Artificial intelligence driven diagnostic model for detecting paranasal sinus opacification in computed tomography...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Visual analysis of paranasal sinuses (PNS) on computed tomography (CT) images is time-consuming, labor-intensive, and subjective.
• Artificial intelligence (AI) and machine learning (ML) tools are being developed to automate radiological image analysis.
• Current methods for PNS analysis lack efficiency and objectivity.

Purpose of the Study:

• To develop and evaluate a coding-free ML model for automated identification of PNS on CT images.
• To assess the accuracy and utility of AI in PNS CT image interpretation.
• To provide an efficient and objective alternative to manual PNS analysis.

Main Methods:

• A dataset of 19,119 anonymous coronal CT images from 90 studies was utilized.
• Images were annotated for sinus locations, names, and opacification status.
• A coding-free ML model was trained using the YOLOv2 algorithm and evaluated with F1 score and Intersection over Union (IoU) metrics.

Main Results:

• The ML model achieved a mean F1 score of 0.89 and a mean IoU50 of 79% on the testing dataset.
• The model demonstrated highest accuracy in detecting normal sphenoid sinuses.
• The lowest accuracy was observed in detecting opacified frontal sinuses.

Conclusions:

• AI and ML can automate the interpretation of PNS CT images effectively.
• A coding-free ML model can be developed and deployed for automated PNS identification with accuracy comparable to custom-coded models.
• This approach offers a promising solution for efficient and objective radiological image analysis.