Identification of demyelinating lesions and application of McDonald criteria when confronted with white matter lesions on brain MRI

Introduction: White matter lesions (WML) on magnetic resonance imaging (MRI) are common in clinical practice. When analyzing WML, radiologists sometimes propose a pathophysiological mechanism to explain the observed MRI abnormalities, which can be a source of anxiety for patients. In some cases, discordance may appear between the patient's clinical symptoms and the identification of the MRI-appearing WML, leading to extensive diagnostic work-up. To avoid misdiagnosis, the analysis of WML should be standardized, and a consensual MRI reading approach is needed.

Objective: To analyze the MRI WML identification process, associated diagnosis approach, and misinterpretations in physicians involved in WML routine practice.

Methods: Through a survey distributed online to practitioners involved in WML diagnostic work-up, we described the leading causes of MRI expertise misdiagnosis and associated factors: clinical experience, physicians' subspecialty and location of practice, and type of device used to complete the survey. The survey consisted of sixteen T2-weighted images MRI analysis, from which ten were guided (binary response to lesion location identification), four were not shown (multiple possible answers), and two were associated with dissemination in space (DIS) McDonald criteria application. Two independent, experienced practitioners determined the correct answers before the participants' completion.

Results: In total, 364 participants from the French Neuro Radiological (SFNR), French Neurological (SFN), and French Multiple Sclerosis (SFSEP) societies completed the survey entirely. According to lesion identification, 34.3% and 16.9% of the participants correctly identified juxtacortical and periventricular lesions, respectively, whereas 56.3% correctly identified non-guided lesions. Application of the 2017 McDonald's DIS criteria was correct for 35.3% of the participants. According to the global survey scoring, factors independently associated with correct answers in multivariate analysis were MS-expert subspecialty (P<0.001), young clinical practitioners (P=0.02), and the use of a computer instead of a smartphone to perform WML analysis (P=0.03).

Conclusion: Our results highlight the difficulties regarding WML analysis in clinical practice and suggest that radiologists and neurologists should rely on each other to ensure the diagnosis of multiple sclerosis and related disorders and limit misdiagnoses.