MICCAI 2021 Daily – Wednesday

In this paper, Ivona proposes a variational topic inference method for automatic chest X-ray report generation , with a probabilistic latent variable model. It introduces a set of latent variables which are defined as a topic of each of the sentences that belong to the report. These latent variable topics are learned by aligning the visual and language features in the latent space. Ivona Najdenkoska is a PhD student at the University of Amsterdam (UvA), under the supervision of Xiantong Zhen and Marcel Worring. She is part of two lab groups – the AIM Lab (AI for Medical Imaging), which is a collaboration between UvA and the Inception Institute of AI in Abu Dhabi, and the Multimedia Analytics Amsterdam Lab (MultiX), which is based in the Informatics Institute at UvA. Her first ever paper as a PhD student, which proposes a new model for medical report generation based on chest X-rays, has been selected for an oral this year – a huge achievement! She speaks to us ahead of her presentation and poster session today. Variational Topic Inference for Chest X-Ray Report Generation 4 DAILY MICCAI Wednesday Oral Presentation