A quality assessment tool for focused abdominal sonography for trauma examinations using artificial intelligence

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Emergency Medicine Diagnostics

Background:

• Current methods for assessing Focused Abdominal Sonography for Trauma (FAST) image quality lack defined criteria.
• Existing tools often evaluate sonographer skill rather than examination quality.
• There is a need for a standardized, reliable system for FAST image quality assessment.

Purpose of the Study:

• To develop a grading system for FAST examinations with high inter-coder agreement.
• To enable the creation of an automated assessment tool for FAST examinations using artificial intelligence (AI).

Main Methods:

• Five coders assigned quality scores (1-5) to FAST clips, with 10% reviewed in triplicate for reliability.
• An AI model was trained using a dataset of 1,514 clips, split into training, validation, and test sets.
• The AI model was developed to predict FAST examination quality scores, distinguishing passing (score ≥3) from failing clips.

Main Results:

• The final dataset achieved 94% agreement between coders on pass/fail predictions, with a Krippendorff α of 66%.
• The AI model demonstrated a fivefold reduction in clips needing manual review compared to unassisted reviewers.
• The AI model correctly identified 85% of passing clips while significantly decreasing the review burden.

Conclusions:

• AI integration shows significant potential for enhancing the accuracy of FAST image evaluation.
• The developed AI tool can substantially alleviate the workload for expert physicians reviewing FAST examinations.
• This approach paves the way for more efficient and reliable quality control in emergency sonography.