Bioluminescence Tomography Based on One-Dimensional Convolutional Neural Networks

Abstract

Bioluminescent tomography (BLT) has increasingly important applications in preclinical studies. However, the simplified photon propagation model and the inherent ill-posedness of the inverse problem limit the quality of BLT reconstruction. In order to improve the reconstruction accuracy of positioning and reconstruction efficiency, this paper presents a deep-learning optical reconstruction method based on one-dimensional convolutional neural networks (1DCNN). The nonlinear mapping relationship between the surface photon flux density and the distribution of the internal bioluminescence sources is directly established, which fundamentally avoids solving the ill-posed inverse problem iteratively. Compared with the previous reconstruction method based on multilayer perceptron, the training parameters in the 1DCNN are greatly reduced and the learning efficiency of the model is improved. Simulations verify the superiority and stability of the 1DCNN method, and the in vivo experimental results further show the potential of the proposed method in practical applications.

Bioluminescence as a functional tool for visualizing and controlling neuronal activity in vivo

“… 154 , 155 The application of advanced models and deep neural networks may be a solution to increase the versatility for multisource reconstruction and have been shown to reduce the localization error down to . 155 , 156 …”

“…154,155 The application of advanced models and deep neural networks may be a solution to increase the versatility for multisource reconstruction and have been shown to reduce the localization error down to ≈1 mm. 155,156 5.3 Bioengineering: Autonomous Luciferase/Luciferin Systems and Faster Luciferases Since the first report of the small and bright Nluc 28 luciferase, more than 30 different Nluc derivatives have been established as fusion proteins with solute and metabolic sensors (Fig. 2).…”

“…However, due to the high-scattering properties of the brain and photon absorption by tissue chromophores such as melanin, hemoglobin, and other proteins, quantitative description of source intensity and localization precision is usually limited to ≈3 mm 154 , 155 . The application of advanced models and deep neural networks may be a solution to increase the versatility for multisource reconstruction and have been shown to reduce the localization error down to ≈1 mm 155 , 156 …”