Our work in "WsiCaption: Multiple Instance Generation of Pathology Reports for Gigapixel Whole-Slide Images” aims to automatically generate pathological reports by a whole slide image (WSI). The whole slide image is the foundation of digital pathology, which is critical to the diagnosis and treatment of cancer. However, writing a pathological report is labor intensive and error prone. To achieve automatic report generation, we curate the largest WSI-report dataset and design a model for text generation based on WSIs. The dataset is named PathText which contains nearly 10000 WSI-text pairs. And our proposed model MI-Gen can transform WSIs to text directly. We believe our work shows the potential of generative model in the pathology field.
Being selected to present our paper shows great appreciation of my work. I feel very honored and I hope it will be a good start for my academic career.
At MICCAI 2024, I'm interested in learning more about Foundation Models in Clinical. Since my work is about visual-language models for WSIs, I am curious about such multimodal models applications in different medical scenes. I also care about the generalization ability of these foundation models in medical area.
While I am at the conference, I am eager to communicate with many researchers and become good friends.