Necrotizing enterocolitis (NEC) is a critical gastrointestinal disease in preterm infants and remains a leading cause of morbidity and mortality in neonatal intensive care units [1]-[4]. Early identification is essential to prevent severe complications, yet diagnosis remains challenging due to nonspecific early symptoms [5][6]. Previous studies have explored physiological indicators, including perfusion-related parameters and thermal regulation, to assess neonatal circulatory status and detect early abnormalities [7]-[10].

Several physiological monitoring strategies have been explored as adjuncts to clinical assessment for earlier NEC identification. Peripheral perfusion indices, heart rate variability, and near-infrared spectroscopy have shown promise individually, yet each captures only one dimension of the underlying pathophysiology [5]-[7][9][11]-[13]. Skin temperature at peripheral and central sites provides an accessible, non-contact correlate of circulatory redistribution: as splanchnic vasoconstriction occurs, abdominal–peripheral temperature gradients may widen in proportion to haemodynamic compromise [9][10][14]. However, whether such gradients are detectable with sufficient resolution during mild or suspected NEC, before overt clinical deterioration, has not been systematically evaluated. The present pilot study therefore addressed two specific objectives: to determine whether continuous, multi-site skin temperature recording is technically feasible in a NICU environment during a standardised 20-minute epoch, and to characterise the preliminary temperature signal features, including inter-site gradients and intra-individual variability in a cohort of preterm infants with clinically suspected NEC.

A total of 16 preterm infants clinically suspected of NEC were included in the analysis. According to the modified Bell’s criteria, 9 infants were classified as stage I (suspected NEC), 6 as stage II (definite NEC), and 1 as stage III (advanced NEC). The single stage III case was identified retrospectively: the infant was enrolled at stage I presentation and progressed to stage III within 24 hours following the monitoring epoch, as confirmed by abdominal radiograph demonstrating intestinal perforation and clinical deterioration requiring surgical consultation.

Continuous physiological monitoring was successfully implemented in all 16 participants, demonstrating the feasibility of real-time, non-invasive data acquisition in a NICU setting. Regarding the pre-specified feasibility outcomes: sensor placement was achieved at all three anatomical sites in all 16 infants (100% placement success rate) without the need for repositioning after initial application. The mean proportion of analyzable recording time per participant was 97.3% (range: 94.1% - 100%), exceeding the 80% a priori threshold in every case. A total of three transient signal-loss events were recorded across the cohort (0.19 events per participant-session), each lasting fewer than 30 seconds and attributable to brief cable movement during routine nursing care. No skin-related adverse events, including pressure marks, erythema, or skin breakdown, were observed in any participant. All physiological data were successfully recorded and stored for analysis.

Analysis of temperature data examined four pre-specified signal features across the 20-minute monitoring epoch for each subject (Table 2). Mean abdominal skin temperature across all subjects was 36.4˚C ± 0.6˚C (range: 35.1˚C - 37.4˚C), while mean hand and foot temperatures were 35.8˚C ± 0.7˚C (range: 34.6˚C - 36.9˚C) and 35.5˚C ± 0.8˚C (range: 34.2˚C - 36.8˚C), respectively. No statistically meaningful between-subject separation was observed across NEC stages for any measurement site. The mean abdomen–foot temperature differential (ΔT) across subjects was 0.9˚C ± 0.6˚C (range: 0.1˚C - 2.1˚C), and values did not demonstrate consistent elevation or directionality sufficient to distinguish stage I from stage II cases. In addition, intra-individual temperature variability remained low at all three measurement sites, with within-subject standard deviation values over the monitoring epoch ranging from 0.2˚C to 0.4˚C across subjects and sites (Table 2).

Table 2. Summary of pre-specified temperature signal features across all subjects (n = 16).

Overall, the findings indicate that temperature-based measurements alone did not provide sufficient differentiation for identifying early-stage NEC in this cohort. The limited variability observed may be attributable to the relatively mild disease severity and small sample size (Figures 2-4).

Figure 2. Measuring an infant’s hand temperature.

Figure 3. Measuring an infant’s abdominal temperature.

Figure 4. Measuring an infant’s foot temperature.

This pilot study evaluated the technical feasibility of continuous, multi-site skin temperature monitoring in preterm infants with clinically suspected NEC. An important limitation is that this study did not test prospective “early detection” in the strict sense: infants were enrolled only after clinical suspicion of NEC had already been raised, and monitoring was conducted over a single 20-minute epoch. The design therefore evaluates monitoring feasibility and signal characterisation, not the ability to detect NEC before clinical recognition. Contrary to initial expectations, temperature variations across measurement sites did not demonstrate clear or consistent patterns during the monitoring period.

One possible explanation is that most enrolled cases were classified as mild (stage I - II), where physiological changes may not yet be sufficiently pronounced to produce detectable temperature differences. Previous studies have shown that physiological signals in preterm infants can be highly variable and influenced by developmental status and environmental factors [17]-[20]. This finding suggests that temperature alone may lack sensitivity as an early biomarker for NEC.

Importantly, these results provide valuable insight into the limitations of single-parameter monitoring approaches. While temperature monitoring is non-invasive and clinically convenient, it may not adequately capture the complex physiological changes associated with early NEC. Prior studies have emphasized that peripheral perfusion and hemodynamic responses are multifactorial and may not be fully reflected by a single physiological parameter [21][22].

This study demonstrates the feasibility of continuous, real-time physiological monitoring in a neonatal intensive care setting. The integration of additional physiological parameters may offer a more comprehensive and sensitive approach for early detection. Previous research has suggested that multi-modal monitoring, including cardiovascular and autonomic indicators, may improve the assessment of neonatal physiological stability [6][13][23][24].

Future research should focus on multi-modal monitoring strategies and larger cohorts, particularly including more severe NEC cases, to better characterize physiological patterns and improve early diagnostic capabilities. In addition, clinical care strategies such as kangaroo care and supportive developmental interventions may further influence physiological regulation and should be considered in future investigations [25][26]-[30].

This study evaluated the feasibility of multi-site skin temperature monitoring for the early assessment of necrotizing enterocolitis in preterm infants. The findings demonstrated that temperature variations alone did not provide sufficient discriminatory power for identifying early-stage NEC, particularly in cases with mild disease severity.

These results suggest that temperature-based monitoring, when used in isolation, may have limited clinical utility for early NEC detection. However, the study confirms the feasibility of continuous, non-invasive physiological monitoring in the neonatal intensive care environment.

Importantly, the findings highlight the need for integrating additional physiological parameters, such as multi-modal physiological data, to enhance sensitivity and diagnostic accuracy. Future studies with larger sample sizes and more severe NEC cases are warranted to further validate and refine this monitoring approach.

The authors gratefully acknowledge the support of Taichung Veterans General Hospital and Feng Chia University for providing resources for this research project. Special thanks are extended to Dr. Te-Ming Wang for his valuable clinical guidance and professional expertise throughout the study.