Overview of the MEDIQA-Sum Task at ImageCLEF 2023: Summarization and Classification of Doctor-Patient Conversations

This paper presents the overview of the MEDIQA-Sum task at ImageCLEF 2023. MEDIQA-Sum 2023 includes three subtasks, in which a doctor-patient dialogue source is given, and participants were tasked with (A) dialogue2topic classification, e.g. classifying the conversation into one of twenty section header categories, (B) dialogue2note snippet generation, e.g. generating clinical note section text additionally given the clinical section header, and (C) dialogue2note full note summarization, e.g. generating a full clinical note. Twelve teams participated with a total of 48 runs. The best teams achieved 0.8 Accuracy on topic classification (subtask A) and ROUGE-1 scores of 0.43 and 0.49 F1, for subtasks B and C, respectively.