Fluorescence separation based on the spatiotemporal Gaussian mixture model for dynamic fluorescence molecular tomography.

Dynamic fluorescence molecular tomography (DFMT) is a promising imaging method that can furnish three-dimensional information regarding the absorption, distribution, and excretion of fluorescent probes in organisms. Achieving precise dynamic fluorescence images is the linchpin for realizing high-resolution, high-sensitivity, and high-precision tomography. Traditional preprocessing methods for dynamic fluorescence images often face challenges due to the non-specificity of fluorescent probes in living organisms, requiring complex imaging systems or biological interventions. These methods can result in significant processing errors, negatively impacting the imaging accuracy of DFMT. In this study, we present, a novel, to the best of our knowledge, strategy based on the spatiotemporal Gaussian mixture model (STGMM) for the processing of dynamic fluorescence images. The STGMM is primarily divided into four components: dataset construction, time domain prior information, spatial Gaussian fitting with time prior, and fluorescence separation. Numerical simulations and in vivo experimental results demonstrate that our proposed method significantly enhances image processing speed and accuracy compared to existing methods, especially when faced with fluorescence interference from other organs. Our research contributes to substantial reductions in time and processing complexity, providing robust support for dynamic imaging applications.