arXiv subjects

Wanqing Zhang

Publications and source records attributed to Wanqing Zhang.

7 recordsLinked to original sources

arXiv2026

Spatial Sampling of Hemispherical Arrays for Three-Dimensional Photoacoustic Computed Tomography

Three-dimensional (3D) photoacoustic computed tomography (PACT) is a powerful noninvasive biomedical imaging modality that provides volumetric data for structural and functional assessment \textit{in vivo}. To maximize angular coverage and mitigate limited-view artifacts, modern 3D PACT systems frequently employ hemispherical transducer arrays. While substantial effort has been devoted to improving image quality through post-processing algorithms, the intrinsic impact of the hardware array layout on the baseline image quality remains underexplored. In this study, we systematically investigate how the spatial sampling characteristics of hemispherical array distributions affect imaging performance under fixed hardware constraints. We propose a uniform-spacing sampling criterion to generate three representative array distributions and evaluate their performance using quantitative metrics across static, noise-perturbed, reduced-element, and rotational acquisition scenarios. Across these diverse testing regimes, the Fibonacci distribution consistently demonstrates superior structural robustness and more globally balanced reconstruction quality, a result we attribute to its highly isotropic sampling properties. These findings demonstrate that an optimal hardware-level sampling strategy is critical for maintaining global reconstruction stability. Ultimately, this framework establishes a rigorous quantitative methodology for benchmarking hemispherical array distributions and provides practical design guidance for the future development of 3D PACT systems.

physics.med-ph↗

arXiv2025

FEAT: A Multi-Agent Forensic AI System with Domain-Adapted Large Language Model for Automated Cause-of-Death Analysis

Forensic cause-of-death determination faces systemic challenges, including workforce shortages and diagnostic variability, particularly in high-volume systems like China's medicolegal infrastructure. We introduce FEAT (ForEnsic AgenT), a multi-agent AI framework that automates and standardizes death investigations through a domain-adapted large language model. FEAT's application-oriented architecture integrates: (i) a central Planner for task decomposition, (ii) specialized Local Solvers for evidence analysis, (iii) a Memory & Reflection module for iterative refinement, and (iv) a Global Solver for conclusion synthesis. The system employs tool-augmented reasoning, hierarchical retrieval-augmented generation, forensic-tuned LLMs, and human-in-the-loop feedback to ensure legal and medical validity. In evaluations across diverse Chinese case cohorts, FEAT outperformed state-of-the-art AI systems in both long-form autopsy analyses and concise cause-of-death conclusions. It demonstrated robust generalization across six geographic regions and achieved high expert concordance in blinded validations. Senior pathologists validated FEAT's outputs as comparable to those of human experts, with improved detection of subtle evidentiary nuances. To our knowledge, FEAT is the first LLM-based AI agent system dedicated to forensic medicine, offering scalable, consistent death certification while maintaining expert-level rigor. By integrating AI efficiency with human oversight, this work could advance equitable access to reliable medicolegal services while addressing critical capacity constraints in forensic systems.

cs.AI↗

arXiv2024

Large-vocabulary forensic pathological analyses via prototypical cross-modal contrastive learning

Forensic pathology is critical in determining the cause and manner of death through post-mortem examinations, both macroscopic and microscopic. The field, however, grapples with issues such as outcome variability, laborious processes, and a scarcity of trained professionals. This paper presents SongCi, an innovative visual-language model (VLM) designed specifically for forensic pathology. SongCi utilizes advanced prototypical cross-modal self-supervised contrastive learning to enhance the accuracy, efficiency, and generalizability of forensic analyses. It was pre-trained and evaluated on a comprehensive multi-center dataset, which includes over 16 million high-resolution image patches, 2,228 vision-language pairs of post-mortem whole slide images (WSIs), and corresponding gross key findings, along with 471 distinct diagnostic outcomes. Our findings indicate that SongCi surpasses existing multi-modal AI models in many forensic pathology tasks, performs comparably to experienced forensic pathologists and significantly better than less experienced ones, and provides detailed multi-modal explainability, offering critical assistance in forensic investigations. To the best of our knowledge, SongCi is the first VLM specifically developed for forensic pathological analysis and the first large-vocabulary computational pathology (CPath) model that directly processes gigapixel WSIs in forensic science.

eess.IV↗

arXiv2020

Jointly Sparse Signal Recovery and Support Recovery via Deep Learning with Applications in MIMO-based Grant-Free Random Access

In this paper, we investigate jointly sparse signal recovery and jointly sparse support recovery in Multiple Measurement Vector (MMV) models for complex signals, which arise in many applications in communications and signal processing. Recent key applications include channel estimation and device activity detection in MIMO-based grant-free random access which is proposed to support massive machine-type communications (mMTC) for Internet of Things (IoT). Utilizing techniques in compressive sensing, optimization and deep learning, we propose two model-driven approaches, based on the standard auto-encoder structure for real numbers. One is to jointly design the common measurement matrix and jointly sparse signal recovery method, and the other aims to jointly design the common measurement matrix and jointly sparse support recovery method. The proposed model-driven approaches can effectively utilize features of sparsity patterns in designing common measurement matrices and adjusting model-driven decoders, and can greatly benefit from the underlying state-of-the-art recovery methods with theoretical guarantee. Hence, the obtained common measurement matrices and recovery methods can significantly outperform the underlying advanced recovery methods. We conduct extensive numerical results on channel estimation and device activity detection in MIMO-based grant-free random access. The numerical results show that the proposed approaches provide pilot sequences and channel estimation or device activity detection methods which can achieve higher estimation or detection accuracy with shorter computation time than existing ones. Furthermore, the numerical results explain how such gains are achieved via the proposed approaches.

eess.SP↗

arXiv2020

Jointly Sparse Support Recovery via Deep Auto-encoder with Applications in MIMO-based Grant-Free Random Access for mMTC

In this paper, a data-driven approach is proposed to jointly design the common sensing (measurement) matrix and jointly support recovery method for complex signals, using a standard deep auto-encoder for real numbers. The auto-encoder in the proposed approach includes an encoder that mimics the noisy linear measurement process for jointly sparse signals with a common sensing matrix, and a decoder that approximately performs jointly sparse support recovery based on the empirical covariance matrix of noisy linear measurements. The proposed approach can effectively utilize the feature of common support and properties of sparsity patterns to achieve high recovery accuracy, and has significantly shorter computation time than existing methods. We also study an application example, i.e., device activity detection in Multiple-Input Multiple-Output (MIMO)-based grant-free random access for massive machine type communications (mMTC). The numerical results show that the proposed approach can provide pilot sequences and device activity detection with better detection accuracy and substantially shorter computation time than well-known recovery methods.

cs.IT↗

arXiv2020

Jointly Sparse Signal Recovery via Deep Auto-Encoder and Parallel Coordinate Descent Unrolling

In this paper, utilizing techniques in compressed sensing, parallel optimization and deep learning, we propose a model-driven approach to jointly design the common measurement matrix and GROUP LASSO-based jointly sparse signal recovery method for complex sparse signals, based on the standard auto-encoder structure for real numbers. The encoder achieves noisy linear compression for jointly sparse signals, with a common measurement matrix. The GROUP LASSO-based decoder realizes jointly sparse signal recovery based on an iterative parallel-coordinate descent (PCD) algorithm which is proposed to solve GROUP LASSO in a parallel manner. In particular, the decoder consists of an approximation part which unfolds (several iterations of) the proposed iterative algorithm to obtain an approximate solution of GROUP LASSO and a correction part which reduces the difference between the approximate solution and the actual jointly sparse signals. The proposed model-driven approach achieves higher recovery accuracy with less computation time than the classic GROUP LASSO method, and the gain significantly increases in the presence of extra structures in sparse patterns. The common measurement matrix obtained by the proposed model-driven approach is also suitable for the classic GROUP LASSO method. We consider an application example, i.e., channel estimation in Multiple-Input Multiple-Output (MIMO)-based grant-free random access which is proposed to support massive machine-type communications (mMTC) for Internet of Things (IoT). By numerical results, we demonstrate the substantial gains of the proposed model-driven approach over GROUP LASSO and AMP when the number of jointly sparse signals is not very large.

cs.IT↗

arXiv2019

Joint Design of Measurement Matrix and Sparse Support Recovery Method via Deep Auto-encoder

Sparse support recovery arises in many applications in communications and signal processing. Existing methods tackle sparse support recovery problems for a given measurement matrix, and cannot flexibly exploit the properties of sparsity patterns for improving performance. In this letter, we propose a data-driven approach to jointly design the measurement matrix and support recovery method for complex sparse signals, using auto-encoder in deep learning. The proposed architecture includes two components, an auto-encoder and a hard thresholding module. The proposed auto-encoder successfully handles complex signals using standard auto-encoder for real numbers. The proposed approach can effectively exploit properties of sparsity patterns, and is especially useful when these underlying properties do not have analytic models. In addition, the proposed approach can achieve sparse support recovery with low computational complexity. Experiments are conducted on an application ex-ample, device activity detection in grant-free massive access for massive machine type communications (mMTC). Numerical results show that the proposed approach achieves significantly better performance with much less computation time than classic methods, in the presence of extra structures in sparsity patterns.

cs.IT↗