Motivation

There is a rapid accumulation of different types of clinical content, including medical records and clinical case reports. These are generally written not only in English but actually in a variety of languages. Some of the clinical reports can be very long and thus it is challenging for domain experts to read and keep track of key clinical insights. LLMs have shown promising results for summarization approaches that can be useful to condense lengthy clinical cases into a shorter version of text, reducing the size of the initial text while preserving key clinical informational elements. Thus there is a pressing need to evaluate and benchmark how well clinical summarization works for case reports written in different languages.

We propose a new task called MultiClinSum covering the automatic summarization of lengthy clinical case reports written in different languages, namely English, Spanish, French, and Portuguese. The MultiClinSum task will rely on a corpus of manually selected full clinical case reports and their corresponding clinical case report summaries derived from case report publications written in the previously mentioned languages. For evaluation proposes, automatically generated summaries will be compared against manually generated summaries generated by the original authors, exploring Rouge-2 scores and  BERTScore for evaluation assessment.

As clinical case reports do share commonalities with medical discharge summaries, insights provided by the MultiClinSum results can be of practical relevance also for clinical records summarization scenarios