wisdomik/MedicalNarratives

Multi-modal models are data hungry. While datasets with natural images are abundant, medical image datasets can not afford the same luxury. To enable representation learning for medical images at scale, we turn to YouTube, a platform with a large reservoir of open-source medical pedagogical videos. We curate MedicalNarratives, a dataset 4.7M medical image-text pairs, with 1M samples containing dense annotations in the form of traces and bounding boxes. Similar to \textit{think-aloud} studies where instructors speak while hovering their mouse cursor movements over relevant image regions, 1M images in MedicalNarratives contains localized mouse traces in image pixels, creating a spatial association between the text and pixels.

Please Note: this version contain some bad images we have since removed and would reupload soon.

https://medical-narratives.github.io/static/images/ct_examples.png