An adaptive stacking generalization integrated with Raman spectroscopy feature enhancement algorithm for fine glioma...

Area of Science:

• Neuro-oncology
• Biomedical Spectroscopy
• Machine Learning in Medicine

Background:

• Accurate intraoperative grading of brain gliomas is essential for personalized treatment and improved patient outcomes.
• Raman spectroscopy offers non-invasive, real-time biomolecular analysis for potential in situ tumor grading.
• Distinguishing between low- and high-grade gliomas is challenging due to subtle spectral differences and low signal-to-noise ratios.

Purpose of the Study:

• To develop an advanced algorithm for accurate intraoperative brain glioma grading using Raman spectroscopy.
• To overcome limitations of conventional algorithms in detecting subtle spectral features of gliomas.
• To enhance the capabilities of portable, non-invasive tumor grading devices.

Main Methods:

• Proposed an adaptive stacking generalization integrated with Raman spectroscopy feature enhancement (ASG-RSFE) algorithm.
• Implemented a Raman characteristic peak ratio approach for spectral feature enhancement.
• Utilized the butterfly optimization algorithm (BOA) to optimize the stacking ensemble strategy by dynamically adjusting sample weights.

Main Results:

• The ASG-RSFE algorithm achieved over 80% accuracy in classifying normal brain tissue, low-grade gliomas, and high-grade gliomas.
• Demonstrated significant improvements in accuracy and generalization compared to conventional recognition algorithms.
• Successfully amplified subtle biomolecular changes indicative of glioma lesions.

Conclusions:

• The ASG-RSFE algorithm shows high potential for accurate intraoperative brain glioma grading.
• This approach can facilitate the development of portable, cost-effective, and non-invasive tumor grading instruments.
• Enables clinicians to create more precise and personalized surgical strategies for glioma patients.