Robust reconstruction of fluorescence molecular tomography with an optimized illumination pattern

Abstract

Fluorescence molecular tomography (FMT) is an emerging powerful tool for biomedical research. There are two factors that influence FMT reconstruction most effectively. The first one is the regularization techniques. Traditional methods such as Tikhonov regularization suffer from low resolution and poor signal to noise ratio. Therefore sparse regularization techniques have been introduced to improve the reconstruction quality. The second factor is the illumination pattern. A better illumination pattern ensures the quantity and quality of the information content of the data set thus leading to better reconstructions. In this work, we take advantage of the discrete formulation of the forward problem to give a rigorous definition of an illumination pattern as well as the admissible set of the patterns. We add restrictions in the admissible set as different types of regularizers to a discrepancy functional, generating another inverse problem with the illumination pattern as unknown. Both inverse problems of reconstructing the fluorescence distribution and finding the optimal illumination pattern are solved by fast efficient iterative algorithms. Numerical experiments have shown that with suitable choice of the regularization parameters the two-step approach converges to an optimal illumination pattern quickly and the reconstruction result is improved significantly, regardless of the initial illumination setting.

Non-invasive visualization of amyloid-beta deposits in Alzheimer amyloidosis mice using magnetic resonance imaging and fluorescence molecular tomography

“…First, the data acquisition procedure for a single FMT measurement is now 5-10 minutes considering the point-by-point laser scanning illumination, which limits the temporal resolution of FMT. The optimization of illumination patterns and usage of structured light illumination can significantly reduce the scanning time and improve the temporal resolution of FMT (Liu et al, 2020;Ren et al, 2020a;Streeter et al, 2019). Second, in the current setup, FMT-MRI was implemented in sequential mode, i.e., the mouse was measured by standalone FMT and an MRI scanner independently.…”

“…In the present work, we successfully applied our customized software platform to efficiently tackle these steps. Our STIFT platform has the potential to further enhance the automation of FMT data acquisition by setting optimal parameters, such as the illumination pattern (Liu et al, 2019).…”

Reconstruction of fluorophore absorption and fluorescence lifetime using early photon mesoscopic fluorescence molecular tomography: a phantom study

“…In the last decade, fluorescence tomography has begun to widely use the so-called compressed-sensing-like reconstruction algorithms, which minimize the Lp norm of the sought solution, where 0≤p≤1 (see, e.g., Refs. 1, 19–22, 29, 30, and 44–56). The most popular of them are iterative shrinkage thresholding algorithms 30 , 45 , 46 , 50 , 51 , 53 , 55 , 56 and algorithms with total variation (TV) regularization 29 , 30 , 44 , 47 …”

Uniqueness and numerical inversion in the time-domain fluorescence diffuse optical tomography

“…The FDOT based on RTE model or DE model has drawn extensive attention of researchers in recent years. For such inverse problems, three kinds of measurements, like time domain [18,22], continuous wave [13,25,29] and frequency domain [10,27] are used. For these optical imaging problems, the necessary regularization techniques such as Tikhonov regularization, sparse regularization methods and hybrid regularization methods have been introduced to overcome the ill-posedness of the inverse problems [11,15,25].…”

“…For such inverse problems, three kinds of measurements, like time domain [18,22], continuous wave [13,25,29] and frequency domain [10,27] are used. For these optical imaging problems, the necessary regularization techniques such as Tikhonov regularization, sparse regularization methods and hybrid regularization methods have been introduced to overcome the ill-posedness of the inverse problems [11,15,25]. Especially, the sparse regularization methods have additional advantages in promoting sparsity and higher spatial resolution for the cases that the target is relatively small compared to the background.…”