Export
Yacong Li
· 2 previous affiliations
- Researcher · Since 2022
- Harbin Institute of Technology2016–2022
- PhD student · 2016–2022
Research Overview
Yacong Li focuses on computational modeling of cardiac electrophysiology and the application of machine learning to computer vision tasks. Early research investigates the mechanisms of biological pacemakers and drug-induced cardiotoxicity using cellular models. More recently, Li has collaborated on weakly-supervised learning and active learning frameworks for salient object detection.
Research Trajectory
Yacong Li’s early research trajectory is characterized by first-author contributions to computational biology, specifically focusing on the electrophysiological modeling of human ventricular myocytes and cardiac tissue. In 2022, Li led investigations into the development and stability of biological pacemakers. In Effect of cell coupling between pacemaker cells on the biological pacemaker in cardiac tissue model, Li examined how gene modifications—such as the overexpression of the HCN gene to increase hyperpolarization-activated current and the knockout of the Kir2.1 gene—can transform ventricular myocytes into pacemaker cells. This work extended previous single-cell models to the tissue level, systematically analyzing how cell coupling influences pacemaking ability in a broader biological context.
During the same period, Li applied these computational modeling techniques to pharmacology and toxicology. As the lead author of Effect of arsenic trioxide on human ventricular myocytes: a model study, Li investigated the cardiotoxic side effects of arsenic trioxide, a drug commonly used to treat acute promyelocytic leukemia. The study aimed to elucidate the underlying mechanisms by which clinically relevant dosages of the drug induce severe arrhythmias, such as long QT syndrome and torsade de pointes, providing a computational framework for understanding drug-induced cardiac events.
By 2025, the available evidence indicates a shift toward collaborative work in computer vision and machine learning methodologies. Participating as a co-author, Li contributed to Pixel is All You Need: Adversarial Spatio-Temporal Ensemble Active Learning for Salient Object Detection. This research addresses the efficiency of weakly-supervised learning, proposing an adversarial spatio-temporal ensemble active learning framework. The project demonstrates how models trained on sparsely annotated point-labeled datasets can achieve performance equivalent to those trained on densely annotated data, utilizing adversarial attacks to trigger uncertainty and guide the active learning process.
Current Directions
The most recent papers in the evidence point to a collaborative focus on label-efficient machine learning and computer vision. Recent work investigates the challenges of salient object detection under weakly-supervised conditions. This includes exploring how active learning frameworks and adversarial techniques can be combined to minimize the manual labeling effort required for training robust visual models, ensuring that sparsely annotated data can yield high-performance outcomes.
Profile updated
Loading new papers…
Loading research neighbors…
See something wrong? Report a mistake