Glioblastoma Detection with Hyperspectral Image Analysis through Optimal Wavelength Selection.

Area of Science:

• Neurosurgery
• Medical Imaging
• Artificial Intelligence

Background:

• Glioblastoma is an aggressive brain tumor requiring precise surgical removal.
• Accurate tumor delineation is critical to maximize resection while minimizing neurological damage.
• Intraoperative imaging techniques can enhance surgical precision and patient outcomes.

Purpose of the Study:

• To investigate the efficacy of hyperspectral imaging (HSI) for intraoperative glioblastoma detection.
• To develop and evaluate an AI-driven image analysis framework for HSI data.
• To assess the potential of HSI in improving glioblastoma resection guided by AI.

Main Methods:

• Acquisition and analysis of 24 HSI datasets from 14 glioblastoma patients.
• Application of spectral and spatial dimensionality reduction techniques.
• Implementation and testing of AI-based classification models, including a multi-layer perceptron optimized with the Ant Colony algorithm.

Main Results:

• The AI framework achieved an 86.65% macro F1 score for glioblastoma detection using 20 selected hyperspectral wavelengths.
• The proposed method demonstrated superior performance compared to existing state-of-the-art approaches.
• Dimensionality reduction techniques facilitated clinical applicability of HSI.

Conclusions:

• Hyperspectral imaging, coupled with an effective AI analysis framework, shows significant potential for intraoperative glioblastoma detection.
• The study highlights the feasibility of using informative spectral wavelengths for grade 4 brain tumor identification.
• This non-invasive imaging approach can advance image-guided glioblastoma surgery and improve patient care.