Li-Chin Chen received her Ph.D. in the Graduate Institute of Biomedical Electronics and Bioinformatics at National Taiwan University, and currently affiliated with the Research Center for Information Technology Innovation, Academia Sinica. Her research focuses on time-series data analysis, anomaly detection, and multi-modality fusion, latent space disentanglement and interpolation in deep neural networks, particularly in the medical domain.

• Transferring disease progress from prevalent cases to rare cases

After the establishment of health information systems, a large amount of patient data has been collected in digital format. A prevalent issue arises from the abundance of cases that are common, while the occurrences of rare cases remain limited. This greatly impedes the potential application of machine learning in the management of such rare cases. Leveraging the distinctive attributes of deep learning, she conducted research on the transferability of disease progression from one condition to another, prognosticating the disease progress by employing methodologies like self-supervised learning, GANs, and transfer learning.

• Multi-modality fusion through multi-modal learning

Patient data can be captured across various digital formats, encompassing tabular/structured data, textual summaries, visual imagery, and auditory recordings. Each modality encapsulates distinctive and consequential insights into both the patient and the disorder. Consequently, she investigates into the methods of assimilating and harnessing the underlying representations inherent in each modality. This entailed the exploration of multimodal fusion learning incorporating both simultaneous and asynchronous data, latent representation disentanglement and interpolation, while leveraging state-of-the-art pretraining models and capitalizing on the inherent attributes of each modality.

Having previously engaged in the developing and adopting informatics applications, optimizing clinical workflows, and integrating IoT devices with existing legacy systems, integrating deep neural network applications into transactional systems as a prospective avenue for future exploration.