Cosmic spike removal based on multi-feature fusion for long-time integrated Raman spectral detection

Area of Science:

• Spectroscopy
• Analytical Chemistry
• Data Science

Background:

• Raman spectral detection requires long integration times, increasing cosmic spike interference.
• Distinguishing cosmic spikes from Raman peaks is challenging due to spike variants.
• Accurate cosmic spike identification is crucial for reliable substance detection.

Purpose of the Study:

• To develop an accurate method for identifying and removing cosmic spikes in Raman spectra.
• To address the challenge of discriminating cosmic spikes from Raman peaks, including various spike types.
• To improve the quality of Raman spectra for enhanced substance identification and quantification.

Main Methods:

• A multi-feature fusion approach was employed for cosmic spike removal.
• The method involved spike rough screening, boundary localization, fine discrimination, and interpolating correction.
• Spike discrimination utilized a scoring system based on features like 'Widths in Pixels'.

Main Results:

• The method successfully identified and corrected typical and variant cosmic spikes (narrow/low, high/wide, overlapping).
• A high removal rate of 98.8% was achieved for over 3300 cosmic spikes across diverse sample types.
• The approach demonstrated good accuracy and versatility on breast tissue and pure substance spectra.

Conclusions:

• Multi-feature fusion enables efficient single-spectrum cosmic spike removal without affecting Raman peaks.
• The method provides high-quality Raman spectra for accurate substance identification and quantification.
• This technique facilitates automatic preprocessing of large spectral datasets for deep learning applications.