Figure 1. Flow chart of the present study. (a) The specimens had tumours, and the stained frozen section was physically close to the tumour. (b) The SD of voltage signals in either channel was greater than 80 µV.

In the first stage, we studied 25 snap-frozen human liver specimens (weighing between 1.3 mg and 58.1 mg): 24 from remnants of pre- and post-LT wedge biopsies and one from a discarded donor liver at Leiden University Medical Center (LUMC) in August 2022. The specimens were thawed at room temperature and scanned using the Raman system under ambient room light. The liver specimens were refrozen at -80 degrees Celsius and thawed again for TG quantification. Given that the penetration depth of the incident laser beam was at least 1mm,²⁶ only specimens heavier than 10 mg (thicker than 1 mm) were included to avoid the insufficiency to quantify TG of low-mass specimens^27,28. The fat content derived from the analyses using the Raman and reflectance channels was then compared with the TG quantification results to assess the Raman system’s accuracy in steatosis estimation.

In the second stage, at Vanderbilt University Medical Center (VUMC) in November 2022, 70 snap-frozen human liver specimens weighing between 0.24 g and 2.58 g were obtained through the Cooperative Human Tissue Network. These liver specimens were split on dry ice: one half was thawed in cold water and scanned under ambient light using the investigated Raman system, while the other half was cryo-sliced and stained with Oil Red O (ORO) for histological analysis. The Raman system user (HG) and the expert pathologist (AES) performed the Raman and histopathological assessments independently, blinded to the results of each other assessment. Four liver specimens were excluded due to proximity to tumours or due to Raman data collection issues. The Raman results were compared to expert pathologist evaluations of steatosis.