Measuring Impact of Radiologist-AI Collaboration: Efficiency, Accuracy, and Clinical Impact

Much progress has been made in Radiology AI model performance, but only a handful of studies have reported the impact of deploying AI models in the radiology workflow. This study explores the impact of AI integration in radiology by examining the efficiency and accuracy of radiologists with and with...

Full description

Saved in:
Bibliographic Details
Main Authors: Purkayastha, Saptarshi, Isaac, Rohan, Veldandi, Amulya, Saxena, Rijul, Singh, Pallavi, Vaswani, Pallavi, Krupinski, Elizabeth A, Danish, Joshua A, Gichoya, Judy Wawira
Format: Conference Proceeding
Language:English
Subjects:
Accuracy
AI in Healthcare
Artificial intelligence
Biomedical imaging
Biomedical measurement
Collaboration
Diagnostic Efficiency
Human-in-the-loop
Radiology
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Much progress has been made in Radiology AI model performance, but only a handful of studies have reported the impact of deploying AI models in the radiology workflow. This study explores the impact of AI integration in radiology by examining the efficiency and accuracy of radiologists with and without AI assistance. The results reveal a statistically significant reduction in diagnosis time with AI across all four datasets, with radiologists spending, on average, 11 seconds less per study (Non-AI: 55 seconds, AI-Assisted: 44 seconds). Moreover, we explored changes in diagnostic accuracy, with radiologists experiencing varied outcomes. Notably, some radiologists demonstrated improved accuracy with AI (such as an increase from 84% to 95%), while others maintained or slightly decreased their accuracy. The study highlights the need for personalized AI implementation strategies, considering individual characteristics and experience levels.
ISSN:1945-8452
DOI:10.1109/ISBI56570.2024.10635618