Publications by authors named "Shurong Pan"
Bootstrap each lead's latent: A novel method for self-supervised learning of multilead electrocardiograms.
Comput Methods Programs Biomed
December 2024
Article Synopsis
- Electrocardiograms (ECGs) are vital for diagnosing cardiovascular diseases, but training deep learning models for automated detection normally depends on expensive and time-consuming manual labeling.* -
- The proposed BELL method (bootstrap each lead's latent) enhances model performance by using self-supervised learning to leverage unlabeled ECG data, minimizing the reliance on labeled data during training.* -
- BELL outperforms previous models, showing improved performance in downstream tasks with limited training data and demonstrating adaptability to uncurated real-world ECG data, reducing the need for manual cardiologist labels.*
The ultrathin thickness of 2D layered materials affords the control of their properties through defects, surface modification, and electrostatic fields more efficiently compared with bulk architecture. In particular, patterning design, such as moiré superlattice patterns and spatially periodic dielectric structures, are demonstrated to possess the ability to precisely control the local atomic and electronic environment at large scale, thus providing extra degrees of freedom to realize tailored material properties and device functionality. Here, the scalable atomic-scale patterning in superionic cuprous telluride by using the bonding difference at nonequivalent copper sites is reported.
View Article and Find Full Text PDF