Aux-MVNet: Auxiliary Classifier-Based Multi-View Convolutional Neural Network for Maxillary Sinusitis Diagnosis on...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Sinusitis Diagnostics

Background:

• Computed tomography (CT) is the gold standard for sinusitis diagnosis but involves high costs and radiation.
• Radiography is a preliminary imaging technique for sinusitis, guiding the need for CT scans.
• Current diagnostic methods for sinusitis can be resource-intensive and time-consuming.

Purpose of the Study:

• To develop a multi-view convolutional neural network (CNN) for accurate sinusitis severity estimation using only radiographs.
• To simplify the diagnostic workflow for paranasal sinus views.
• To reduce reliance on CT scans by leveraging standard X-ray imaging.

Main Methods:

• A multi-view CNN with a cascaded architecture was designed.
• The network analyzes Waters' view and Caldwell's view radiographs.
• The CNN simultaneously performs maxillary sinus localization and sinusitis classification.

Main Results:

• The proposed CNN achieved an average area under the curve (AUC) of 0.722 for maxillary sinusitis classification.
• AUCs of 0.750 for left and 0.700 for right maxillary sinusitis were obtained.
• The network demonstrated accurate sinusitis severity estimation without CT assistance.

Conclusions:

• A multi-view CNN can accurately estimate sinusitis severity from radiographs, offering a viable alternative to CT scans.
• This AI-driven approach simplifies sinusitis diagnosis and potentially reduces healthcare costs and radiation exposure.
• The cascaded CNN architecture effectively handles both localization and classification tasks for maxillary sinusitis.