Raman spectroscopy with an improved support vector machine for discrimination of thyroid and parathyroid tissues.

Area of Science:

• Biomedical Optics
• Surgical Pathology
• Chemometrics

Background:

• Thyroid surgery carries a risk of inadvertent parathyroid gland removal, leading to complications.
• Accurate intraoperative discrimination between thyroid and parathyroid tissues is clinically significant.
• Current methods for tissue discrimination lack specificity, necessitating improved diagnostic tools.

Purpose of the Study:

• To develop and validate a Raman spectroscopy-based method for discriminating thyroid and parathyroid tissues.
• To enhance the classification accuracy of support vector machine (SVM) algorithms using spectral data.
• To assess the diagnostic potential of an optimized SVM algorithm combined with partial least squares (PLS) for surgical guidance.

Main Methods:

• Raman spectroscopy was performed on tissue samples from 43 individuals.
• Data dimensionality was reduced using partial least squares (PLS).
• An improved support vector machine (SVM) algorithm, optimized via a genetic algorithm (GA), was employed for spectral analysis and classification.

Main Results:

• The developed PLS-GA-SVM algorithm demonstrated high diagnostic accuracy and reliability.
• The algorithm achieved a sensitivity of 94.67% and an accuracy of 94.44% in discriminating between thyroid and parathyroid tissues.
• The optimization of the SVM algorithm significantly improved classification performance in spectral analysis.

Conclusions:

• Raman spectroscopy combined with the PLS-GA-SVM algorithm shows significant potential for real-time, in vivo tissue discrimination during thyroid surgery.
• This technique offers a reliable and accurate method to prevent inadvertent parathyroid gland resection.
• Further clinical validation is warranted to integrate this technology into standard surgical practice.