Korean J Fam Pract.  2021 Dec;11(6):422-437. 10.21215/kjfp.2021.11.6.422.

A Level of Chest X-Ray Nodule Detection Difficulty with and without Artifical Intelligence Based Automatic Detection Assist in Family Medicine Residents

• Yoon C
• Kim DK
• Kim DH

• KMID: 2523465
• DOI: http://doi.org/10.21215/kjfp.2021.11.6.422

Abstract

Background
Chest x-rays are one of the most commonly used radiological studies, and one of the most common findings of chest x-rays at primary care facilities is lung nodules. Recently, artificial intelligence (AI) tools trained to find nodules in chest x-rays have improved significantly. This study aims to find if the level of difficulty can be lowered when detecting nodules in chest radiographic images using an AI tool for family medicine residents.
Methods
Five of the chest x-rays were randomly selected to create the survey. A survey e-mail was sent to 966 family medicine residents. A paired t-test was conducted to check the difference in scores before and after the AI-based reading was provided, and a paired t-test was separately conducted for each year. A one-way analysis of variance (ANOVA) and Scheffe’s post hoc test were conducted to analyze the contribution scores of AI-based reading.
Results
In the paired t-test, the difficulty level decreased after the AI-based reading was provided compared to before the AI-based reading was provided. Furthermore, the difference was statistically significant in all chest x-rays. In the one-way ANOVA, the AI-based contribution score was significantly different according to the difficulty level as F=5.322 (P<0.001). The Scheffe test confirmed that the AI-based contribution was higher in difficulty level four than in levels one and two.
Conclusion
AI-based reading can reduce the level of difficulty when family medicine residents read chest radiography images. AI-based reading is more helpful when reading difficult chest radiographic images.

Keyword

Artificial Intelligence; Chest X-Ray; Lung Nodule; Family Medicine Resident; Surveys