Detection of Distal Radius and Wrist Fracture Through Plain Radiography Using Artificial Intelligence: A Meta-Analysis of Diagnostic Test Accuracy Data

Purpose: Distal radius and wrist fractures are common orthopedic injuries that necessitate precise and prompt diagnosis for optimal treatment. Traditional diagnostic methods depend solely on readers' interpretation of conventional radiography and can vary in accuracy. With the advancement of artificial intelligence (AI), There is a growing potential to enhance the precision and consistency of fracture detection in medical imaging. This study aims to assess the effectiveness of AI algorithms in diagnosing distal radius and wrist fractures using plain radiography.

Methods/Materials: A systematic search in medical and electronic databases, including Pubmed, Web of Science, and IEEE was implemented up to December 2023, focusing on the application of AI for detecting distal radius and wrist fractures in plain X-rays. Eligible studies were those employing AI algorithms with available data to construct a 2*2 contingency table. Statistical analysis was performed using R software (version 4.0.3) with the mada package.

Results: The analysis included ten studies incorporating 6,006 radiographs with 3,072 confirmed fractures. AI algorithms demonstrated a sensitivity of 93.2% (95% CI: 88.1-96.2%, I2=92.1%), specificity of 91.1% (95% CI: 72.7-97.5%, I2=96.8%), and a false-positive rate of 8.9% (95% CI: 2.5-27.3%). The diagnostic odds ratio was high at 140.78 (95% CI: 31.92-620.89). Additionally, AI exhibited a positive Likelihood Ratio (LR+) of 10.51 (95% CI: 3.07-35.93) and a negative Likelihood Ratio (LR-) of 0.07 (95% CI: 0.04-0.14), indicating its substantial diagnostic accuracy.

Conclusions: The study indicates that AI has substantial potential in accurately detecting distal radius and wrist fractures from plain X-ray images. The observed high diagnostic accuracy and the significant odds ratio suggest that AI could be a highly effective supplement to traditional radiological methods. Its application may enhance the consistency and precision of diagnoses. However, the noted heterogeneity in the results highlights the need for further research to standardize AI applications and ensure their reliable integration into various clinical settings.