Computer-Aided Comparative Assessment of Inflammatory Lesions in MRI of the Spine in Axial Spondyloarthritis Patients
Year: 2014
Type: dissertation
Abstract: Detection of inflammatory lesions in magnetic resonance imaging (MRI) of the spine in axial Spondyloarthritis (SpA) patients is a labor-intensive task. Current scoring methods often require an expert reader to visually assess 23 vertebral units (VU) on multiple MRI slices. Furthermore, monitoring of progressive changes consists of a manual search and alignment of the 23 VUs in both baseline and follow-up scans, followed by side-by-side comparison of two images. We present a semi-automated framework for comparative visualization of inflammatory lesions in MRI of the spine in axial SpA patients. The first part of our method localizes and segments the VUs using a fully automatic algorithm based on quasi-periodicity of vertebral column intensity profile and tissue class probability maps. The second part uses these segmentations to align baseline and follow-up scans by deriving a locally rigid transform from a pre-computed deformable registration. To facilitate visual comparison, the co-registered baseline and follow-up images are fused in a single color-encoded difference image. This allows an expert reader to efficiently interpret progressive changes while focusing on just one image. Experiments in which an expert reader used the proposed framework to score inflammatory changes in axial SpA patients were performed. Results showed strong agreement with ground truth scores for cases of full lesion occurrence/remission.
Author: Evgeni Aizenberg
Language: English
Topic: Medical Imaging and Analysis
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W199565108