Longitudinal visualization for exploratory analysis of multiple sclerosis lesions. (October 2022)

• Record Type: Journal Article
• Authors: Sugathan, Sherin; Bartsch, Hauke; Riemer, Frank; Grüner, Renate; Lawonn, Kai; Smit, Noeska
• Is Part Of: Computers & graphics. Volume 107(2022)
• Page Start: 208
• Page End: 219
• Publication Date: 2022-10
• Subjects: Multiple sclerosis -- Longitudinal visualization -- Lesion analysis -- Magnetic resonance imaging -- White matter lesions; Computer graphics -- Periodicals; 006.6
• DOI: 10.1016/j.cag.2022.07.023
• Languages: English
• ISSNs: 0097-8493

Abstract

Abstract: In multiple sclerosis (MS), the amount of brain damage, anatomical location, shape, and changes are important aspects that help medical researchers and clinicians to understand the temporal patterns of the disease. Interactive visualization for longitudinal MS data can support studies aimed at exploratory analysis of lesion and healthy tissue topology. Existing visualizations in this context comprise bar charts and summary measures, such as absolute numbers and volumes to summarize lesion trajectories over time, as well as summary measures such as volume changes. These techniques can work well for datasets having dual time point comparisons. For frequent follow-up scans, understanding patterns from multimodal data is difficult without suitable visualization approaches. As a solution, we propose a visualization application, wherein we present lesion exploration tools through interactive visualizations that are suitable for large time-series data. In addition to various volumetric and temporal exploration facilities, we include an interactive stacked area graph with other integrated features that enable comparison of lesion features, such as intensity or volume change. We derive the input data for the longitudinal visualizations from automated lesion tracking. For cases with a larger number of follow-ups, our visualization design can provide useful summary information while allowing medical researchers and clinicians to study features at lower granularities. We demonstrate the utility of our visualization on simulated datasets through an evaluation with domain experts. Graphical abstract: Highlights: Support for exploratory studies on large longitudinal MS datasets. Solve the difficulty in analyzing intensity from large longitudinal MS datasets. Solve the difficulty in perceiving the overall trend of lesion activity over time. Linked visualizations are proposed for detailed exploration of MS lesions. Insights on 3D longitudinal trends can be clearly studied.