Machine learning-assisted serum Raman spectroscopy for risk stratification in men with suspected prostate cancer: a retrospective proof-of-concept study

Machine learning on dried-serum Raman spectra distinguishes men with histologically confirmed prostate cancer from symptomatic controls, with AUC about 0.77 overall and about 0.78 in PSA 4–10 ng/mL.

• Why it matters: PSA has limited specificity, especially for PSA 4–10 ng/mL, creating a need for blood-based tests that better stratify risk before imaging or biopsy.
• What they did: Archived sera from 119 prostate cancer cases and 60 symptomatic controls were measured by dried-serum Raman spectroscopy, then used in supervised classifiers to output predicted probabilities. Performance was assessed in the full cohort and in the PSA 4–10 ng/mL subgroup (54 cases, 41 controls).
• The result: Predicted probabilities were higher in cancer cases, and discrimination stayed moderate in the PSA grey zone, supporting Raman spectroscopy as a potential second-line adjunct pending prospective validation.

AI summary of the abstract · may contain errors · not medical advice · read the paper.