Date of Award

Spring 2022

Degree Name

Master of Medical Science (Physician Assistant)

Department

Physician Assistant; College of Health Sciences

First Advisor

Zachary Weik

Abstract

Introduction: The emergency department (ED) has had an appreciable increase in volume over the past two decades with more individuals utilizing the ED now than a decade ago which has placed a strain on healthcare resources. Specifically adding to the rising costs in healthcare are diagnostics including chest x-rays. To combat this, institutions across the nation are developing deep-learning and artificial intelligence (AI) programs to assist in reading these diagnostics. Therefore, this systematic review analyzes the use of AI programs by clinicians in the ED to read diagnostic imaging compared to those who do not use these algorithms.

Methods: This was an analysis of available studies utilized publicly available databases PubMed, Cochrane Library, Ovid, Elsevier, and EBSCOHost. The search parameters included peer-reviewed articles of adult patients from Jan. 2016-Jan. 2021. Search terms included: “Emergency Department/Room”, “Artificial Intelligence”, “Deep-Learning”, “Chest X-ray”, “Imaging”, and “Diagnostics”.

Results: Eleven articles matched the search criteria. All eleven journals found AI programs performed equally to clinicians and occasionally better. AI programs showed a marked decrease in time spent interpreting the diagnostics. These programs were found to be more sensitive, but not necessarily more specific (93% accuracy versus 85% in AI reading for neurologic pathologies).

Conclusion: Artificial intelligence and deep-learning systems have shown increasing viability and capacity to perform as well or better than radiologists in diagnostic metrics like chest radiographs, CT, and MRI readings. However, at present, AI programs in the ED are best utilized in augmenting radiologists instead of outrightly replacing them in practice.

Share

COinS
 

Diagnostic Imaging Read By AI-Assisted Technology within the Emergency Department

Introduction: The emergency department (ED) has had an appreciable increase in volume over the past two decades with more individuals utilizing the ED now than a decade ago which has placed a strain on healthcare resources. Specifically adding to the rising costs in healthcare are diagnostics including chest x-rays. To combat this, institutions across the nation are developing deep-learning and artificial intelligence (AI) programs to assist in reading these diagnostics. Therefore, this systematic review analyzes the use of AI programs by clinicians in the ED to read diagnostic imaging compared to those who do not use these algorithms.

Methods: This was an analysis of available studies utilized publicly available databases PubMed, Cochrane Library, Ovid, Elsevier, and EBSCOHost. The search parameters included peer-reviewed articles of adult patients from Jan. 2016-Jan. 2021. Search terms included: “Emergency Department/Room”, “Artificial Intelligence”, “Deep-Learning”, “Chest X-ray”, “Imaging”, and “Diagnostics”.

Results: Eleven articles matched the search criteria. All eleven journals found AI programs performed equally to clinicians and occasionally better. AI programs showed a marked decrease in time spent interpreting the diagnostics. These programs were found to be more sensitive, but not necessarily more specific (93% accuracy versus 85% in AI reading for neurologic pathologies).

Conclusion: Artificial intelligence and deep-learning systems have shown increasing viability and capacity to perform as well or better than radiologists in diagnostic metrics like chest radiographs, CT, and MRI readings. However, at present, AI programs in the ED are best utilized in augmenting radiologists instead of outrightly replacing them in practice.