Oedema is a common clinical finding in critically ill neonates and may reflect systemic illness such as congestive heart failure, hepatic cirrhosis, nephrotic syndrome, sepsis, and acute kidney injury. Oedema is characterised by tissue swelling due to water accumulation in the interstitial space. Currently, the gold standard in clinical practice is visual assessment, which is subjective and limited in accuracy. Alternative methods, such as ultrasound and bioimpedance, have been explored; however, they are unsuitable in neonates and do not provide direct water quantification. Near-infrared spectroscopy (NIRS) is a non-invasive optical method that could measure water content through light interaction between near-infrared light and OH particles within the tissue. This study validated NIRS for oedema assessment using an ex vivo porcine skin model, where controlled oedema was induced by phosphate-buffered saline (PBS) injection. Continuous spectroscopic data were collected via optical fibres positioned perpendicularly and parallel to the tissue. Regression models were developed and evaluated using the spectral data, with partial least squares (PLS) regression outperforming ridge regression (RR) and support vector regression (SVR). Notably, spectra acquired in the parallel configuration yielded superior results (R² = 0.97, RMSE = 0.15). These findings support the potential of NIRS as a reliable, quantitative tool for neonatal oedema assessment.

Generalised oedema is characterised by tissue swelling due to excess water accumulation. It is a frequent complication encountered in infants admitted to paediatric and neonatal intensive care. In clinical practice, oedema assessment depends on a combination of weight monitoring, fluid balance calculations and direct visual inspection of the skin. However, all of these lack accuracy and sensitivity, especially in critically ill infants. We report the use of near infrared (NIR) spectroscopy using a portable spectrometer and an optical fibre with a custom attachment allowing continuous recordings for 30 min. Data were collected from 20 neonates and infants admitted to the intensive care unit, with or without oedema, across three consecutive days. Statistically significant differences were found in the 1450 nm features, amplitude (p = 0.009), ratio (p < 0.001) and band area (p < 0.001), and the 970 nm ratio followed the same direction (p = 0.005), all with higher values in the oedema group. A late fusion model, combining spectral data and clinical features, effectively differentiated between oedematous and non-oedematous infants, achieving AUC 0.909 ± 0.016, p = 0.004. This proof-of-concept study suggests that near-infrared spectroscopy can detect oedema in neonates and infants, as assessed against clinical evaluation, laying the groundwork for its potential application as a non-invasive clinical detection and monitoring tool.