Leaderboard Ad

New Algorithm Offers Clearer Path to CRS Diagnoses

Clinical Validation of an Automated Deep-Learning-Based Algorithm for Quantitative Sinus Computed Tomography Analysis

On Demand – Best of Scientific Oral Presentation

 

Diagnosing chronic rhinosinusitis (CRS) with more precision and objectivity is the goal of a new technique for automated computed tomography (CT) analysis using a deep learning-based algorithm.

Conner J. Massey, MD, a resident at the University of Colorado Anschutz Medical Campus in Aurora, Colorado, said this new approach will give greater nuance to standard radiology interpretations, which can be somewhat imprecise and vague in terms of characterizing disease burden on sinus CTs. “Visual-based scoring systems that aim to complete this task are often easy to learn and use but require a trained evaluator with knowledge of sinus anatomy, are only semi-objective, and are subject to human bias,” he said.

The new algorithm addresses these shortcomings by using automated technology that can analyze a sinus CT scan in about a minute, producing a highly precise quantification of sinus cavity opacification down to the percent.

“To our knowledge, this is the first such algorithm capable of completing this task,” said Dr. Massey, who is presenting the details of the algorithm in an on-demand session in the Scientific Oral Presentations education platform.

“Apart from creating a highly-detailed picture of mucosal disease burden, we have leveraged the algorithm to detect and characterize other important radiological entities, such as osteitis/neo-osteogenesis of the sinus wall,” he said.

In a research study of 88 subjects, the algorithm successfully segmented 100% of scans and calculated percent sinus opacification for all individual sinus cavities in a diverse CRS population with CT images obtained on different machines utilizing varied acquisition protocols.

The algorithm also produces a number of other readouts that are not easily obtainable, such as mean Hounsfield units of opacified regions and sinus cavity volumes.

“This technology has the potential to fundamentally change how mucosal disease burden is quantified in sinus CT scans,” Dr. Massey said. “Importantly, it obviates the need for a human assessor for this task. We expect that the detailed data produced from volumetric analysis will be invaluable to radiologists, ordering physicians, researchers, and patients.”

Top