ASGE Annual Postgraduate Course: Clinical Challenges and Real-Time Decision Making in GI Endoscopy (On-demand)

Video Transcription

Video Summary

In this video, Taposh Roy from Kaiser Permanente discusses annotation and segmentation. He begins by explaining that annotation is the addition of extra information to a document or file, and it can be done for various types of information such as text, images, sound, etc. He provides examples of how images can be annotated, including labeling images of animals or recognizing objects and their parts at a pixel level. Taposh emphasizes the importance of accurate annotation in healthcare and mentions the Hypercavaser dataset, which contains annotated gastrointestinal images. He also discusses the machine learning paradigms of supervised and unsupervised learning, explaining that supervised learning relies on labeled data while unsupervised learning aims to find patterns without prior tagging information. Taposh introduces the concept of automatic image descriptions and mentions work done in non-medical fields. He briefly mentions clustering algorithms such as mean shift, normalized cuts, and level sets, which are used for image segmentation. Finally, he concludes by mentioning that there is potential for using unsupervised clustering methods in medical image processing in the future.

Asset Subtitle

Taposh Roy

Keywords

Taposh Roy

annotation

segmentation

healthcare

supervised learning