Introduction

Hepatic steatosis (HS), particularly macrovesicular steatosis (MaS) in donor livers, has been found to be linked to a higher likelihood of graft dysfunction following liver transplantation (LT).1,2 Severe MaS (≥60%) is typically viewed as high risk and leads to organs being discarded.3 As the donor liver pool expands to accommodate the growing number of patients on waiting lists, it is increasingly essential to conduct thorough assessments of survival benefits among candidates.4,5High-quality estimations of risk factors are crucial, as donors with less favourable characteristics are now more frequently considered for transplantation. The recovery rate of donor livers with moderate (30%-60%) MaS has increased over the years, while other risk factors of mortality or graft loss were mitigated6,7. However, the discard rate of moderately and severely steatotic livers is still high because of their higher risk of graft dysfunction.8,9 According to a recent study on 17,801 liver transplant recipients, compared with recipients of grafts with 0-10% MaS, the hazard of graft failure was found to be 53% and 25% higher among the recipients of grafts with >30% MaS and 10%-30% MaS, respectively.10 Therefore, precise classifications and quantification of steatotic donor livers, alongside a clear-cut threshold at different degrees of MaS, are crucial for optimizing liver allocation and maximizing the survival benefit of liver recipients. At present, donor livers are assessed macroscopically at the time of recovery.11 This assessment is often subjective and dependent on the experience of the donor surgeon. If necessary, frozen sections can be obtained, but this often is associated with significant delays and requires the availability of an expert pathologist.12 A tool that can provide objective, quantitative and rapid analysis of (donor) liver fat content is needed. Optical spectroscopic tools, including infrared (IR), reflectance, and Raman, have shown significant potential in evaluating HS due to their minimally invasive nature and rapid analysis capabilities.11,13 Recent studies on both animal models and human livers have demonstrated that IR spectroscopy14–16 and reflectance spectroscopy17,18 could evaluate various degrees of global HS (with out differentiating MaS); however, those applications required either switching off (directing away) surgical lights or using an invasive optical needle in the operating room (OR). Furthermore, complex spectral analyses of acquired data were a necessity. Raman spectroscopy offers some distinct advantages. It is sensitive to rotations and vibrations of chemical bonds, making its inherent high molecular specificity ideal for characterizing biological materials.11,19 Pre-clinical studies on ex vivo rodent liver specimens revealed that Raman spectroscopy could quantify HS in animal models using signals in the high wavenumber region 2800–3000 cm−1, and the results agreed well with pathology ratings.20–22 Confocal Raman microscopy has been reported to effectively analyze the sizes of lipid droplets in rodent livers.23–25 This offers insights relevant to microvesicular steatosis (MiS) and MaS. However, obtaining a Raman image can be considered too time-consuming to be helpful when assessing the HS of donor livers, as each pixel requires spectral analyses. The exploration of HS in LT using conventional Raman spectroscopy lags IR and reflectance spectroscopy. Hewitt et al. (2015)20 and Pacia et al. (2018)21proved the concept that conventional Raman spectrometers could assess the fat content of rodent livers. However, neither of the two studies discussed HS assessment at morphological levels. Navigating the challenges associated with weak signal detection and fluorescence interferences, we engineered a filter-based 1064-nm Raman system validated through examinations using phantom models and duck liver samples.26 In our previous study, employing readily interpretable voltage intensities to quickly quantify the relative fat content within the examined liver samples, this multi-channel system functioned reliably (r2 = 0.934) under normal and intense ambient light conditions26. In the present study, we analyzed 95 specimens in two sets from two medical centers, utilizing our filter-based 1064-nm Raman system. For the initial set of 16 specimens, fat contents converted from Raman and reflectance intensities were contrasted with triglyceride (TG) quantification results. Subsequently, for the more extensive set of 66 specimens, fat contents derived from Raman and reflectance intensities were compared to evaluations of MaS and global HS (including MiS and MaS) performed by an expert pathologist and to determinations of positive pixels made by an algorithm.