Sign in Subscribe
Cable News Network

Automated Radiologic Scoring of Computed Tomography of Paranasal Sinuses

Strategic Insights

— 1 min read

A breakthrough study conducted by researchers at the University of Toronto has made significant strides in developing an automated algorithm to determine the radiologic severity of chronic rhinosinusitis (CRS) using computed tomography (CT) scans. This innovation combines a convolutional neural network (CNN) for sinus segmentation with post-processing to compute the Lund-Mackay score (LMS) directly from CT scans. The researchers' model achieved a high degree of accuracy in automatically computing LMS, paving the way for standardized and accurate reporting of paranasal sinus CT.

Key Takeaways:

  • The researchers developed an automated algorithm that combines a CNN for sinus segmentation with post-processing to compute LMS directly from CT scans.
  • The algorithm achieved a mean Dice score of 0.85 for all sinus regions, except for the osteomeatal complex.
  • Individual Dice scores were 0.95, 0.71, 0.78, 0.93, and 0.86 for the maxillary, anterior ethmoid, posterior ethmoid, sphenoid, and frontal sinuses, respectively.
  • The LMS model showed a high degree of accuracy with a score of 0.92, 0.99, 0.99, 0.97, and 0.99 for the maxillary, anterior ethmoid, posterior ethmoid, sphenoid, and frontal sinuses, respectively.
  • The researchers used 1,399 CT scans and manually labelled 13,668 coronal images for individual sinuses, demonstrating the scalability of the algorithm.
  • The study's findings have implications for the standardization and automation of radiologic scoring of paranasal sinus CT.

Statistics:

  • 1,399 CT scans were used to train and test the algorithm.
  • 13,668 coronal images were manually labelling for individual sinuses.
  • The algorithm achieved a mean Dice score of 0.85 for all sinus regions.
  • Individual Dice scores ranged from 0.71 to 0.95 for different sinuses.
  • The LMS model showed a high degree of accuracy with scores ranging from 0.86 to 0.99 for different sinuses.

Sources:

  • The use of a convolutional neural network to automate radiologic scoring of computed tomography of paranasal sinuses. BioMedical Engineering OnLine, 2025,24(1):1-11. (BioMedical Engineering OnLine - http://www.biomedical-engineering-online.com/).
  • NewsRx. Reports Outline Biomedical Engineering Research from University of Toronto (The use of a convolutional neural network to automate radiologic scoring of computed tomography of paranasal sinuses). Health & Medicine Week. May 23, 2025; p 319.

Read more