Sunday, November 30 | 9:30 a.m.-9:40 a.m. | S1-SSHN01-4 | Room E352
In this Sunday morning session, researchers will present findings from a study that showed how deep learning (DL) can improve MRI's ability to identify retropharyngeal edema (RPE) in patients with acute neck infection -- a condition that may indicate potential for disease progression.
A team led by study presenter Jussi Hirvonen, MD, PhD, of Tampere University in Finland, developed a deep-learning algorithm and conducted a study that included 479 patients with acute neck infections who underwent axial T2-weighted water-only Dixon MRI imaging. The model was made up of a convolutional neural network for categorizing individual image slices and an algorithm for patient-level classification based on stacks of slices.
The group tracked the model's accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) and compared these results to radiologists' interpretations. Of the total patient cohort, 51% were RPE-positive and 49% were RPE-negative.
What did Hirvonen and colleagues find? The model had high AUCs at both slice- and patient-level data (0.941 and 0.948, respectively). Attend this session to find out more about how the team's results could improve triage and treatment of patients with RPE.
