Shangbin Chen

Dendritic spines are key postsynaptic structures whose morphology, density, and remodeling provide structural indicators of synaptic plasticity and neurological disease. Longitudinal two-photon microscopy enables in vivo observation of dendritic-spine dynamics, yet practical z-stacks are frequently affected by out-of-focus fluorescence, axial distortion, low contrast, and inter-slice displacement, while high-quality pixel-level annotations remain scarce. We present an integrated computational workflow that combines percentile-based intensity normalization, sequential slice-wise affine registration, uncertainty-aware mean teacher (UA-MT) semantic segmentation, watershed-based instance separation, cross-plane linking, and interpretable morphometric classification. DeepD3 was reimplemented in PyTorch and trained using labeled and unlabeled images. A simplified ball-and-stick model estimated head length, head width, an attachment-distance surrogate for neck length, and total length; a KD-tree was used only to accelerate nearest-neighbor queries. Across three quantitatively evaluated stacks (two local and one external), the UA-MT-enhanced model achieved mean spine IoU and Dice values of 0.467 and 0.636, respectively, compared with 0.447 and 0.617 for the supervised baseline. The watershed-based downstream pipeline correctly classified 99 of 125 retained objects (79.2%). Given the small evaluation set, these results are interpreted as feasibility evidence rather than broad cross-dataset validation. After parameter configuration and model training, the accompanying aSpineGUI executes the trained workflow and produces segmentation masks, morphometric measurements, and class summaries; a six-day example illustrates potential use in repeated longitudinal analysis without constituting validated identity-level tracking.

Alzheimer’s disease (AD) classification from structural magnetic resonance imaging (MRI) may benefit from weak supervision that uses clinically meaningful but imperfect supervisory signals. We evaluated a weakly supervised framework in which a multilayer perceptron (MLP) trained on age, sex, and Mini-Mental State Examination (MMSE) scores generated clinical pseudo-labels to initialize a patch-based fully convolutional network (FCN). For 260 Alzheimer’s Disease Neuroimaging Initiative (ADNI) training participants, subsequent refinement combined 80% of the preceding MRI-model probability with 20% of the participant’s ground-truth diagnostic label. This design preserves a dominant pseudo-label/self-training component while using partial diagnostic guidance to stabilize refinement. The FCN generated whole-brain probability maps, and selected voxel probabilities were classified by a second MLP. The framework was developed using ADNI (n = 417). Using ADNI validation data only, iteration 3 and a classification threshold of 0.5 were selected and then applied unchanged to the held-out ADNI test set and the external AIBL (n = 182), FHS (n = 102), and NACC (n = 265) cohorts. The selected model achieved F1 scores of 0.853 in ADNI, 0.707 in AIBL, 0.765 in FHS, and 0.807 in NACC. These results support the feasibility and cross-cohort transferability of clinical pseudo-label-based weak supervision for MRI classification. The framework is not intended to be label-free; rather, it provides a transparent strategy for integrating imperfect clinical pseudo-labels with partially weighted diagnostic guidance during training.

Alzheimer’s disease (AD), the leading cause of dementia worldwide, necessitates early and accurate diagnosis to mitigate its progressive impact. Current diagnostic approaches rely heavily on costly and time-consuming expert evaluations, often lacking sufficient sensitivity and specificity. To address these limitations, we propose a novel weakly supervised deep learning framework that integrates fully convolutional networks and multilayer perceptrons for automated AD classification using multimodal MRI and clinical data. Our framework leverages pseudo-labels derived from clinical information to train the model in an iterative weakly supervised manner, significantly reducing the dependency on fully annotated datasets. The model was trained on the ADNI dataset and externally validated on three independent cohorts: AIBL, FHS, and NACC. The proposed method achieved promising performance with an average F1-score of 85.3% and an AUC of 0.808, comparable to fully supervised methods and neurologist evaluations. The model demonstrated robust generalization across diverse populations and imaging protocols. By integrating this framework into routine brain MRI examinations, our approach enables cost-effective, opportunistic screening for early AD risk detection without requiring dedicated diagnostic protocols. This strategy holds potential for reducing healthcare disparities and facilitating timely intervention, particularly in resource-limited settings.

Alzheimer’s disease is a formidable health challenge due to lack of effective therapeutic modalities. The excitation/inhibition imbalance in the early stage of Alzheimer’s disease can be potentially considered as a central link between structural brain pathology and cognitive dysfunction. However, the role and effects of reactive astrocytes in the neuronal excitability in early Alzheimer’s disease remain unclear. Here, we present a tripartite synaptic model integrating the interactions between neurons and astrocytes than can clarify the role of astrocytes in the regulation of excitation/inhibition. Our model integrates the cation channel transient receptor potential ankyrin 1, whose activation triggers calcium influx, thereby enhancing the fidelity of astrocyte calcium dynamics. Constrained by physiological data, we demonstrate that amyloid-β can activate astrocytes to release glial neurotransmitters, thereby mediating the hyperexcitability of nearby neurons. We also investigate the astrocyte-mediated symbiosis of two neurotransmitters, glutamate and gamma-aminobutyric acid, at the glutamatergic synapse in the context of Alzheimer’s disease, to predict the inhibitory compensatory response to excitotoxicity. During excitotoxicity, astrocytes can use the coupling of the excitatory amino acid transporter and gamma-aminobutyric acid transporter to control the concentration ratio of glutamate and gamma-aminobutyric acid in the synaptic cleft, and may convert both through the intracellular gamma-aminobutyric acid synthesis pathway. Our findings reveal that the coding efficiency of neurons diminished as the effects of amyloid-β deepened, establishing a direct link between the pathological features of Alzheimer’s disease and cognitive dysfunction. These simulations suggest that astrocytes play a critical role in regulating the neuronal excitation/inhibition balance in the early stage of Alzheimer’s disease, thereby influencing the subsequent processes of information transmission, learning, and memory. The pathways characterized by our model present potential therapeutic targets for Alzheimer’s disease.

Photon-counting computed tomography has demonstrated significant advancements in recent years; however, micro photon-counting CT (Micro-PCCT) systems are still limited by pixel-wise detector response nonuniformity, which degrades measurement uniformity across detector pixels and commonly produces ring artifacts in reconstructed images. Existing calibration methods exhibit limited generalizability in complex multi-material scenarios, such as contrast-enhanced imaging. This study introduces a Signal-to-Nonuniformity Error Polynomial Calibration (STEPC) framework based on measurement nonuniformity error modeling to address this issue. STEPC first fits multi-energy projections using a 2D polynomial surface to generate ideal references, then applies a nonlinear multi-energy polynomial model to predict and correct pixel-wise nonuniformity errors. The model is calibrated using homogeneous slab phantoms of different materials, including PMMA, aluminum, and iodinated contrast agents, enabling correction for both non-contrast and contrast-enhanced imaging. Experiments were performed on a custom Micro-PCCT system with phantoms and mouse. Correction performance of STEPC was evaluated using the mean local standard deviation (MLSD) in the projection domain and the ring artifact deviation (RAD) on the reconstructed images. Compared with existing methods, STEPC achieved an average MLSD reduction of at least 21.58% and reduced RAD by at least 14.18%, consistently yielding the best performance in both non-contrast and contrast-enhanced scenarios. Furthermore, STEPC can be readily extended to compensate for beam hardening effects within the same calibration framework. Quantitative material decomposition results indicate that the proposed method preserves measurement accuracy across different basis materials. These results demonstrate that STEPC provides a robust and practical calibration framework that significantly improves detector measurement uniformity in Micro-PCCT systems for multi-material spectral imaging.

Objective: To measure the steady-state corneal biomechanical properties related to postoperative corneal ectasia, keratoconus, glaucoma and other ophthalmic diseases, we propose a novel in vivo measurement method. Methods: By precisely manipulating ambient negative pressure via a suction device to achieve controlled in vivo corneal inflation, we analyzed the coupling relationship between corneal deformation response and negative pressure loading. The displacement corresponding to the corneal initial configuration under 0 mmHg was extrapolated. The inverse finite element analysis (FEA) technique was employed to calculate the segmental Young's modulus of the cornea under different steady-state intraocular pressure (IOP) conditions. Results: In vivo experiments on 3-month-old rabbit eyes demonstrated a Young's modulus of 0.2–0.3 MPa at physiological IOP (10–12 mmHg), along with an increasing trend of corneal stiffness across a pressure range of 0–85 mmHg. Conclusion: The proposed non-invasive measurement method exhibits stability and consistency in parameter inversion under different IOPs, indicating its significant clinical potential. Significance: Providing a new approach for corneal biomechanical assessment is beneficial for the diagnosis and treatment of ophthalmic diseases.

SignificanceConfocal endomicroscopic image stitching can expand the field of view and improve examination efficiency. However, due to interference from the rolling shutter effect, traditional stitching methods may produce misalignments, leading to structural distortion and artifacts. Suppressing the rolling shutter effect in confocal endomicroscopic images can effectively enhance stitching quality.AimWe propose a Dual-Path Gaussian U-Net (DGU-Net)-based framework for confocal endomicroscopic image stitching. The parallel dual-encoder paths of DGU-Net extract Gaussian features and conventional features at different resolutions, respectively, achieving more precise gland segmentation masks. Based on these masks, we filter stable frames and optimize feature matching to effectively suppress rolling shutter interference and improve stitching quality.ApproachWe annotated a segmentation dataset comprising 80 rat confocal laser endomicroscopy (CLE) images to train the segmentation network and validated the frame selection method’s effectiveness in suppressing the rolling shutter effect on consecutively acquired rat CLE video sequences. The stitching results generated from the filtered stable image sequences were compared with conventional methods.ResultsExperimental results demonstrate that DGU-Net achieves superior performance with a Dice score of 85.17 on CLE datasets, significantly outperforming existing segmentation networks. Compared with Auto-Stitching, our method improves regional consistency across the panoramic image by eliminating artifacts caused by mismatches while delivering enhanced stitching accuracy and image quality.ConclusionsThe proposed method effectively accomplishes confocal image stitching tasks, significantly enhancing endomicroscopic examination efficiency and contributing to improved diagnostic outcomes.