A novel IagPLS baseline correction method for glioma identification using Raman spectroscopy

Area of Science:

• Biomedical Engineering
• Spectroscopy
• Computational Biology

Background:

• Gliomas are aggressive central nervous system tumors requiring advanced diagnostic tools.
• Raman spectroscopy offers non-invasive molecular fingerprinting but suffers from autofluorescence interference.
• Existing baseline correction methods fail to adequately address noise, feature preservation, and spectral heterogeneity.

Purpose of the Study:

• To develop an improved baseline correction method for Raman spectra of brain tissues.
• To enhance the accuracy and efficiency of intraoperative glioma diagnosis using Raman spectroscopy.
• To provide a robust pre-processing tool for optical biopsy systems.

Main Methods:

• Proposed an improved adaptive gradient-derived penalized least squares (IagPLS) method.
• Integrated curvature-driven dynamic regularization, SHAP algorithm-guided feature protection, and quantum-inspired global optimization.
• Validated the method on 423 clinical Raman spectra (157 normal, 266 glioma tissues).

Main Results:

• IagPLS achieved 96.1% glioma identification accuracy with random forest classification, outperforming airPLS (89.4%) and agdPLS (87.0%).
• Significantly improved spectral feature peak prominence (82.05% vs. agdPLS) and reduced negative residual area (89.79% vs. airPLS).
• Demonstrated rapid processing (<0.1s per correction) and enhanced biological interpretability.

Conclusions:

• IagPLS is a superior spectral correction technique for intraoperative glioma diagnosis via Raman spectroscopy.
• The method offers improved accuracy, speed, and interpretability for optical biopsy systems.
• The algorithmic framework is adaptable for multimodal biomedical spectral analysis in precision medicine.