A Comparison of Brain Age Estimation And Brain Parenchymal Fraction as Imaging Markers in Multiple Sclerosis

Abstract

Background and objectives: By integrating information provided by neuroimaging repositories, brain age prediction provides an estimate of the biological age of the brain for the individual participant. The aims of the study were to apply a brain age model on brain MRI data from participants with multiple sclerosis (MS) and to assess clinical associations and compare performance with the established total brain volume and brain parenchymal fraction.

Methods: Participants with MS (n = 1514, 72 % female, n = 1088 longitudinal) and healthy controls (HC) (n = 862, 55 % female, n = 289 longitudinal) were included from Oslo and Karolinska University Hospitals. Structural 3D T1-weighted MRI data were processed using a harmonised pipeline. We estimated brain age based on an independent training set (n = 35474, age range 3 - 89 years). We used linear regression and linear mixed effects models to test for associations between brain age and scanner, disease modifying treatments, Expanded Disability Status Scale (EDSS), MS phenotypes, and disease duration.

Results: The model revealed reliable brain age predictions for the training set (cor = 0.94) and for the HCs at baseline (cor = 0.94). The estimated brain age of participants with MS was on average 6.5 years higher than that of HCs (t = 12.2, p = 2.6 x 10 -34 ). Longitudinal data from participants with MS revealed an accelerated brain ageing of 22 % compared to chronological ageing (t = 6.5, p = 1.0 x 10 -10 ) and significant associations between brain age and both EDSS (t=3.8, p = 1.6 x 10 -4 ) and disease duration (t = 4.3, p = 2.5 x 10 -5 ), with similar effect sizes as those obtained using the established total brain volume and brain parenchymal fraction.

Conclusions: MS participants showed higher brain age compared to HCs and accelerated brain ageing compared to chronological ageing. The clinical associations with disease duration and severity support the further development of the brain age prediction framework to offer an intuitive individual global imaging marker for disease progression and disability in MS.

Note:

Ethics: This project was approved by the Regional Ethics Review Board in Stockholm (ID: 2009/2107-31/2, 2018/2711-32) and the Institutional Review Boards (IRB) at Huddinge Hospital, Stockholm, Sweden (ID: 21/95), the University of Oslo (UiO) and OUH (ID: 2011/1846 A and 2016/102). Study participants provided signed informed consent prior to study enrolment at the respective sites according to the Declaration of Helsinki.

Funding Information: The funders had no role in study design, in data collection, analysis or interpretation, or writing of the manuscript

Declaration of Interests: All other authors declare no competing interests.

Keywords: multiple sclerosis, brain age estimation, machine learning, longitudinal