Serum Raman spectroscopy as a blood-based triage test for bladder cancer: a retrospective proof-of-concept study

Background Bladder cancer remains one of the most common urological malignancies. Current diagnostic pathways rely heavily on cystoscopy, an invasive investigation with significant cost and resource implications. A rapid blood-based assay capable of improving patient stratification before cystoscopy could reduce unnecessary invasive investigations while maintaining diagnostic sensitivity. Methods Archived serum samples from patients with histologically confirmed bladder cancer and symptomatic control subjects were analysed using dried-serum Raman spectroscopy. Raman spectra were incorporated into supervised machine-learning classification models to generate predicted probabilities of bladder cancer. Diagnostic performance was evaluated using receiver operating characteristic (ROC) analysis. Results The study included 49 patients with bladder cancer and 97 symptomatic controls. Predicted probabilities of bladder cancer were substantially higher in cancer patients than in controls. The Raman classifier demonstrated excellent diagnostic discrimination (AUC approximately 0.90). Incorporation of age into the classification model produced little additional improvement (AUC approximately 0.91). Conclusions Machine learning-assisted analysis of dried-serum Raman spectra demonstrates promising diagnostic performance for bladder cancer. Prospective evaluation is warranted to determine its role as a minimally invasive triage test within contemporary bladder cancer diagnostic pathways.