Evaluating AI rib fracture detections using follow-up CT scans

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Rib fractures are common trauma injuries.
• Accurate diagnosis is essential for patient management.
• Initial CT scans can miss subtle fractures.

Purpose of the Study:

• To compare AI-driven rib fracture diagnosis using initial and follow-up CT scans.
• To evaluate the impact of AI assistance on fracture detection rates.
• To determine the optimal diagnostic criteria for rib fractures.

Main Methods:

• Retrospective analysis of 113 trauma patients with initial and follow-up CT scans.
• AI software (V2.1.0, Huiying Medical Technology Co.) used for rib fracture detection.
• Comparison of diagnostic sensitivity and specificity among radiologists, AI, and AI-assisted radiologists.

Main Results:

• Follow-up CT identified 712 rib fractures, while initial CT missed 127 (17.84%) and had 4 false positives.
• AI-assisted radiologists (Group 3) achieved higher sensitivity (91.57%) than radiologists alone (Group 1: 82.16%) and AI alone (Group 2: 79.35%).
• AI alone (Group 2) showed lower specificity (84.90%) compared to radiologists (Group 1: 99.80%) and AI-assisted radiologists (Group 3: 99.70%).

Conclusions:

• AI-assisted diagnosis enhances rib fracture detection rates.
• Follow-up CT is recommended as the gold standard for diagnosing rib fractures.
• Radiologist oversight significantly reduces AI-related misdiagnoses.