|
What would a flight simulator for medical AI look like? As AI becomes more autonomous, it needs a way to prove its effectiveness and safety before taking care of patients.
In this episode of JAMA+ AI Conversations, Bon Ku, MD, MPP, Program Manager at the Advanced Research Projects Agency for Health (ARPA-H), and JAMA+ AI Associate Editor Yulin Hswen, ScD, MPH, speak about the future of AI in health care and the importance of funding ambitious ideas. Ku discusses the need for rigorous evaluation of medical AI, the risk of AI automation bias and alarm fatigue, and how to co-design medical AI with patients.
Listen now on Spotify | Apple Podcasts | YouTube | JAMA.com.
Editor’s Picks in this week’s JAMA+ AI:
- In a study of an AI-based diagnostic tool, a locally deployable system applied to electrocardiography (ECG) images identified patients with transthyretin amyloid cardiomyopathy (ATTR-CM) across multinational retrospective and screening cohorts. Using AI-ECG may help prioritize selected patients for imaging studies where access to cardiac imaging is otherwise limited. (JAMA)
- A deep learning model applied to 2-dimensional cardiac ultrasound video clips accurately identified patients with moderate or greater aortic stenosis. The model accepts multiple echocardiographic views, can be applied to handheld ultrasound, and reduces the technical expertise required for analysis. (JAMA Cardiology)
- An accompanying Editorial indicates that AI-guided transthoracic echocardiography (TTE) may help nonexpert operators obtain interpretable studies and help triage patients with aortic stenosis. Still, advancing to truly autonomous TTE will require additional feasibility testing in the community, not only in an echocardiography laboratory. (JAMA Cardiology)
|
