A Belgian study assessing relapsing multiple sclerosis activity through automated magnetic resonance imaging volumetry

Aim This study assessed to what extent icobrain ms, a commercially available artificial intelligence (AI)-assisted magnetic resonance imaging (MRI) volumetry software, detected disease activity in patients with clinically and radiologically stable relapsing multiple sclerosis (RMS), and estimated its impact on radiologists’ assessments.Patients and methods We retrospectively enrolled 111 RMS patients with two consecutive MRI scans at one year interval, without relapses, progression or radiological changes while remaining on stable therapy. MRI scans were analyzed by four radiologists acting independently, and by icobrain ms. The radiologists performed their assessments both without and with icobrain ms support and categorized patient status based on lesion and brain volume changes.Results Radiologists’ detection of lesion activity non-significantly increased from 6.8% to 8.6% (McNemar p = 0.215; ordinal regression p = 0.06; sign-test p = 0.08) using icobrain ms. Detection of accelerated brain volume loss increased from 2.0% to 7.7% (all p < 0.01). Inter-reader agreement improved from 36% to 56% for lesions, and from 26% to 40% for brain loss (Kendall’s W, p < 0.001). Icobrain ms detected new or significantly enlarged lesions in 16.2% and accelerated brain volume loss in 6.3% of participants.Conclusion AI-assisted MRI volumetry may support radiologists to detect subclinical disease activity in patients characterized as having stable disease.