Lung Computed Tomography (CT) images play an important role in the diagnosis and treatment of patients with COVID- 19. However, manually identifying lesions in CT images is very time-consuming and requires specialized medical knowledge. The accuracy and integrity of the existing models are still not ideal due to the reasons such as scattered distribution, small lesions and blurred boundaries. Therefore, we propose a segmentation model based on U-Net architecture by combining Mirror-symmetry Boundary Guided (MBG) module and then adding Spatial Attention Dilation (SAD) convolution Module, called MSU-Net. The SAD module uses spatial attention mechanism and dilated convolution and multi-scale feature extraction strategy to enhance the network’s ability to recognize small lesions. The MBG module further enhances the model’s ability to capture more complex context information and boundary details, making the model more robust and able to effectively deal with the challenges of fuzzy boundaries. The proposed method has shown superior performance in terms of the Dice coefficient, Hausdorff distance, and Sensitivity on the publicly available COVID-19 CT dataset.
SignificanceFluorescence molecular tomography (FMT) is a promising imaging modality, which has played a key role in disease progression and treatment response. However, the quality of FMT reconstruction is limited by the strong scattering and inadequate surface measurements, which makes it a highly ill-posed problem. Improving the quality of FMT reconstruction is crucial to meet the actual clinical application requirements.AimWe propose an algorithm, neighbor-based adaptive sparsity orthogonal least square (NASOLS), to improve the quality of FMT reconstruction.ApproachThe proposed NASOLS does not require sparsity prior information and is designed to efficiently establish a support set using a neighbor expansion strategy based on the orthogonal least squares algorithm. The performance of the algorithm was tested through numerical simulations, physical phantom experiments, and small animal experiments.ResultsThe results of the experiments demonstrated that the NASOLS significantly improves the reconstruction of images according to indicators, especially for double-target reconstruction.ConclusionNASOLS can recover the fluorescence target with a good location error according to simulation experiments, phantom experiments and small mice experiments. This method is suitable for sparsity target reconstruction, and it would be applied to early detection of tumors.
Bioluminescence tomography (BLT) is an effective noninvasive molecular imaging modality, it has shown great potential for studying and monitoring disease progression in pre-clinical imaging. As the BLT is an inherent highly ill-posed inverse problem, it is still a challenge to obtain an accurate reconstruction result. Some algorithms have been proposed to solve highly ill posedness of inverse problems. Nevertheless, Existing methods always need to consume large time or have low interpretability. Thus, in this paper, we proposed a novel model-driven deep learning network, which unfolding the Fast Iterative Shrinkage Thresholding Algorithm (FISTA) algorithm into a deep network, named FISTA-Net to overcome the above shortcoming. FISTA-Net is formed from three modules, gradient descent module, proximal mapping module and accelerate module. Key parameters of FISTA-Net including the gradient step size, thresholding value are learned from training data. The experimental results, evaluated both visually and quantitatively, show that the FISTA-Net can achieve a high-quality reconstruction result of BLT.
Bioluminescence tomography (BLT) reconstruction is an ill-posed problem. A class of strategy based on the permissible region (PR) reduces the ill-posed by reducing the space. However, in multi-objective reconstruction, the strategy is challenging to fit the sources of different positions. In this study, a subspace decision (SD) method is proposed, which transforms the traditional single permissible region into multiple spatially continuous subspaces by clustering, and performs spatial shrinkage optimization for each of them. In addition, a plug-and-play sliding single polyline module is introduced to analyze and cluster the reconstruction results each time to obtain the number and distribution of subspaces contained in the results. SD method does not rely on any specific reconstruction or clustering algorithm, so it has great flexibility. Experiment results show that the SD approach can more accurately obtain the spatial distribution information of different numbers of sources distributed in different locations and ensure the quality of multi-source BLT reconstruction. Keywords: Bioluminescence Tomography, Inverse Problem, Subspace, Clustering, Permissible Region.
Sparse regularization methods have been widely used in fluorescence molecular tomography (FMT) for stable three-dimensional reconstruction. Generally, ℓ1-regularization-based methods allow for utilizing the sparsity nature of the target distribution. However, in addition to sparsity, the spatial structure information should be exploited as well. A joint ℓ1 and Laplacian manifold regularization model is proposed to improve the reconstruction performance, and two algorithms (with and without Barzilai–Borwein strategy) are presented to solve the regularization model. Numerical studies and in vivo experiment demonstrate that the proposed Gradient projection-resolved Laplacian manifold regularization method for the joint model performed better than the comparative algorithm for ℓ1 minimization method in both spatial aggregation and location accuracy.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.