Author Of 1 Presentation
LB1246 - Future implementation of automated analysis tools for Multiple Sclerosis on conventional magnetic resonance imaging. (ID 2132)
Abstract
Background
Magnetic resonance imaging (MRI) is imperative for the detection and characterization of Multiple Sclerosis (MS) lesions in the central nervous system. The revised McDonald's criteria of 2017 involve brain and spinal cord MRI lesions with respect to dissemination in time and space to aid in establishing the diagnosis. Furthermore, MRI is the principal tool of tracking brain and spinal cord changes and monitoring treatment effects and disease progression. Manual evaluation of multiple evolving MS lesions and particularly estimation of brain atrophy is difficult due to the time-consuming nature of longitudinal assessment, the complexity of brain volume estimation, and is subject to significant inter-observer variability.
Objectives
This study aims to survey current commercial and freeware automated tools for lesion identification and brain volume monitoring. We evaluate the feasibility and identify barriers in the adoption of computerized tools in the clinical setting.
Methods
A literature search was performed in PubMed, and Google Scholar databases and publications on automated image evaluation tools in multiple sclerosis were identified and reviewed. Findings in other neurologic populations supplemented limited evidence on reliability and validity.
Results
We evaluated various existing automated software packages suitable and specifically developed for the multiple sclerosis population, including SepINRIA, Icobrain, DeepMedic, and others. We confirmed the benefits of image analysis automation. We describe differences between available software models, their advantages and disadvantages. Notably, we identify challenges faced by existing software implementations representing an obstacle to their wide adoption, such as hardware requirements, price of purchase and maintenance, absence of a gold standard, and uncertainty of healthcare benefits.
Conclusions
After exploring the barriers, we propose solutions to integrating automated image analysis into routine practice through the development of a quality assurance and decision support system.