Rspssl: A Novel High-fidelity Raman Spectral Preprocessing Scheme To Enhance Biomedical Applications And Chemical Resolution Visualization

The article discusses a novel Raman spectral preprocessing model, RSBPCNN, which uses a patches strategy and self-supervision strategy to improve feature learning ability and avoid intensity interference from baselines. The model can function independently in arbitrary cross-device spectral denoising and baseline corrections without further data training. It has been evaluated through an experimental data preprocessing trial and shows promising results in various biomedical applications.