Deep Dive: An Introduction to AI in Critical Care Medicine-Online-Shortcuts Causing Bias in Medical Imaging

Video Transcription

Video Summary

Judy, an interventional radiologist from Emory University, provides insights into the complexities of data interpretation using AI, focusing on the implications of shortcuts in deep learning. She highlights how shortcuts can lead models to incorrect conclusions by relying on superficial cues rather than core data, such as AI predicting pneumonia on ICU X-rays based on contextual clues rather than actual evidence. Judy illustrates this with examples from medical imaging, dermatology, and more, noting that non-contributory elements like radiographic markers or background artifacts can mislead AI. She stresses that these issues extend beyond imaging, citing biases in AI models affecting real-world applications like healthcare recommendations. New AI architectures, such as foundation models, continue to encode demographic biases, making it crucial to address shortcuts. Judy emphasizes the need for domain experts to provide ongoing feedback to mitigate biased outcomes, ensuring AI models serve all demographics fairly and effectively.

Keywords

AI shortcuts

deep learning

medical imaging

demographic biases

domain experts

healthcare AI

	Society of Critical Care Medicine 500 Midway Drive Mount Prospect, IL 60056 USA Phone: +1 847 827-6888 Fax: +1 847 439-7226 Email: support@sccm.org	Contact Us About SCCM Newsroom Advertising & Sponsorship DONATE	MySCCM LearnICU Patients & Families Surviving Sepsis Campaign Critical Care Societies Collaborative	GET OUR NEWSLETTER
© Society of Critical Care Medicine. All rights reserved. \| Privacy Statement \| Terms & Conditions The Society of Critical Care Medicine, SCCM, and Critical Care Congress are registered trademarks of the Society of Critical Care Medicine.