Machine learning models could help create a more standardized, reproducible and efficient way of grading Crohn’s disease (CD) severity in the small bowel based on CT imaging.
New research published in Academic Radiology compared the use of a machine learning model to the performance of two radiologists in assessing the severity of CD in a cohort of computed tomography enterography(CTE) scans and found that a hybrid model could hold clinical value in streamlining this often subjective task.
“Advancements in artificial intelligence, computer vision, and machine learning provide novel approaches for standardized assessments and explainable CD quantitation,” Ashish P. Wasnik, MD, with the department of radiology at the University of Michigan, and colleagues wrote. “Previous work shows automated methods are capable of efficiently extracting and quantifying conventional CD imaging features, as defined by radiologists and employed in clinical practice, thereby enabling automated and reproducible assessments.”
For the study, experts compared the severity scores of two radiologists interpreting 236 CTEs to those produced by a hybrid machine learning model that was a combination of deep-learning, 3-D CNN, and Random Forest model. Each was tasked with classifying disease severity at each mini segment of the distal and terminal ileum. Precision, sensitivity, weighted Cohen’s score and accuracy were used to compare performance.
The model’s precision and sensitivity ranged from 42.4% to 84.1% for severity categories on the test set. However, an improvement was observed in its Cohen’s score (κ = 0.83) and accuracy (70.7%), both of which were in line with the radiologists' (κ = 0.87, accuracy = 76.3%).
Further, the model was able to accurately predict disease length and identify the portion of total ileum containing moderate to severe disease with greater than 90% precision, exercising “expert level judgment,” the authors highlighted.
“The proposed model combines clinically inspired features with the abstract predictors of a deep learning model, presenting an automated solution for CD severity predictive analysis,” the experts explained. “The severity assessments provide detailed descriptions of intestinal damage along with clinically relevant measures that are challenging to collect manually.”
With the right adjustments, similar models could “revolutionize” the way CD is assessed, the authors said, adding that such automation could reduce radiologist burden while also improving disease management and patient care.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She joined Innovate Healthcare in 2021 and has since put her unique expertise to use in her editorial role with Health Imaging.