Fast Hyperspectral Diffuse Optical Imaging Method with Joint Sparsity.

Area of Science:

• Biomedical Imaging
• Medical Physics
• Computational Imaging

Background:

• Diffuse optical tomography (DOT) is crucial for clinical diagnosis and treatment.
• Reconstructing DOT images becomes challenging with increasing data complexity (more wavelengths).

Purpose of the Study:

• To address the computational challenges in hyperspectral DOT (hyDOT) imaging.
• To develop a faster and more accurate image reconstruction algorithm for hyDOT.

Main Methods:

• The hyperspectral DOT inverse problem is framed as a multiple-measurement vector (MMV) problem.
• A novel fast stochastic greedy algorithm, MStoGradMP, combined with mini-batching is proposed.
• Exploits the joint sparsity of the images for reconstruction.

Main Results:

• The proposed MStoGradMP algorithm achieves higher reconstruction accuracy compared to traditional methods.
• Demonstrates significantly reduced running time for image reconstruction.
• Effective in handling the complexity of hyperspectral DOT data.

Conclusions:

• The developed algorithm offers a computationally efficient and accurate solution for hyDOT image reconstruction.
• This advancement can improve the clinical applicability of DOT imaging.
• Highlights the benefit of MMV approaches and sparsity exploitation in complex imaging problems.