publications

2026

1. 

   Learning to Select Visual In-Context Demonstrations

   Eugene Lee, Yu-Chi Lin, and Jiajie Diao

   In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026

2. 

   Semantic-Guided Pruning of Visual Tokens for Vision-Language models

   Zheng-Wei Liu, Eugene Lee, Jui-Huang Tsai, and 2 more authors

   In ICASSP 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2026

3. 

   Adaptive Thermal Control For BioFPGA-Based Nucleic Acid Amplification Tests

   Yi-Xuan Ran, Yi-Wei Lin, Jhan-Yi Liao, and 2 more authors

   In ICASSP 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2026

2025

1. 

   Understanding the role of polyethylene glycol coating in reducing the subcellular toxicity of MXene nanoparticles using a large multimodal model

   Aditya Yadav, Eugene Lee, Rui Chen, and 3 more authors

   Materials Today Bio, 2025

   Abs DOI

   

   MXenes exhibit remarkable properties such as high electrical conductivity, mechanical strength, and versatile surface chemistry, positioning them as promising candidates for various applications in energy storage, biomedical engineering, and environmental remediation. However, concerns regarding their potential toxicity necessitate a deeper understanding of their interactions with biological systems. This study investigates the subcellular interactions of MXenes, focusing on the mitochondria, endoplasmic reticulum (ER), and lysosomes. Using structured illumination microscopy (SIM), the observations reveal that MXenes exhibit selective toxicity, primarily causing structural damage to the ER, while leaving mitochondria and lysosomes largely unaffected. This organelle-specific toxicity appears to be linked to the negative surface charge of MXenes, which modulates their interactions with cellular components. Notably, surface modification of MXenes with polyethylene glycol (PEG) significantly reduces ER toxicity, thereby improving biocompatibility and offering a promising strategy for safer biomedical applications. Building on these findings, it also introduces a novel methodology that employs a large multimodal model (LMM), a state-of-the-art artificial intelligence (AI) framework, for the automated analysis and interpretation of super-resolution microscopy images.

2. 

   M*: On-Chip Microfluidic Operations With A-Star for Portable Diagnostics

   Cheng-Hsuan Hsieh, Eugene Lee, Peng Hsien Chi, and 2 more authors

   In ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025

   DOI

   

3. 

   Modulating inter-mitochondrial contacts to increase membrane potential for mitigating blue light damage

   Yuxin Wang, Kangqiang Qiu, Weiwei Zou, and 8 more authors

   bioRxiv, 2025

   PDF

   

4. 

   A Versatile Near-Infrared Fluorescent Probe for Fast Assessment of Lysosomal Status via a Large Multimodal Model

   Rui Chen, Eugene Lee, Yuxin Wang, and 5 more authors

   Aggregate, 2025

   Abs DOI PDF

   

   ABSTRACT Lysosomes are essential organelles for cells that act as the “recycling center” for decomposing biomolecules and clearing out damaged organelles. The status of lysosomes is tightly regulated by cells to maintain normal homeostasis. To monitor subcellular lysosomal status, super-resolution imaging has emerged as a promising technology that surpasses conventional imaging limitations, offering extraordinary visualization capability. However, existing fluorescent probes for super-resolution imaging still suffer from significant drawbacks, such as complex synthesis, poor intracellular stability, and the lack of near-infrared (NIR) imaging capability. Besides, to quantitatively analyze fluorescence images, traditional human-driven image interpretation is time-consuming and prone to information loss and human error. To tackle these challenges, we first developed a quinolinium-based fluorescent probe, PA-2, for NIR super-resolution imaging of lysosomes with low cytotoxicity and stable fluorescence. Harnessing PA-2’s strong resistance to photobleaching, the lysosomal dynamic statuses, encompassing autophagy, mitochondria-lysosome contacts, and mitophagy, were successfully visualized. Building on this, we next demonstrate a novel approach leveraging a large multimodal model (LMM), an advanced artificial intelligence (AI) tool, for automated analysis of super-resolution images. The LMM accurately interprets images of PA-2 and predicts lysosomal status under various drug treatments with remarkable speed, precision, and explainability, significantly outperforming human experts in image analysis. To sum up, this work highlights the strong potential of combining advanced fluorescent probe design with AI-assisted image interpretation to drive revolutionary innovation in bioimaging and beyond.

2024

1. 

   Monitoring Escherichia coli in water through real-time loop-mediated isothermal amplification on biochips

   Yuxin Wang, Yun-Sheng Chan, Eugene Lee, and 3 more authors

   Micromachines, 2024

   

2023

1. 

   Cross-resolution flow propagation for foveated video super-resolution

   Eugene Lee, Lien-Feng Hsu, Evan Chen, and 1 more author

   In Proceedings of the IEEE/CVF winter conference on applications of computer vision, 2023

   PDF

   

2021

1. 

   Few-shot and continual learning with attentive independent mechanisms

   Eugene Lee, Cheng-Han Huang, and Chen-Yi Lee

   In Proceedings of the IEEE/CVF international conference on computer vision, 2021

   PDF

   

2. 

   PPG-based smart wearable device with energy-efficient computing for mobile health-care applications

   Eugene Lee and Chen-Yi Lee

   IEEE Sensors Journal, 2021

   

2020

1. 

   Meta-rppg: Remote heart rate estimation using a transductive meta-learner

   Eugene Lee, Evan Chen, and Chen-Yi Lee

   In European conference on computer vision, 2020

   PDF

   

2. 

   Cross-domain adaptation for biometric identification using photoplethysmogram

   Eugene Lee, Annie Ho, Yi-Ting Wang, and 2 more authors

   In ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020

   PDF

   

3. 

   Neuralscale: Efficient scaling of neurons for resource-constrained deep neural networks

   Eugene Lee and Chen-Yi Lee

   In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020

   PDF

   

2019

1. 

   Centralized State Sensing using Sensor Array on Wearable Device

   Eugene Lee, Tsu-Jui Hsu, and Chen-Yi Lee

   In 2019 IEEE International Symposium on Circuits and Systems (ISCAS), 2019

   DOI PDF