arXiv subjects

Zhu Chen

Publications and source records attributed to Zhu Chen.

At least 19 recordsLinked to original sources

arXiv2026

Semantic-Guided Multimodal Preprocessing for Vision Transformer-Based Clear Cell Renal Cell Carcinoma Grading

Clear cell renal cell carcinoma (CCRCC) grading is essential for treatment planning, yet existing approaches either analyze patch-level images directly or focus solely on nuclei-level classification, without linking to final tumor grading. We propose a semantic-guided multimodal preprocessing method that integrates nuclei classification maps from existing pre-trained models with RGB histopathology images for Vision Transformer (ViT)-based CCRCC grading. Our approach employs classification map channel concatenation and multiplicative modulation, with optimized overlays to leverage nuclei grading information, while preserving RGB textural features. Evaluation of multiple preprocessing strategies demonstrates that semantic-guided enhancement achieves 0.916 balanced accuracy, outperforming RGB-only baseline (0.707) and max-voting aggregation from prior studies (0.427). Sensitivity analysis reveals that this 21 percentage point improvement over baseline persists even under simulated perturbation at rates matching current state-of-the-art nuclei classification model error thresholds, suggesting both effective semantic utilization and practical robustness. These findings show that preprocessing-based multimodal fusion can leverage the diagnostic potential of existing imperfect nuclei classifiers, effectively bridging previously isolated fine-grained nuclear-level analysis with coarse-grained ViT-based patch classification. Per-class recall was consistent across grades (0.93, 0.91, 0.91), indicating that gains are not concentrated in the majority class. Because the sensitivity analysis perturbs ground-truth maps rather than predictions from an actual nuclei model, this result characterizes robustness under simulated error rather than deployment with a real upstream model, which remains for future work.

cs.CV↗

arXiv2026

STEAM: A Spatio-TEmporal Alignment Mixture-of-Experts Model with Hierarchical Pre-training for EEG Decoding

Brain-computer interfaces (BCIs) have been widely used in motor rehabilitation, disease diagnosis, and other neural engineering scenarios. However, conventional neural signal decoding algorithms often suffer from limited generalizability and high adaptation costs, motivating recent interest in BCI foundation models. Existing approaches still struggle to jointly achieve general transferability, accurate decoding, and efficient downstream adaptation. We present STEAM, a hierarchical transfer framework that reconciles general-purpose representation learning with paradigm-specific specialization in EEG foundation models. The framework is instantiated as a dual-branch spatio-temporal encoder in which a shared soft mixture-of-experts (SSMoE) module aligns the spatial and temporal branches, allowing complementary representations to exchange information through a compact set of soft slots. Across seven downstream datasets and fourteen evaluation settings, STEAM attains the best average rank among the compared methods at a competitive inference cost measured in FLOPs. Building upon the Stage-I general initialization, the hierarchical pre-training strategy further specializes the model to a target paradigm without retraining from scratch, yielding consistent gains in paradigm-specific decoding accuracy.

cs.CV↗

arXiv2026

Comparative analysis of missing data imputation methods for CSST survey: Impact on photometric redshift estimation performance

Improving the accuracy of photometric redshifts (photo-$z$) is essential for reliable statistical studies of cosmology and galaxy evolution. However, missing photometric bands are a common observational challenge that can significantly degrade photo-$z$ estimation accuracy. In this work, we present a systematic evaluation of data imputation methods aimed at improving photo-$z$ performance. We benchmark a range of representative machine learning (ML) and deep learning (DL) architectures, identifying k-nearest neighbors (KNN) and the attention-based SAITS model as the leading performers. These models are then applied to China Space Station Survey Telescope (CSST) mock data to assess their performance under realistic observational conditions. Our results show that KNN yields the highest accuracy under idealized missing completely at random (MCAR) conditions with complete training sets, whereas robustness tests reveal that SAITS significantly outperforms KNN when training data is incomplete or when applied to realistic mixed-mechanism scenarios. We find that domain consistency between training and testing missingness patterns is a prerequisite for optimal performance, highlighting the risks of domain shift in supervised regression tasks. Furthermore, our analysis demonstrates that while general imputation models are highly effective for MCAR and missing at random (MAR) data, they are detrimental when applied to missing not at random (MNAR) data arising from flux limits, as statistical models fail to capture the physical information inherent in these non-detections. Consequently, we advocate for more sophisticated architectures capable of disentangling stochastic missingness from physical non-detections to address these distinct mechanisms individually.

astro-ph.GA↗

arXiv2026

EEG-FM-Compass: Progress, Benchmarking, and Future Directions for EEG Foundation Models

Electroencephalography (EEG) foundation models (FMs) have recently emerged as a promising paradigm for brain-computer interfaces, aiming to learn transferable neural representations from large-scale heterogeneous recordings. Despite rapid progress, a fair and comprehensive comparison of existing EEG FMs is still lacking, owing to inconsistent pre-training objectives, preprocessing choices, and downstream evaluation protocols. To fill this gap, we present EEG-FM-Compass. We first review 55 representative models and organize their design choices into a unified taxonomic framework including data standardization, model architectures, and self-supervised pre-training strategies. We then evaluate 12 open source FMs and competitive specialist baselines across 13 EEG datasets spanning nine brain-computer interface paradigms. Emphasizing real-world deployments, we consider both cross-subject generalization under a leave-one-subject-out protocol and rapid calibration under a within-subject few-shot setting. We further compare full-parameter fine-tuning with linear probing to assess the transferability of pre-trained representations, and examine the relationship between model scale and downstream performance. Our results indicate that: 1) linear probing is frequently insufficient; 2) specialist models trained from scratch remain competitive across many tasks; and 3) larger FMs do not necessarily yield better generalization performance under current data regimes and training practices.

cs.LG↗

arXiv2025

LSTM-MDNz: Estimating Quasar Photometric Redshifts with an LSTM-Augmented Mixture Density Network

Quasar photometric redshifts are essential for studying cosmology and large-scale structures. However, their complex spectral energy distributions cause significant redshift-color degeneracy, limiting the accuracy of traditional methods. To overcome this, we introduce LSTM-MDNz, a novel end-to-end deep learning model combining long short-term memory networks (LSTM) with mixture density networks (MDN). The model directly uses multi-band photometric fluxes and associated errors as wavelength-ordered sequential inputs, eliminating the need for manual feature engineering while enabling simultaneous point estimation and probability distribution function (PDF) prediction of quasar redshifts. We integrate data from four major sky surveys-SDSS, DESI-LS, WISE, and GALEX-to assemble a sample of over 550,000 spectroscopically confirmed quasars ($0 \leq z_{\mathrm{spec}} \leq 5$) across 14 ultraviolet to infrared bands for model training and testing. Experimental results show that using all 14 bands yields optimal performance, with a normalized median absolute deviation ($\sigma_{\mathrm{NMAD}}$) of 0.037 and an outlier rate ($f_{\mathrm{out}}$) of 3.5\% on the test set. These values represent reductions of 29\% and 56\%, respectively, compared to the commonly adopted SDSS+WISE band set. Probability integral transform ($\mathrm{PIT}$) and continuous ranked probability score ($\mathrm{CRPS}$) analyses confirm that the predicted PDFs align closely with the true redshift distribution. Band-ablation experiments further highlight the essential role of ultraviolet and infrared data in alleviating color degeneracy and reducing systematic bias. This study demonstrates the effectiveness of multi-band fusion in improving quasar photo-z accuracy and offers a ready-to-use estimation framework for future surveys like LSST, CSST, and Euclid.

astro-ph.GA↗

arXiv2025

BALNet: Deep Learning-Based Detection and Measurement of Broad Absorption Lines in Quasar Spectra

Broad absorption line (BAL) quasars serve as critical probes for understanding active galactic nucleus (AGN) outflows, black hole accretion, and cosmic evolution. To address the limitations of manual classification in large-scale spectroscopic surveys - where the number of quasar spectra is growing exponentially - we propose BALNet, a deep learning approach consisting of a one-dimensional convolutional neural network (1D-CNN) and bidirectional long short-term memory (Bi-LSTM) networks to automatically detect BAL troughs in quasar spectra. BALNet enables both the identification of BAL quasars and the measurement of their BAL troughs. We construct a simulated dataset for training and testing by combining non-BAL quasar spectra and BAL troughs, both derived from SDSS DR16 observations. Experimental results in the testing set show that: (1) BAL trough detection achieves 83.0% completeness, 90.7% purity, and an F1-score of 86.7%; (2) BAL quasar classification achieves 90.8% completeness and 94.4% purity; (3) the predicted BAL velocities agree closely with simulated ground truth labels, confirming BALNet's robustness and accuracy. When applied to the SDSS DR16 data within the redshift range 1.5<z<5.7, at least one BAL trough is detected in 20.4% of spectra. Notably, more than a quarter of these are newly identified sources with significant absorption, 8.8% correspond to redshifted systems, and some narrow/weak absorption features were missed. BALNet greatly improves the efficiency of large-scale BAL trough detection and enables more effective scientific analysis of quasar spectra.

astro-ph.GA↗

arXiv2025

Segment Anything for Cell Tracking

Tracking cells and detecting mitotic events in time-lapse microscopy image sequences is a crucial task in biomedical research. However, it remains highly challenging due to dividing objects, low signal-tonoise ratios, indistinct boundaries, dense clusters, and the visually similar appearance of individual cells. Existing deep learning-based methods rely on manually labeled datasets for training, which is both costly and time-consuming. Moreover, their generalizability to unseen datasets remains limited due to the vast diversity of microscopy data. To overcome these limitations, we propose a zero-shot cell tracking framework by integrating Segment Anything 2 (SAM2), a large foundation model designed for general image and video segmentation, into the tracking pipeline. As a fully-unsupervised approach, our method does not depend on or inherit biases from any specific training dataset, allowing it to generalize across diverse microscopy datasets without finetuning. Our approach achieves competitive accuracy in both 2D and large-scale 3D time-lapse microscopy videos while eliminating the need for dataset-specific adaptation.

cs.CV↗

arXiv2025

MIRepNet: A Pipeline and Foundation Model for EEG-Based Motor Imagery Classification

Brain-computer interfaces (BCIs) enable direct communication between the brain and external devices. Recent EEG foundation models aim to learn generalized representations across diverse BCI paradigms. However, these approaches overlook fundamental paradigm-specific neurophysiological distinctions, limiting their generalization ability. Importantly, in practical BCI deployments, the specific paradigm such as motor imagery (MI) for stroke rehabilitation or assistive robotics, is generally determined prior to data acquisition. This paper proposes MIRepNet, the first EEG foundation model tailored for the MI paradigm. MIRepNet comprises a high-quality EEG preprocessing pipeline incorporating a neurophysiologically-informed channel template, adaptable to EEG headsets with arbitrary electrode configurations. Furthermore, we introduce a hybrid pretraining strategy that combines self-supervised masked token reconstruction and supervised MI classification, facilitating rapid adaptation and accurate decoding on novel downstream MI tasks with fewer than 30 trials per class. Extensive evaluations across five public MI datasets demonstrated that MIRepNet consistently achieved state-of-the-art performance, significantly outperforming both specialized and generalized EEG models. Our code will be available on GitHub\footnote{https://github.com/staraink/MIRepNet}.

cs.CV↗

arXiv2025

CLEAN-MI: A Scalable and Efficient Pipeline for Constructing High-Quality Neurodata in Motor Imagery Paradigm

The construction of large-scale, high-quality datasets is a fundamental prerequisite for developing robust and generalizable foundation models in motor imagery (MI)-based brain-computer interfaces (BCIs). However, EEG signals collected from different subjects and devices are often plagued by low signal-to-noise ratio, heterogeneity in electrode configurations, and substantial inter-subject variability, posing significant challenges for effective model training. In this paper, we propose CLEAN-MI, a scalable and systematic data construction pipeline for constructing large-scale, efficient, and accurate neurodata in the MI paradigm. CLEAN-MI integrates frequency band filtering, channel template selection, subject screening, and marginal distribution alignment to systematically filter out irrelevant or low-quality data and standardize multi-source EEG datasets. We demonstrate the effectiveness of CLEAN-MI on multiple public MI datasets, achieving consistent improvements in data quality and classification performance.

cs.CE↗

arXiv2025

Meta-Calibration of the Cosmic Magnification Coefficient: Toward Unbiased Weak Lensing Reconstruction by Counting Galaxies

Weak lensing alters galaxy sizes and fluxes, influencing the clustering patterns of galaxies through cosmic magnification. This effect enables the reconstruction of weak lensing convergence $\hat{\kappa}$ maps for DES and DECaLS by linearly combining galaxy overdensities across magnitude bins in the $g$, $r$, and $z$ photometry bands \citep{Qin+,Qin2+}. In this study, we enhance the lensing reconstruction method by addressing biases in the magnification coefficient estimation, which arise from incomplete consideration of selection effects, especially those induced by photometric redshift (photo-$z$) selection. Using a Random Forest-based photo-$z$ estimation for DECaLS and DES galaxies, we quantify the impact of photo-$z$ induced selection on magnification coefficient estimation. Our results show that neglecting photo-$z$ selection introduces significant biases in the magnification coefficient, leading to deviations in the reconstructed convergence map amplitude $A$, with values ranging from 0.4 to 3.5 depending on the survey, redshift, and magnitude cuts. By incorporating an improved magnification coefficient estimation that accounts for photo-$z$ selection, these biases are significantly reduced, with $A$ converging to $\sim 1$ as the magnitude cuts approach optimal values. This improvement is consistently observed across DES and DECaLS datasets and redshift bins, despite differences in survey strategies and depths. Our findings highlight the importance of addressing photo-$z$ induced selection to achieve unbiased weak lensing reconstructions and accurate cosmic magnification measurements.

astro-ph.CO↗

arXiv2025

Improving Photometric Redshift Estimation for CSST Mock Catalog Using SED Templates Calibrated with Perturbation Algorithm

Photometric redshifts of galaxies obtained by multi-wavelength data are widely used in photometric surveys because of its high efficiency. Although various methods have been developed, template fitting is still adopted as one of the most popular approaches. Its accuracy strongly depends on the quality of the Spectral Energy Distribution (SED) templates, which can be calibrated using broadband photometric data from galaxies with known spectroscopic redshifts. Such calibration is expected to improve photometric redshift accuracy, as the calibrated templates will align with observed photometric data more closely. The upcoming China Space Station Survey Telescope (CSST) is one of the Stage IV surveys, which aiming for high precision cosmological studies. To improve the accuracy of photometric redshift estimation for CSST, we calibrated the CWW+KIN templates using a perturbation algorithm with broadband photometric data from the CSST mock catalog. This calibration used a training set consisting of approximately 4,500 galaxies, which is 10% of the total galaxy sample. The outlier fraction and scatter of the photometric redshifts derived from the calibrated templates are 2.55% and 0.036, respectively. Compared to the CWW+KIN templates, these values are reduced by 34% and 23%, respectively. This demonstrates that SED templates calibrated with a small training set can effectively optimize photometric redshift accuracy for future large-scale surveys like CSST, especially with limited spectral training data.

astro-ph.CO↗

arXiv2025

Galaxy Morphology Classification via Deep Semi-Supervised Learning with Limited Labeled Data

Galaxy morphology classification plays a crucial role in understanding the structure and evolution of the universe. With galaxy observation data growing exponentially, machine learning has become a core technology for this classification task. However, traditional machine learning methods predominantly rely on supervised learning frameworks, and their dependence on large of labeled samples limits practical applications. To address this challenge, we propose an innovative hybrid semi-supervised model, GC-SWGAN, designed to tackle galaxy morphology classification under conditions of limited labeled data. This model integrates semi-supervised generative adversarial networks (SGAN) with Wasserstein GAN with gradient penalty (WGAN-GP), establishing a multi-task learning framework. Within this framework, the discriminator and classifier are designed independently while sharing part of the architecture. By collaborating with the generator, the model significantly enhances both classification performance and sample generation capabilities, while also improving convergence and stability during training. Experimental results demonstrate that, on the Galaxy10 DECals dataset, GC-SWGAN achieves comparable or even superior classification accuracy (exceeding 75%) using only one-fifth of the labeled samples typically required by conventional fully supervised methods. Under identical labeled conditions, the model displays excellent generalization performance, attaining approximately 84% classification accuracy. Notably, in extreme scenarios where only 10\% of the data is labeled, GC-SWGAN still achieves high classification accuracy (over 68%), fully demonstrating its stability and effectiveness in low-labeled data environments. Furthermore, galaxy images generated by GC-SWGAN are visually similar to real samples.

astro-ph.GA↗

arXiv2025

CellStyle: Improved Zero-Shot Cell Segmentation via Style Transfer

Cell microscopy data are abundant; however, corresponding segmentation annotations remain scarce. Moreover, variations in cell types, imaging devices, and staining techniques introduce significant domain gaps between datasets. As a result, even large, pretrained segmentation models trained on diverse datasets (source datasets) struggle to generalize to unseen datasets (target datasets). To overcome this generalization problem, we propose CellStyle, which improves the segmentation quality of such models without requiring labels for the target dataset, thereby enabling zero-shot adaptation. CellStyle transfers the attributes of an unannotated target dataset, such as texture, color, and noise, to the annotated source dataset. This transfer is performed while preserving the cell shapes of the source images, ensuring that the existing source annotations can still be used while maintaining the visual characteristics of the target dataset. The styled synthetic images with the existing annotations enable the finetuning of a generalist segmentation model for application to the unannotated target data. We demonstrate that CellStyle significantly improves zero-shot cell segmentation performance across diverse datasets by finetuning multiple segmentation models on the style-transferred data. The code will be made publicly available.

cs.LG↗

arXiv2025

Unsupervised Learning for Feature Extraction and Temporal Alignment of 3D+t Point Clouds of Zebrafish Embryos

Zebrafish are widely used in biomedical research and developmental stages of their embryos often need to be synchronized for further analysis. We present an unsupervised approach to extract descriptive features from 3D+t point clouds of zebrafish embryos and subsequently use those features to temporally align corresponding developmental stages. An autoencoder architecture is proposed to learn a descriptive representation of the point clouds and we designed a deep regression network for their temporal alignment. We achieve a high alignment accuracy with an average mismatch of only 3.83 minutes over an experimental duration of 5.3 hours. As a fully-unsupervised approach, there is no manual labeling effort required and unlike manual analyses the method easily scales. Besides, the alignment without human annotation of the data also avoids any influence caused by subjective bias.

cs.CV↗

arXiv2025

No Free Lunch in Annotation either: An objective evaluation of foundation models for streamlining annotation in animal tracking

We analyze the capabilities of foundation models addressing the tedious task of generating annotations for animal tracking. Annotating a large amount of data is vital and can be a make-or-break factor for the robustness of a tracking model. Robustness is particularly crucial in animal tracking, as accurate tracking over long time horizons is essential for capturing the behavior of animals. However, generating additional annotations using foundation models can be counterproductive, as the quality of the annotations is just as important. Poorly annotated data can introduce noise and inaccuracies, ultimately compromising the performance and accuracy of the trained model. Over-reliance on automated annotations without ensuring precision can lead to diminished results, making careful oversight and quality control essential in the annotation process. Ultimately, we demonstrate that a thoughtful combination of automated annotations and manually annotated data is a valuable strategy, yielding an IDF1 score of 80.8 against blind usage of SAM2 video with an IDF1 score of 65.6.

cs.CV↗

arXiv2024

From Galaxy Zoo DECaLS to BASS/MzLS: detailed galaxy morphology classification with unsupervised domain adaption

The DESI Legacy Imaging Surveys (DESI-LIS) comprise three distinct surveys: the Dark Energy Camera Legacy Survey (DECaLS), the Beijing-Arizona Sky Survey (BASS), and the Mayall z-band Legacy Survey (MzLS). The citizen science project Galaxy Zoo DECaLS 5 (GZD-5) has provided extensive and detailed morphology labels for a sample of 253,287 galaxies within the DECaLS survey. This dataset has been foundational for numerous deep learning-based galaxy morphology classification studies. However, due to differences in signal-to-noise ratios and resolutions between the DECaLS images and those from BASS and MzLS (collectively referred to as BMz), a neural network trained on DECaLS images cannot be directly applied to BMz images due to distributional mismatch. In this study, we explore an unsupervised domain adaptation (UDA) method that fine-tunes a source domain model trained on DECaLS images with GZD-5 labels to BMz images, aiming to reduce bias in galaxy morphology classification within the BMz survey. Our source domain model, used as a starting point for UDA, achieves performance on the DECaLS galaxies' validation set comparable to the results of related works. For BMz galaxies, the fine-tuned target domain model significantly improves performance compared to the direct application of the source domain model, reaching a level comparable to that of the source domain. We also release a catalogue of detailed morphology classifications for 248,088 galaxies within the BMz survey, accompanied by usage recommendations.

astro-ph.GA↗

arXiv2024

Cross-Survey Image Transformation: Enhancing SDSS and DECaLS Images to Near-HSC Quality for Advanced Astronomical Analysis

This study focuses on transforming galaxy images between astronomical surveys, specifically enhancing images from the Sloan Digital Sky Survey (SDSS) and the Dark Energy Camera Legacy Survey (DECaLS) to achieve quality comparable to the Hyper Suprime-Cam survey (HSC). We proposed a hybrid model called Pix2WGAN, which integrates the pix2pix framework with the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) to convert low-quality observational images into high-quality counterparts. Our model successfully transformed DECaLS images into pseudo-HSC images, yielding impressive results and significantly enhancing the identification of complex structures, such as galaxy spiral arms and tidal tails, which may have been overlooked in the original DECaLS images. Moreover, Pix2WGAN effectively addresses issues like artifacts, noise, and blurriness in both source and target images. In addition to the basic Pix2WGAN model, we further developed an advanced architecture called Cascaded Pix2WGAN, which incorporates a multi-stage training mechanism designed to bridge the quality gap between SDSS and HSC images, demonstrating similarly promising outcomes. We systematically assessed the similarity between the model-generated pseudo-HSC images and actual HSC images using various metrics, including Root Mean Squared Error (RMSE), Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index (SSIM), along with perceptual metrics such as Learned Perceptual Image Patch Similarity (LPIPS) and Fr\'echet Inception Distance (FID). The results indicate that images transformed by our model outperform both the original SDSS and DECaLS images across nearly all evaluation metrics. Our research is expected to provide significant technical support for astronomical data analysis, cross-survey image integration, and high-precision astrometry.

astro-ph.IM↗

arXiv2024

Photometric Redshift Estimation for CSST Survey with LSTM Neural Networks

Accurate estimation of photometric redshifts (photo-$z$s) is crucial for cosmological surveys. Various methods have been developed for this purpose, such as template fitting methods and machine learning techniques, each with its own applications, advantages, and limitations. In this study, we propose a new approach that utilizes a deep learning model based on Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM) to predict photo-$z$. Unlike many existing machine learning models, our method requires only flux measurements from different observed filters as input. The model can automatically learn the complex relationships between the flux data across different wavelengths, eliminating the need for manually extracted or derived input features, thereby providing precise photo-$z$ estimates. The effectiveness of our proposed model is evaluated using simulated data from the Chinese Space Station Telescope (CSST) sourced from the Hubble Space Telescope Advanced Camera for Surveys (HST-ACS) and the COSMOS catalog, considering anticipated instrument effects of the future CSST. Results from experiments demonstrate that our LSTM model, compared to commonly used template fitting and machine learning approaches, requires minimal input parameters and achieves high precision in photo-$z$ estimation. For instance, when trained on the same dataset and provided only with photometric fluxes as input features, the proposed LSTM model yields one-third of the outliers $f_{out}$ observed with a Multi-Layer Perceptron Neural Network (MLP) model, while the normalized median absolute deviation $\rm \sigma_{NMAD}$ is only two-thirds that of the MLP model. This study presents a novel approach to accurately estimate photo-$z$s of galaxies using photometric data from large-scale survey projects.

astro-ph.GA↗