| |
| |
|
|
| |
A joint initiative between Stanford Medicine and the Stanford Institute for Human-Centered Artificial Intelligence (HAI) to guide the responsible use of AI across biomedical research, education, and patient care.
|
| |
|
|
|
| |
| |
| |
A side-by-side comparison of real RNA tiles and synthetically generated tiles (Figure courtesy of Francisco Carrillo Perez and Olivier Gevaert)
|
|
| |
|
| |
| |
The team anticipates that synthetic data created by the model will allow for a more comprehensive training tool for cancer-identifying algorithms. So far, the tool has created one million realistic images across five cancer types (lung adenocarcinoma, kidney renal papillary cell carcinoma, cervical squamous cell carcinoma, colon adenocarcinoma, and glioblastoma), which the team has shared with other researchers to help inform additional studies.
This research is critical to furthering the conversation about whether synthetic data can help support this type of cancer imaging.
Q: What’s your take on training AI models on synthetic data?
Send an email to: [email protected]
|
|
| |
|
| |
| |
|
|
| |
The European Parliament recently passed the Artificial Intelligence Act an expansive piece of legislation that sets out rules for developers of AI systems as well as restrictions on how the technology can be used.
The law will classify and regulate AI applications based on their risk and potential harm. Depending on their risk assessment, AI applications will undergo security, transparency and quality assessments; if deemed too risky, some AI applications will be banned.
Similar to the EU’s General Data Protection Regulation, the AI Act applies to providers outside of Europe who have products in the EU.
Some experts speculate that the stricter data governance and ethics reviews of the AI Act will impact clinical trials and research that use AI.
|
| |
|
|
|
| |
| |
| |
The Coalition for Health AI (CHAI) recently announced its board of directors and newly appointed CEO, Brian Anderson.
Formed in April 2023, CHAI is a community of academic health systems, organizations, and expert practitioners of AI and data science. Stanford Medicine is a founding member of CHAI. The organization’s primary mission is to develop criteria around evaluating, reporting, and monitoring health AI, and to establish standards that promote transparency and credibility of health AI. It has already created several advisory boards that will guide key aspects of health AI tools so that they meet the needs of all people.
Stanford Health Care’s chief data scientist and RAISE Health faculty member, Nigam Shah, is one of the nine board members who will advise CHAI as it develops “guidelines and guardrails” that drive high-quality health care by promoting the adoption of credible, fair, and transparent health AI systems.
|
|
| |
|
| |
| |
Nigam Shah, PhD, is a professor of medicine at Stanford University, and chief data scientist for Stanford Health Care.
|
|
 |
|
| |
Nigam Shah, PhD, is a professor of medicine at Stanford University, and chief data scientist for Stanford Health Care.
|
|
|
| |
|
| |
| |
| |
How was CHAI first established?
The idea of CHAI was first outlined in an editorial in StatNews in March 2022. That led to a coalition of the willing at multiple institutions, including Stanford Medicine. Over 2023, it became apparent that, to accomplish the collective vision, we'd need a legal entity. CHAI was officially registered in January 2024 as a nonprofit with Stanford Medicine, the Mayo Clinic, Duke University, and Johns Hopkins University as the founders. Now, 16 other medical centers and four companies have joined as founding partners. Another 2,000 people have registered their interest in becoming members and contributing to CHAI's mission. It is an amazing response to the idea of creating community consensus around responsible AI.
Why is it important for an organization like CHAI to exist, especially now?
There are roughly 600 health systems in the United States, roughly 150 professional medical societies, bodies such as the National Academy of Medicine, and patient organizations. There’s a broad desire to tap into the promise of AI to better health. However, no one entity has the answers. Rather than create hundreds of competing frameworks, lifecycles, and guidelines, it’s better to spend the time creating national consensus. That helps us all do things right — as much as possible — from day one.
Other than your appointment to the board of directors, what is Stanford’s involvement with CHAI?
Several individuals at Stanford Medicine have contributed to the Blueprint for trustworthy AI and more are contributing towards the next update of this document. Going forward, there are many opportunities for other leaders at Stanford Medicine to contribute to the various CHAI working streams — of which there are now more than a dozen.
What role do you see yourself playing?
Once we achieve lift off, I hope I can focus more on contributing reusable tools — such as the FURM assessment that we built as part of RAISE Health — to the CHAI community for broad use. I also hope to help create a working instance of the envisioned nationwide network of assurance laboratories, working closely with other AI experts.
|
| |
|
|
|
| |
| |
| |
Pediatric cardiologist Charitha Reddy, MD, teamed up with engineers and computer scientists to develop a model that accurately and reliably estimates the heart’s left ventricle function in children.
Although the tool is not yet in the clinic, Reddy imagines that soon such a tool could help parents screen for heart-pumping weakness in children with heart problems who live far from a hospital. She also thinks it could improve care in rural areas with fewer cardiologists.
|
|
| |
|
| |
| |
Panelists from L-R: Nigam Shah, Natalie Pageler, David Magnus, Sylvia Plevritis, Michael Pfeffer
|
|
| |
|
| |
| |
We want to provide you with the latest news and information on topics about responsible AI in health that are of most interest to you. Please take our poll and rank the topics which you would like to learn more about (1= most relevant and 6 = least relevant). This will help inform future topics covered in our newsletter.
|
|
| |
|
|
|
