Artificial Intelligence in Laryngeal Endoscopy: Systematic Review and Meta-Analysis

Area of Science:

• Otolaryngology
• Medical Imaging
• Artificial Intelligence

Background:

• Early diagnosis of laryngeal lesions is crucial for timely treatment and optimal organ function.
• Artificial intelligence (AI) is increasingly used to enhance medical imaging quality and diagnostic accuracy.
• Laryngeal endoscopy plays a key role in visualizing laryngeal structures.

Purpose of the Study:

• To investigate the diagnostic utility of artificial intelligence (AI) in laryngeal endoscopy.
• To evaluate the accuracy, sensitivity, and specificity of AI models in assessing laryngeal lesions.

Main Methods:

• A systematic search of five databases was conducted for studies using AI-enhanced models for laryngeal lesion image analysis.
• Outcomes including accuracy, sensitivity, and specificity were analyzed.
• Eleven studies with a low risk of bias were included in the analysis.

Main Results:

• AI models demonstrated very high overall accuracy, ranging from 0.806 to 0.997.
• Pooled sensitivity and specificity for identifying healthy tissue were 0.91 and 0.97, respectively.
• For differentiating benign from malignant lesions, pooled sensitivity was 0.91 and specificity was 0.94.

Conclusions:

• AI shows exceptionally high accuracy, sensitivity, and specificity in assessing laryngeal lesion images.
• The diagnostic performance of AI models was not significantly different between narrow band imaging and white light endoscopy.
• AI holds significant promise for improving the diagnosis of laryngeal pathologies.