An artificial intelligence algorithm that identifies middle turbinate pneumatisation (concha bullosa) on sinus...

Area of Science:

• Artificial Intelligence
• Medical Imaging Analysis
• Deep Learning

Background:

• The middle turbinate is the largest of the three turbinates in the nose.
• Pneumatization of the middle turbinate, known as concha bullosa, is a common anatomical variant.
• Concha bullosa can be associated with sinonasal symptoms and may complicate endoscopic sinus surgery.

Purpose of the Study:

• To develop and evaluate a convolutional neural network (CNN) algorithm for detecting middle turbinate pneumatization (concha bullosa).
• To assess the diagnostic accuracy of the CNN in identifying concha bullosa on coronal sinus computed tomography (CT) images.

Main Methods:

• Retrospective collection of high-resolution CT scans of the paranasal sinuses.
• Retraining the classification layer of the Inception-V3 model using transfer learning in Python.
• Utilizing segmentation analysis to potentially enhance diagnostic performance.

Main Results:

• The trained CNN achieved a diagnostic accuracy of 81% (95% CI: 73.0-89.0%).
• The area under the curve (AUC) for the CNN model was 0.93, indicating high discriminative ability.

Conclusions:

• The developed CNN algorithm demonstrates a high accuracy in identifying middle turbinate pneumatization.
• This AI tool shows potential for clinical application in rhinology and otolaryngology for diagnosing anatomical variants.