ARRS 2022 Abstracts

RETURN TO ABSTRACT LISTING


E1700. Common Medicolegal Issues with AI Discrepancies: A Case-Based Educational Series
Authors
  1. Steven Rothenberg; University of Alabama at Birmingham; University of Maryland School of Medicine
  2. Andrew Smith; University of Alabama at Birmingham
  3. Srini Tridandapani; University of Alabama at Birmingham
Background
Recently there has been an explosion of commercially available artificial intelligence (AI) imaging products. Although vendors are limited to marketing the specific intended use of this technology, implementation is customized to the request of the practice. In our experience with AI workflows, we have uncovered several recurring issues that carry medical or legal risk due to AI-physician discrepancies. It is important for the radiologist to understand the fundamentals of medical malpractice as it relates to AI to handle these scenarios appropriately.

Educational Goals / Teaching Points
The goal of this presentation is to educate radiologists on how to handle common medicolegal issues that arise with the initial AI implementation phase and subsequent diagnostic use. Specific workflow solutions for both the practice and individual radiologist will be illustrated through a series of three common discrepancy scenarios.

Key Anatomic/Physiologic Issues and Imaging Findings/Techniques
In case 1, a newly implemented AI system for case prioritization identifies a missed PE from an overnight read. An example systematic pathway for escalation of care is reviewed. In case 2, an enlarging suspicious lung nodule on a chest CT is detected by an AI algorithm. The AI also identified this nodule as measuring greater than 6 mm in a prior study that was not reported, potentially resulting in a delay in diagnosis. Professional considerations for reporting with a lens toward risk management are discussed. In case 3, AI detects and characterizes a potential malignancy with high confidence. The radiologist does not agree. Several ways to document and save results are illustrated.

Conclusion
Although AI has the potential to improve the quality of care for patients, there are recurring scenarios that carry medicolegal risk. It is important for the radiologist to understand what to do when these situations occur and what workflow considerations should be documented to decrease departmental risk.