Acute Stroke Imaging

In this video, Dr. Luca Giancardo discusses three innovative projects. The first project focuses on using deep learning to bridge the gap between advanced and more accessible imaging modalities for efficient endovascular stroke therapy. The second project explores leveraging radiology reports to pre-train AI models, addressing challenges related to manual image annotation. Lastly, the speaker discusses using retina imaging as a cost-effective proxy for CT imaging in stroke care. The video highlights successful outcomes and the potential of foundation models for diverse acute stroke tasks.
1. Introduction to Acute Stroke Imaging Projects
a. Overview of projects pushing imaging boundaries in acute stroke management.
Time Stamp: 00:07-00:43
2. Project 1: Community Screening for Endovascular Stroke Therapy
a. Addressing the unmet need for advanced imaging in endovascular stroke therapy.
b. Proposal to use deep learning to extract information from more CTA images.
Time Stamp: 00:44-2:46
3. Challenges in AI with Clinical Data
a. Highlighting challenges in AI applications to clinical data, especially in dealing with noisy labels.
b. Video of a live demo of a model for identifying stroke core from CTA images.
Time Stamp: 2:46-7:14
4. Project 2: Leveraging Radiology Reports for AI Model Training
a. Introduction to using radiology reports for pre-training AI models; overcoming labeling challenges.
Time Stamp: 7:15-9:36
5. Training 3D Models from Scratch
a. Discussing the development of 3D learning architecture trained from scratch using radiology reports.
Time Stamp: 9:37-12:37
6. Project 3: Retina Imaging as a Proxy for CT Imaging
a. Exploring the feasibility of retina imaging as a cost-effective proxy for CT imaging in stroke care.
Time Stamp: 12:38-14:04
7. Live Demo of Retina Imaging Model
a. Referencing the live demo of a model for assessing stroke probability available on the Internet.
Time Stamp: 14:05-14:36
8. Acknowledgments
a. Expressing gratitude to collaborators/contributors to the projects discussed.
Time Stamp: 14:37-14:52