15 November 2021 in ACS Photonics
Understanding the complexity of animals’ behaviors and intellect relies on decoding the underlying structure and wiring diagram of their brains. Various imaging techniques have been developed to tackle this challenge, yet they still have limitations in mapping all cells in a brain region at mesoscale level. Here we propose an optical multilayer interference tomography (OMLIT) method with serial sectioning capability for mesoscale neuroanatomy. With OMLIT, we can distinguish and reconstruct all cell bodies, large branches of dendrites and myelinated axons, and the vasculature in the region of interest, enabling fast mapping of cell density and classification of certain cell types. Furthermore, using the same form of ultrathin sample section library, our method offers seamless compatibility with tape-based serial scanning electron microscopy, providing a correlative approach to reconstruct both mesoscale and microscale brain atlases in the same tissue.
12 February 2025 in IEEE Transactions on Radiation and Plasma Medical Sciences
Low-dose computed tomography (LDCT) denoising methods based on supervised learning with labeled simulation data have made significant progress. However, these methods usually struggle to directly process unlabeled LDCT images due to inherent biases. While unsupervised methods have been explored to utilize unlabeled LDCT images, they typically involve complex network structures with limited denoising performance. To address these issues, we propose a self-adaptive weight embedded lightweight semi-supervised network (SWELNet) for unlabeled LDCT image denoising, which integrates supervised and unsupervised learning in a lightweight architecture. Unlike other semi-supervised algorithms that only consider the correlations between labeled simulation data and unlabeled data, the proposed SWELNet not only takes into account correlations but also the differences between data. There are three modules in the proposed network respectively for feature extraction, refinement and self-adaptive weight. Specially, the multi-scale convolution feature extraction module (MCFEM) and recursive module (RECM) extract and refine common representations from labeled simulation and unlabeled data with the well-designed. After that, the softmax feature fusion module (SFFM) with self-adaptive weighted learning for forming different feature spaces for two types of data. Extensive experiments using one simulation and two unlabeled datasets demonstrate that the proposed SWELNet outperforms several state-of-the-art baseline network methods in terms of robustness and generalization, as well as inference efficiency. The code is available at https://github.com/nightastars/SWELNet-main.git.
19 January 2026 in IEEE Transactions on Circuits and Systems for Video Technology
Sparse-view computed tomography (SVCT) offers the advantages of accelerated scanning and reduced X-ray radiation dose in different clinical applications. However, it faces a challenge due to incomplete data acquisition, resulting in streak artifacts in the analytically reconstructed CT images. Utilizing self-supervised learning, implicit neural representation (INR) recently has shown great promise in addressing inverse problems such as SVCT reconstruction. Nonetheless, given that the input of original INR only contains coordinate information, it is limited to represent one SVCT instance at a time, and its performance significantly declines when performing cross-instance reconstruction. In this study, we propose a novel self-supervised framework named VFMINR, which leverages generalizable representations extracted from the visual foundation models (VFMs) to tackle the cross-instance reconstruction issue of INR. Specifically, VFMINR first utilizes VFMs to effectively capture the spatial and frequency domain representations of sinograms, and then a fusion module is applied to fuse two domain features into complementary representations. This combination maximizes the utilization of local detail information from the spatial domain and the global structural information from the frequency domain. Subsequently, an adaptive cell decoding strategy is designed to map representations into variable resolution hybrid feature grids, which are integrated into the learning of the INR to enhance its generalizability for different SVCT instances. The VFMs and VFMINR are trained by using only SV sinogram data, and extensive results confirm that the proposed method can effectively handle the generalization problem of INR, while achieving superior performance in image fidelity and artifact suppression. The code is available at: https://github.com/nightastars/VFMINR-main.
1 January 2026 in IEEE Transactions on Computational Imaging