The Informatics e-Learning Hub: Government Regulations and AI

The Total Product Lifecycle in AI for Radiological Imaging

By: Alex Cadotte, PhD

This video provides an overview of the regulation of radiological AI and Machine Learning (ML). After reviewing the Center for Devices and Radiological Health’s (CDRH) mission and vision, key regulatory concepts, and a brief history of AI/ML regulation, the video summarizes how policies could change for new technologies, including autonomous AI and AI-guided imaging.

Learning Objectives:

Learn about the regulation of radiological AI and ML.
Understand CDRH’s mission and vision.
Explore key regulatory concepts.
Grasp the difference between general and special controls for medical devices.
Learn the available pre-market submission pathways.
Get a glimpse of the future of regulation practices.

Overview of FDA Regulatory Framework

By: Howard Chen, MD, MBA

Many of the commercial imaging AI models have already gone through the complex FDA review process. This video highlights the differences between the FDA’s 510(k) premarket notification, a De Novo classification, and Pre-market Approval (PMA).

Because updating ML models is unlike updating typical software, it’s not always clear when a new FDA review submission is required. The video gives an overview of the proposed modifications to the regulatory framework for modifications to AI/ML.

Learning Objectives:

Learn FDA medical device classifications.
Discover major pathways for “software as medical devices” to become commercially available.
Understand the needs surrounding an amended process, and the FDA’s proposal.

Regulatory Issues Related to AI: ACR Initiatives

By: Bibb Allen Jr., MD, FACR

In response to regulatory issues related to AI in diagnostic imaging, the ACR has several ongoing initiatives to ensure the safety and effectiveness of AI for all patients. Along with leveraging existing resources including data registries, the Center for Research and Innovation, and Appropriateness Criteria, the ACR launched the ACR Data Science Institute^® (DSI) in 2017.

The DSI continues to promote clinically relevant, safe, and effective use of AI in clinical practice. Some DSI resources reviewed in this video include structured use cases, AI-Central, AI-LAB, and ACR Connect.

Learning Objectives:

Discover ACR and DSI initiatives to support the use of AI in medical imaging.
Navigate current challenges in the FDA regulatory process for AI algorithms.
Learn how to ensure AI is safe and effective for patients.

Effective Validation of AI Models Prior to Clinical Use

By: Bibb Allen Jr., MD, FACR

Prior to using an AI algorithm in clinical practice, radiologists must ensure it is relevant, safe and effective. FDA-cleared algorithms aren't guaranteed to function well at your practice and can be “brittle,” meaning they may fail to adapt to the conditions outside of the training environment. This video outlines the importance AI validation, the ACR approach to ensuring AI is safe and effective, and how ACR Connect is facilitating distributed validation.

Learning Objectives:

The challenges developers face in ensuring an algorithm will work in all clinical settings.
How to determine if a specific model will work at your practice.
The process for centralized vs. distributed algorithm validation.

Evaluating AI for Use in Your Clinical Practice

By: Bibb Allen Jr., MD, FACR

When evaluating an AI algorithm for clinical practice, radiologists must consider factors beyond FDA clearance. The FDA clearance process focuses on safety and efficacy, but does not ensure generalizability to all practices. Local assessment is essential to ensure a model will work in your practice. This video reviews the FDA clearance process, the importance of evaluating AI before implementation, and how the ACR AI-LAB can be used to evaluate AI using your own data.

Learning Objectives:

What the FDA clearance process of AI algorithms involves.
Specific challenges the FDA faces in the AI clearance process.
How to evaluate AI using your own patient data.

The Informatics Value Proposition for Radiologists

By: Bibb Allen Jr., MD, FACR

The Imaging 3.0 initiative promotes radiologists’ value to the healthcare system beyond imaging interpretation. Informatics resources including clinical decision support, structured reporting, image sharing, registries and AI have been key to the Imaging 3.0 program. This video gives an overview of the Imaging 3.0 initiative and why AI could become radiology’s next value proposition.

Learning Objectives:

Details of the ACR Imaging 3.0 initiative.
The value of the ACR informatics tools.
Why AI could become radiology’s next value proposition.