Transformers & BioNLP: Transforming Words into Care

Transformer models are an increasingly important component of machine learning solutions. In the first part of this webinar, we will review the current and potential use cases for these models, specifically within the contexts of natural language processing and image analysis, along with important considerations for their development.

In the second part of the webinar, we will focus on clinical language processing. Clinical reports are an essential component of electronic health records (EHRs). These reports contain information about a patient’s medical history, including their diagnoses, treatments, and test results. This information can be extracted and analyzed using natural language processing (NLP) techniques, which can benefit clinical trial recruitment, clinical coding, etc. This part of the webinar will introduce clinical language processing and discuss the contribution of external knowledge sources in clinical NLP. We will then explore how contrastive learning can be used to bridge the gap between clinical text processing and diagnostic imaging, as we will explain how it can be used to train predictive models without the need for manual annotation of data.

Following that, we’ll discuss how self-supervised learning can help bridge the gap between clinical text processing and diagnostic imaging. We’ll explain how this technique allows us to train predictive models without heavily relying on manual data annotation, making the process more efficient and scalable.

Objectives

After attending this webinar, the attendee should be able to:

• Review the basics of transformer models and their increasing presence in machine learning methodologies
• Discuss the uses of transformers in Natural Language Processing and image analysis, with particular focus on applications to radiology practice
• Define NLP and its main tasks including named-entity recognition, co-reference resolution, and relation extraction
• Examine contrastive learning, linking clinical textual reports to diagnostic images, and tackling small-sized annotated data sets

CE Credit

SIIM has approved this activity for 1.0 hours of SIIM IIP Credits towards certification and re-certification by the American Board of Imaging Informatics (ABII).