Unsupervised domain adaptation in cross-modality medical data analysis

Start year: 2023

Summary: The great advances in deep learning over the past decades have been powered by ever-bigger models crunching ever-bigger amounts of data. Building and using datasets for AI systems is often artisanal—painstaking and expensive. Also, training on such massive data comes at a price of huge computational and infrastructural costs. Therefore, how to efficiently create a dataset with high-quality samples is becoming a hot research topic, called data-centric AI (DCAI) in machine learning community. DCAI represents the transition from focusing on the model to the underlying data to make building, maintaining, and evaluating datasets easier, cheaper and more repeatable. DCAI aims to provide machine learning-based tools for automated data governance, such as data denoising, data condensation, data augmentation and data quality evaluation. In this research, we focus on applying unsupervised domain adaptation and intend to borrow knowledge cross-modalities to improve data reliability, data interpretability, and data efficiency.