participant’s room remained air conditioner-controlled between 23 and 25 °C at all times. Additionally, an acclimatization period of 30 min was allowed before post-feeding recordings and the average weight of the participants at the time of the recordings allows us to assume that the participants had largely attained their thermoregulatory competence.

The thermal accuracy of the recordings also depends on the image quality. A highly controlled setting without motion artifacts or obstruction of ROIs with, for example, the subject’s hand would be ideal. These conditions could not be met for the noncooperative target population selected. A lot of images in this study were discarded due to image quality issues involving motion artifacts and ROI obstruction, thus, also limiting the clinical applicability of the method. This problem could be reduced in future studies by selecting only one ROI (e.g., the nose-tip) for study.