Zeming Huang

Zeming Huang

Master's student, University of Chinese Academy of Sciences

I am a master’s student in Electronic Information at the University of Chinese Academy of Sciences, working on AI for Health with a particular interest in cardiovascular AI. My research focuses on learning reliable representations from complex, noisy, heterogeneous, and incomplete biomedical data, and developing methods that are useful in real clinical settings.

I work on medical imaging and physiological signals, including prostate ultrasound segmentation and measurement, longitudinal ECG forecasting, and paroxysmal atrial fibrillation detection. Across these projects, I aim to connect robust learning with clinical needs, meaningful validation, and practical deployment. My engineering background in signal processing, physical modeling, and hands-on system building shapes how I approach these problems.

Publications and Manuscripts

NOR-TL architecture showing normal-reference and AF-specific encoders, with weak labels for pooled detection and interval annotations for sample-wise AF prediction.

NOR-TL

Normal-Reference Temporal Learning with Decoupled Supervision for Paroxysmal AF Detection

Zeming Huang, Tao Wang, Chenhao Wu, Ning Liu, Yiquan Wang, Lei Wang, Gong Su, Zhipei Huang, and Fei Qin

Submitted to ICASSP 2027First author

Learning AF detection from large recording-labeled collections and limited interval annotations. NOR-TL combines normal-reference and AF-specific features with decoupled supervision, evaluated through window-level recognition and time-point labeling on clinical-grade ECG recordings.

PDF
MIRAUS Figure 2: MRI-privileged teacher training, residual privileged transfer, and TRUS-only deployment.

MIRAUS

MRI-Informed Residual Adaptation for Single-Frame TRUS Prostate Boundary Measurement

Zeming Huang, Haixin Chen, Chuanzhen Cao, Tao Wang, Zhipei Huang, and Fei Qin

In preparation for submission to MeasurementCo-first author · listed first

Using MRI-derived anatomy during training to support prostate boundary measurement from a single ultrasound frame. The deployed model requires no MRI and supports both automatic and box-assisted inference.

PDF Code Project
Clinical motivation for MRI-informed ultrasound learning, illustrated using Figure 1 from MIRAUS. Clinical context · MIRAUS Fig. 1

PRISM

A Privileged Representation-Informed Synergistic Model for Transrectal Ultrasound-Only Prostate Segmentation

Zeming Huang, Haixin Chen, Chuanzhen Cao, Tao Wang, Zhipei Huang, and Fei Qin

IEEE BIBM 2026 · acceptedCo-first author · listed first

Transferring MRI anatomical information into prompt-assisted ultrasound segmentation. Privileged attention, reliability-aware fusion, and distribution alignment address weak cross-modal correspondence while keeping inference ultrasound-only.

PDF Code
Figure 1 from the individualized ECG forecasting manuscript: population-based model development and a baseline-to-future clinical workflow.

ECG FORECASTING

Individualized ECG Waveform Forecasting Enables Testable Predictions of Longitudinal Morphology

Tao Wang, Zeming Huang, Yiquan Wang, Ning Liu, Zhicheng Hu, Lei Wang, Gong Su, Zhipei Huang, Ming Yin, and Fei Qin

ManuscriptSecond author

Studying individualized ECG waveform forecasts from a baseline recording, with evaluation focused on longitudinal morphology. My contributions include data curation, waveform analysis, and validation.

CLEF Figure 1: historical ECG standardization, morphology-aware pretraining, and latent evolution for future waveform reconstruction.

LONGITUDINAL ECG

A Cross-Temporal Latent Evolution Framework for Year-Scale ECG Waveform Forecasting from Historical Records

Tao Wang, Zhicheng Hu, Gong Su, Zeming Huang, Zhipei Huang, Ming Yin, Chenhao Wu, Xiangao Meng, and Fei Qin

Information Sciences · under reviewFourth author

Investigating year-scale ECG waveform forecasting from historical recordings. This work examines longitudinal morphology and progression-sensitive changes, supported by my contributions to data processing, analysis, and validation.

BioShoulder data pipeline, from motion curation through biomechanical simulation to quality-controlled muscle activations.

BIOSHOULDER

A Large-Scale Biomechanical Dataset for Shoulder Muscle Activations

ManuscriptThird author

A dataset of quality-controlled, simulated shoulder muscle activations derived from human motion. It combines motion curation, musculoskeletal simulation, and activation screening, with benchmarks for motion-to-muscle prediction.

PDF

Research and Engineering Experience

  • 2025.08 – Present, Individualized ECG Waveform Forecasting - University of Chinese Academy of Sciences; clinical collaborators: Fuwai Hospital and Aerospace Center Hospital, Beijing. Built preprocessing and representation-learning pipelines for over one million public ECG records and over 100,000 hospital recordings; contributed to waveform evaluation and follow-up analysis across two longitudinal forecasting studies, including a clinical subset of 959 participants; contributed to NSFC General Program and Beijing Natural Science Foundation proposal preparation.
  • 2026.04 – 2026.09, NOR-TL: Paroxysmal AF Detection - University of Chinese Academy of Sciences; clinical collaborator: Aerospace Center Hospital, Beijing. Developed normal-reference and AF-specific ECG representations with decoupled supervision: weak recording labels guide pooled detection, while interval annotations supervise sample-wise predictions. Evaluated window-level recognition and time-point detection on public, clinical-grade AFDB and external LTAFDB and CPSC2021 recordings with patient- or recording-disjoint partitions. Achieved 98.0% temporal AUROC and 0.838 time-point MCC on 1,425 CPSC2021 recordings. First-author manuscript submitted to ICASSP 2027.
  • 2024.09 – Present, Privileged Learning for Prostate Ultrasound Segmentation and Measurement - developed PRISM and MIRAUS to use MRI-derived information during training while retaining TRUS-only inference; evaluated patient-disjoint segmentation, external generalization, and model-derived calipers for conventional ellipsoid volume estimation; contributed to manuscript writing and grant proposal preparation for a NSFC Regional Science Fund project.
  • 2025.04 – Present, Biomechanical Simulation and Physical Orthopedic Modeling - constructed an ANSYS model of humeral motion using physiological movement constraints; compared simulated kinematics and force responses with collaborator measurements; designed and built a physical orthopedic model to reproduce humeral motion for bench-scale validation. Co-authored BioShoulder, a dataset of simulated shoulder muscle activations, as part of this broader shoulder-biomechanics collaboration (third author).
  • 2024, Deep Learning-Based Underwater Acoustic Channel Estimation - investigated deep learning methods for underwater acoustic channel estimation in noisy and time-varying conditions, with a focus on signal processing and experimental evaluation.
  • 2023.08 – 2023.09, South China Sea Survey Center, Ministry of Natural Resources of the People’s Republic of China - assisted buoy operations, including GPS and CTD sensor deployment and recovery, data reception, and cloud-based data uploading.
  • 2023 – 2024, National Marine Vehicle Design and Manufacturing Competition - participated in AUV modification, underwater testing, troubleshooting, and competition presentation; received Third Prize in the South China regional contest.
  • 2022, Tracked ROV Mechanical Design - led the design of a tracked remotely operated vehicle; completed 3D modeling, motion simulation, and prototype fabrication using SolidWorks, COMSOL, CATIA, and 3D printing.

Education

  • 2024.09 – Present, Master’s studies in Electronic Information, University of Chinese Academy of Sciences, Beijing, China. Research university of China’s national academy of sciences; ranked 54th globally and 5th in China (U.S. News 2026).
  • 2020.09 – 2024.06, Bachelor of Engineering in Marine Engineering and Technology, Sun Yat-sen University, Guangdong, China. National research university under China’s Ministry of Education; ranked 85th globally and 8th in China (U.S. News 2026).

Honors and Awards

  • 2026: Second-Class Academic Scholarship, University of Chinese Academy of Sciences.
  • 2026: Merit Student, University of Chinese Academy of Sciences.
  • 2024: Third Prize, South China Regional Contest, 12th National Marine Vehicle Design and Manufacturing Competition.
  • 2021: Merit Award, Sun Yat-sen University Experimental Skills Competition.
  • 2021: Outstanding Team Member, Sun Yat-sen University Admissions Outreach Association.

Skills

  • Programming: Python, MATLAB
  • Research: Deep learning, ECG signal processing, time-series modeling, medical image analysis
  • Engineering: SolidWorks, AutoCAD, CATIA, COMSOL, Arduino, 3D printing
  • Languages: Mandarin Chinese (Native); English (CET-6, Intermediate)