16 December 2024 in Neuron
18 May 2024 in Nature Communications
Cross-modal analysis of the same whole brain is an ideal strategy to uncover brain function and dysfunction. However, it remains challenging due to the slow speed and destructiveness of traditional whole-brain optical imaging techniques. Here we develop a new platform, termed Photoacoustic Tomography with Temporal Encoding Reconstruction (PATTERN), for non-destructive, high-speed, 3D imaging of ex vivo rodent, ferret, and non-human primate brains. Using an optimally designed image acquisition scheme and an accompanying machine-learning algorithm, PATTERN extracts signals of genetically-encoded probes from photobleaching-based temporal modulation and enables reliable visualization of neural projection in the whole central nervous system with 3D isotropic resolution. Without structural and biological perturbation to the sample, PATTERN can be combined with other whole-brain imaging modalities to acquire the whole-brain image with both high resolution and morphological fidelity. Furthermore, cross-modal transcriptome analysis of an individual brain is achieved by PATTERN imaging. Together, PATTERN provides a compatible and versatile strategy for brain-wide cross-modal analysis at the individual level.
30 May 2025 in IEEE Transactions on Medical Imaging
Anisotropy in imaging systems often results in directional degradation, impairing image quality and complicating subsequent analyses. While multiangle imaging has proven effective in mitigating these effects, it introduces challenges such as extended imaging times and increased excitation doses. To address these limitations in Photoacoustic Tomography (PAT), we propose a novel approach—Diffusion-based Sparse Tomographic Angular Recovery (D-STAR). D-STAR significantly reduces the number of required angles for high-resolution PAT while maintaining image quality comparable to full tomographic angular imaging. By training a diffusion model on a custom 3D PAT dataset, we optimize the balance between spatial and temporal resolutions, signal-to-noise ratio (SNR), and laser exposure. Our experiments with excised brain and vessel phantoms demonstrate that D-STAR produces high-fidelity images suitable for both structural and molecular imaging. This method outperforms existing approaches in static structural recovery and quantitative data extraction, offering substantial improvements in imaging quality, particularly in resolution and contrast. Furthermore, D-STAR enhances flexibility in imaging system design, reducing the need for hardware upgrades while improving temporal resolution and minimizing laser exposure.
11 November 2022 in Science Advances
Achieving optimal behavior requires animals to flexibly retrieve prior knowledge. Here, we show that adult newborn granule cells (anbGCs) mediate emotional state–dependent adaptability of memory retrieval. We find that acute social reward (aSR) enhances memory retrieval by increasing the reactivation of engram cells, while acute social stress (aSS) weakens retrieval and reduces the reactivation. Such bidirectional regulation relies on the activation of distinct populations of anbGCs by aSR and aSS, triggering opposing modifications of dDG activity, which is sufficient to regulate and predict the performance of memory retrieval. Concordantly, in emotional disorder models, aSR-dependent memory adaptability is impaired, while the effect of aSS remains intact. Together, our data revealed that anbGCs mediate adaptability of memory retrieval, allowing animals to flexibly retrieve memory according to the current emotional state, and suggested the essential roles of anbGCs in translating emotional information to the regulation of memory expression.
16 October 2024 in Cell Reports
Home information profoundly influences behavioral states in both humans and animals. However, how "home" is represented in the brain and its role in driving diverse related behaviors remain elusive. Here, we demonstrate that home bedding contains sufficient home information to modulate affective behaviors, including aversion responses, defensive aggression, and mating behaviors. These varied responses to home information are mediated by gama-aminobutyric acid (GABA)ergic neurons in the lateral hypothalamus (LHGABA). Inhibiting LHGABA abolishes, while activating mimics, the effects of home bedding on these behaviors across different contexts. Specifically, projections from LHGABA to the ventral tegmental area (VTA) mediate the relaxation of aversive emotion, while projections to the periaqueductal gray (PAG) initiate defensive concerns. Thus, our data suggest that home information in different contexts converges to activate distinct subgroups of the LHGABA, which, in turn, elicit appropriate affective behaviors in relieving aversion, fighting intruders, or enhancing mating through involving distinct downstream projections.
26 November 2025 in Scientific Reports
Computational pathology leverages advanced deep-learning techniques to analyze medical images with high resolution. However, a trade-off exists between model lightweight, interpretability, and task performance in such real-world scenarios. Knowledge distillation (KD) is widely applied to compress deep learning models while preserving high performance. Despite this, deep learning-based KD often lacks interpretable design, leading to inaccurate attention to images. Inspired by human vision processing, we developed a human vision attention-inspired knowledge distillation (HVisKD) strategy that captures local and global patch relations to construct differentiated features. We employed it in pathological analysis to balance the tradeoff. HVisKD improves performance across various lightweight models in segmentation tasks. More importantly, the attention map of HVisKD showed promoted consistency with human expert-labeled segmentation. Furthermore, we examined HVisKD in a real-world intraoperative pathological diagnosis scenario and achieved an interpretable and fast analysis. Together, HVisKD offers a lightweight and interpretable strategy for computational pathology, aligning deep learning with brain-like information processing for more dependable output.