Intracranial Vessel Wall Segmentation Using a Novel Tiered Loss to Capture Class Inclusion
H Zhou1*, J Xiao2, Z Fan2, 3, 4, D Ruan1, 5, (1) Department of Bioengineering, University of California, Los Angeles, CA 90095, (2) Department of Radiology, University of Southern California, Los Angeles, CA 90033, (3) Department of Radiation Oncology, University of Southern California, Los Angeles, CA 90033, (4) Department of Biomedical Engineering, University of Southern California, Los Angeles, CA 90089 (5) Department of Radiation Oncology, University of California, Los Angeles, CA 90095
Presentations
SU-H400-IePD-F6-1 (Sunday, 7/10/2022) 4:00 PM - 4:30 PM [Eastern Time (GMT-4)]
Exhibit Hall | Forum 6
Purpose: To develop an automated vessel wall segmentation method on T1-weighted intracranial vessel wall magnetic resonance images, with a focus on modeling the inclusion relation between the inner and outer boundaries of the vessel wall.
Methods: We propose a novel method that estimates the inner and outer vessel wall boundaries simultaneously, using a network with a single output channel resembling the level-set function height, different from the multi-channel output as in conventional multi-class and multi-label segmentation. The network is driven by a unique tiered loss that accounts for data fidelity of the lumen and vessel wall classes and regularization in the smoothness of class transition and segmentation boundaries.
Results: The proposed method achieved Dice similarity coefficients (DSC) in 2D of 0.925 ± 0.048, 0.786 ± 0.084, Hausdorff distance (HD) of 0.286 ± 0.436 mm, 0.345 ± 0.419 mm and mean surface distance (MSD) of 0.083 ± 0.037 mm and 0.103 ± 0.032 mm for the lumen and vessel wall, respectively, on a test set with 360 image slices; compared favorably to a benchmark model using multi-label setting that achieved DSC 0.924 ± 0.047, 0.794 ± 0.082, HD 0.298 ± 0.477 mm, 0.394 ± 0.431 mm, and MSD 0.087 ± 0.056 mm, 0.119 ± 0.059 mm. Most importantly, our vessel wall segmentation method achieved significant improvement in morphological integrity and feasibility compared to benchmark multi-label and polar-coordinated segmentation methods.
Conclusion: The proposed method provides a systematic approach to model the inclusion morphology and incorporate it into an optimization infrastructure. It generalizes to other applications where inclusion exists among a (sub)set of classes to be segmented. Improved feasibility in result morphology promises better support for clinical quantification and decision.
Funding Support, Disclosures, and Conflict of Interest: This work is supported in part by NIH/NHLBI R01 HL147355.
Keywords
Modeling, Blood Vessels, Segmentation
Taxonomy
IM- MRI : Machine learning, computer vision
Contact Email