Raman Spectroscopy for Differentiating High- and Low-Grade Canine Cutaneous Mast Cell Tumours

Canine cutaneous mast cell tumours (MCTs) demonstrate variable clinical behaviour. Histological grading remains the primary prognostic approach, but some low-grade tumours behave aggressively, and interobserver variability persists. This underscores the need for complementary and objective biomarkers. Raman spectroscopy (RS) is a non-destructive optical technique based on the inelastic scattering of monochromatic light capable of detecting subtle biochemical changes in tissues. This exploratory study evaluates whether RS could differentiate low-grade from high-grade cutaneous MCTs based on characteristic vibrational signatures. Eighteen canine cutaneous mast cell tumours were graded according to the Kiupel two-tier system, comprising 10 low-grade and 8 high-grade cases. Formalin-fixed tumour samples were analysed, yielding a total of 68 Raman spectra. Spectra were acquired with a 785 nm laser, pre-processed using baseline correction, and normalized to the 1007 cm^-1 phenylalanine band. Data normality was tested with the Shapiro-Wilk test, and group comparisons used the Mann-Whitney U test (α = 0.05). The 1007 cm^-1 band showed no significant difference between Low- and High-Grade tumours (p = 0.068), validating its use for normalization. Conversely, the 860 cm^-1 tyrosine-associated band was significantly higher in high-grade tumours (p = 0.033). Ratios normalized to the 1007 cm^-1 peak consistently discriminated between grades, with significantly higher values in low-grade tumours across multiple spectral regions-particularly 1007/1299 and 1007/1301-1320 (all p < 0.05). Overall, RS identified distinct biochemical fingerprints between tumour grades in canine cutaneous MCTs, especially in spectral regions associated with amino acids, lipids, and nucleic acids. These findings support the potential of RS as a complementary tool for MCT characterization and encourage further validation in larger cohorts with clinical outcome data.