Multi-wavelength graph convolutional network for high-performance sparse multispectral optoacoustic tomography.

Area of Science:

• Biomedical Imaging
• Medical Physics
• Computational Imaging

Background:

• Multispectral optoacoustic tomography (MSOT) offers label-free, high-resolution, deep-tissue biomedical imaging.
• High costs and computational needs limit current MSOT applications for in vivo studies.

Purpose of the Study:

• To develop a cost-effective and computationally efficient MSOT method for high-quality in vivo imaging.
• To address the limitations of sparse data acquisition in MSOT.

Main Methods:

• A multi-wavelength graph convolutional network was proposed for sparse MSOT reconstruction.
• A graph learning framework integrated a multi-wavelength sparse sampling strategy.
• The method modeled correlations in artifact distributions across sparse transducer configurations.

Main Results:

• The proposed method achieved high-performance sparse MSOT imaging with only 16 transducer elements.
• Reconstruction quality was validated with SSIM of 0.92 ± 0.01 and PSNR of 27.74 ± 1.27 in vivo.
• Demonstrated a flexible and practical solution for sparse MSOT.

Conclusions:

• The graph convolutional network effectively solves the ill-conditioned sparse reconstruction problem in MSOT.
• This approach significantly enhances the feasibility of high-quality, low-cost in vivo MSOT imaging.
• The method provides a practical solution for advancing MSOT applications.