Model performance improved with combination of radiological, clinical data
By Lori Solomon HealthDay Reporter
THURSDAY, Aug. 10, 2023 (HealthDay News) — In a study published Aug. 2 in the American Journal of Roentgenology, a radiomic machine learning model predicted Crohn disease diagnosis with better performance than two of three expert radiologists.
Richard X. Liu, Ph.D., from the University of Cincinnati College of Medicine, and colleagues developed a machine learning-based method for predicting ileal Crohn disease using radiomic features of ileal wall and mesenteric fat from noncontrast T2-weighted magnetic resonance (MR) images. Performance of the artificial intelligence (AI) model was compared to the performance of expert radiologists.
The researchers found that for 135 patients, the three radiologists had accuracies of 83.7, 86.7, and 88.1 percent for diagnosing Crohn disease, with consensus accuracy of 88.1 percent. There was substantial inter-radiologist agreement (κ = 0.78). The bowel core was the best-performing region of interest (area under the receiver operating characteristic curve [AUC], 0.95; accuracy, 89.6 percent); whole bowel fat core and whole fat had worse performance (whole bowel AUC, 0.86; fat core AUC, 0.70; whole-fat AUC, 0.73). A clinical-only model achieved an AUC of 0.85 and accuracy of 80.0 percent, while an ensemble model combining bowel-core radiomic and clinical models achieved AUC of 0.98 and accuracy of 93.5 percent. The bowel-core radiomic-only model had better accuracy than radiologist 1 and radiologist 2, but not radiologist 3 or radiologists’ consensus. The ensemble model achieved better accuracy than radiologists’ consensus.
“Deployment of a radiomic-based model using T2-weighted MR data could decrease inter-radiologist variability and increase diagnostic accuracy for pediatric Crohn disease,” the authors write.
Abstract/Full Text (subscription or payment may be required)
Copyright © 2023 HealthDay. All rights reserved.
