ISRA May 2022

Curved Planar Reformatting Of The Small Bowel for Visualization and Quantitative Assessment of Crohn’s Disease from MRE– A Pilot Study

Faten Haj Ali Shinnawi 1 Moti Freiman 2 Angeleene Ang 2 Yael Zaffrani 2 Anat Ilivitzki 3
1Department of Medical Imaging, Hillel Yaffe Medical Center, Hadera, Technion-Institute of Technology, Israel
2Faculty of Biomedical Engineering, Technion-Institute of Technology, Israel
3Department of Medical Imaging, Rambam Health Care Campus, Technion-Institute of Technology, Israel

Introduction: The terminal ileum (TI) is the most commonly affected segment of bowel among Crohn’s disease (CD) patients. MR-enterography (MRE) is rapidly evolving as the imaging of choice for diagnosis and monitoring disease activity. Giving the tortuosity of the small bowel, measuring the length of affected inflamed segments and the distance between them is challenging. Therefore, advanced machine learning techniques were recruited on purpose to present the small bowel in a straightened form and achieve a better assessment.

Purpose: To compare the efficacy and time needed to calculate the length of terminal and distal ileum segments in MRE studies of CD patients with involvement of these segments between manual and software-assisted techniques.

Methods: 18 patients who underwent MRE at Rambam healthcare campus were enrolled. We used the coronal T2-weighted images. For the machine learning measurements, we used in-house software developed in python+vtk environment to annotate the TI central line, reconstruct and visualize 3D multiplanar reformations, and perform disease extent measurements. The results were compared to manual traditional measurement of the same segments. We calculated the time needed to measure the same segments manually and software-assisted. We also assessed intra-observer variability for each method.

Results: the mean length of the terminal ileum segment we measured was 23.9 cm and 25.3 cm when calculated manually and using the software-assisted manner, respectively. The mean time for calculating the length of the bowel segment was 103 seconds in the manual manner compared to 10 seconds in the software-assisted manner. Intra-observer variability was noticed only when performed manually.

Conclusion: Using a unique in-house software for measurements of bowel length showed to be reliable, reproducible, and rapid compared to the manual measurements. When applied to the entire short bowel, radiologists’ reports for MRE studies of CD patients will be more accurate, standardized, and less time consuming.