Spatial and Spectral Reconstruction of Breast Lumpectomy Hyperspectral Images

Area of Science:

• Medical Imaging
• Surgical Technology
• Computational Imaging

Background:

• Hyperspectral imaging (HSI) is a promising technique for intraoperative margin assessment in breast-conserving surgery.
• A key challenge is the trade-off between image quality and acquisition speed, crucial for real-time surgical guidance.
• Current HSI methods face limitations in speed and image fidelity for effective intraoperative use.

Purpose of the Study:

• To develop a deep learning framework for reconstructing high-resolution HSI from low-resolution HSI and high-resolution RGB images.
• To enable fast data acquisition without compromising image quality during surgery.
• To enhance the accuracy of intraoperative margin assessment in breast cancer surgery.

Main Methods:

• A deep learning-based spatial-spectral reconstruction framework was designed.
• The framework utilizes a low-resolution HSI and a high-resolution RGB image as input.
• The model reconstructs a high-resolution HSI, effectively upscaling the spectral and spatial information.

Main Results:

• The framework successfully achieved fast data acquisition while maintaining high image quality.
• The reconstruction method demonstrated robustness against common imaging challenges like motion blur, dead pixels, sensor noise, and specular reflections.
• High-resolution HSI was generated even in complex surgical scenarios.

Conclusions:

• The developed deep learning framework facilitates accurate intraoperative margin assessment using hyperspectral imaging.
• This approach overcomes the speed-quality trade-off, making HSI more practical for real-time surgical applications.
• It offers a significant advancement for improving decision-making during breast-conserving surgery.