Minimally invasive detection of early-stage opisthorchiasis-associated cholangiocarcinoma using label-free surface-enhanced Raman spectroscopy (SERS) of hamster serum

TL;DR

This study shows that a non-invasive technique using Raman spectroscopy and machine learning can detect early signs of a deadly bile duct cancer in hamsters.

Contribution

A label-free SERS-based method for early detection of CCA in a hamster model is proposed and validated.

Findings

SERS combined with machine learning achieved 93% sensitivity and 95% specificity for detecting CCA and precancerous lesions.

The method showed an accuracy of ≥67% with an AUC exceeding 0.67 for early-stage detection.

Histopathology confirmed the progression from inflammation to CCA in the hamster model.

Abstract

Cholangiocarcinoma (CCA) is a deadly cancer often detected late. Current diagnostic methods, such as ultrasound and invasive biopsies, have limitations; there is a critical need for a rapid, minimally invasive and effective strategy for the early diagnosis and staging of CCA.

We aimed to address this need using serum samples and label-free surface-enhanced Raman spectroscopy (SERS) combined with machine learning. CCA development was induced in hamsters using a combination of Opisthorchis viverrini infection and administration of N-nitrosodimethylamine, with induction time courses spanning 1–5 month(s). Normal and pathological stages (inflammation, precancerous lesion, and CCA) were assigned based on histopathological features, as well as the expression of cytokeratin 19 and alpha-fetoprotein. Raman spectra were subjected to dimensionality reduction using principal component analysis, and diagnostic clusters were acquired using partial least-squares discriminant analysis.

Histopathological analysis confirmed a clear path towards CCA, initiated by marked inflammation, progressing to include significant cholangiofibrosis and cholangiofibroma in the precancerous stage, and culminating in definitive CCA tumor development. The integration of SERS and machine learning achieved a diagnostic sensitivity of 93%, specificity of 95%, and accuracy of ≥ 67% for precancerous lesions and CCA, with an area under the receiver operating characteristic curve exceeding 0.67.

Our findings demonstrate that this cost-effective, label-free SERS approach can accurately detect precancerous and cancerous stages of cholangiocarcinoma in a hamster model, highlighting its strong potential for future development as a community-based screening tool.

Figures