Progressive Transformer-Based Generation of Radiology Reports
This incremental approach addresses the problem of generating detailed radiology reports for medical imaging, potentially aiding radiologists in clinical workflows.
The authors tackled radiology report generation by dividing it into two steps: first generating global concepts from images, then refining them into coherent text using transformers, improving state-of-the-art results on two benchmark datasets.
Inspired by Curriculum Learning, we propose a consecutive (i.e., image-to-text-to-text) generation framework where we divide the problem of radiology report generation into two steps. Contrary to generating the full radiology report from the image at once, the model generates global concepts from the image in the first step and then reforms them into finer and coherent texts using a transformer architecture. We follow the transformer-based sequence-to-sequence paradigm at each step. We improve upon the state-of-the-art on two benchmark datasets.