MRI time series modeling of MS lesion development
Introduction
We present a model-based approach to the pixel-wise assessment of the short-term dynamics of new lesion formation in multiple sclerosis (MS), as apparent from serial T2-weighted magnetic resonance imaging (MRI). Serial MRI provides effective and valuable data for understanding both natural history and treatment effects of neurodegenerative diseases. In MS, the current state of knowledge supposes a dynamic sequence of inflammatory/autoimmune, degenerative, and restorative processes that result in the pathological findings of focal demyelination and remyelination, inflammatory infiltrates, axonal loss and fibrillary astrocytosis. Little is known about the timing, sequence and interplay of these processes leading to both white and gray matter damage. From the MRI perspective, lesion pathogenesis comprises an acute phase with blood–brain barrier leakage lasting a few weeks followed by a chronic phase of T2 hyperintensity (T2 lesions) and occasional T1 hypointensity (black holes). Morphometric assessment of MRI lesion burden is a commonly used surrogate in large-scale clinical trials (Filippi et al., 1995, Jacobs et al., 1996, Li and Paty, 1999, Miller et al., 1999), and long-term trends thereof serve as indicators of disease activity and therapeutic effect (Molyneux et al., 1998). T2 hyperintensity is considered pathologically non-specific, but it remains unclear to what extent different pathological processes could be distinguished based on the sequence and dynamics of the T2 signal. That such dissociative potential exists in serial MRI is illustrated, for example, by a recent study on the evolution of enhancing inflammatory lesions into permanent T1-hypointense lesions (black holes), which showed that less than a quarter of all enhancing lesions became “black holes” and that those were preceded by longer duration of enhancement (Bagnato et al., 2003).
We have previously presented a time series analysis strategy to study longitudinal changes in tissue quality (composition) (Meier and Guttmann, 2003). Based on this work, we here examine the dynamics of new T2 lesions in the context of a pathologically motivated mathematical model. Specifically, we aim to answer the following:
- •
How well is the short-term variation of new T2 lesions explained by a model of two opposing processes (e.g. representative of inflammation/degeneration and resorbtion/repair)?
- •
How long is new T2 lesion activity sustained on average? Does this differ from the duration of contrast enhancement ?
- •
Does the spatio-temporal distribution of new T2 hyperintense lesions exhibit a characteristic pattern?
Section snippets
Image acquisition and preprocessing
Weekly to monthly MRI was obtained from 45 MS patients (64% female) over 1 year. Mean subject age was 39 ± 8 years (25–55 years), with 26 patients classified as relapsing–remitting MS (RRMS), 14 as secondary progressive (SPMS) and 5 as “stable MS” (Weiner et al., 2000). The MRI was obtained at 1.5 T with a conventional spin echo sequence with 3 mm contiguous axial slices (axial spin echo, variable echo multi-planar, PDw/T2w, TE = 30/80 ms, TR = 3000 ms, with 0.93 × 0.93 × 3 mm3 nominal voxel
Results
Results are reported in three sections: (1) how well the model fitted each time profile, (2) the spatial patterns of lesion dynamics represented as feature maps and (3) the statistical distribution of duration and residuals over the examined cohort.
Discussion
The presented model-based analysis of lesion formation presents an attempt to approach structural MRI change as a process rather than in terms of the end result. T2 hyperintensity alone is considered non-specific with respect to the outcome of processes such as inflammation, neurodegeneration and repair. The return to T2 isointensity also is not necessarily equivalent with complete lesion repair (Barkhof et al., 2003) nor is T2 “activity” necessary for diffuse WM changes to occur (Goodkin et
Acknowledgments
This work was supported in part by funding provided by the Harvard Center for Neurodegeneration and Repair (Clinical Trials Pilot Study Grant) and the National Multiple Sclerosis Society (RG 3574-A-1), as well as the National Institutes of Health (R01 NS35142; P41 RR13218-01). We would like to thank Drs. Howard Weiner and Ferenc Jolesz for providing the longitudinal MRI data.
References (17)
- et al.
Local magnetization transfer ratio signal inhomogeneity is related to subsequent change in MTR in lesions and normal-appearing white-matter of multiple sclerosis patients
NeuroImage
(2005) - et al.
Time-series analysis of MRI intensity patterns in multiple sclerosis
NeuroImage
(2003) - et al.
Serial magnetic resonance imaging in multiple sclerosis: correlation with attacks, disability, and disease stage
J. Neuroimmunol.
(2000) - et al.
Evolution of T1 black holes in patients with multiple sclerosis imaged monthly for 4 years
Brain
(2003) - et al.
Remyelinated lesions in multiple sclerosis: magnetic resonance image appearance
Arch. Neurol.
(2003) - et al.
MRI contrast uptake in new lesions in relapsing–remitting MS followed at weekly intervals
Neurology
(2003) - et al.
Quantitative assessment of MRI lesion load in monitoring the evolution of multiple sclerosis
Brain
(1995) - et al.
A serial study of new MS lesions and the white matter from which they arise
Neurology
(1998)
Cited by (31)
Estimation of Multiple Sclerosis lesion age on magnetic resonance imaging
2021, NeuroImageCitation Excerpt :The inability to determine the age of a chronic MS lesion may delay diagnosis (Thompson et al., 2018), limit an accurate assessment of treatment response (Traboulsee et al., 2016), and impede longitudinal study of MS lesion dynamics. A number of studies have described the temporal behavior of lesion signal captured on conventional qualitative MRI (Sweeney et al., 2016; Meier and Guttmann, 2003; Meier and Guttmann, 2006; Ghassemi et al., 2015) or advanced quantitative MRI such as quantitative susceptibility mapping (QSM) (Zhang et al., 2019a, Zhang et al., 2016a). Currently there is much interest in studying the longitudinal behavior of chronic active lesions on MRI.
Imaging of noninfectious inflammatory disorders of the spinal cord
2016, Handbook of Clinical NeurologyPREVAIL: Predicting Recovery through Estimation and Visualization of Active and Incident Lesions
2016, NeuroImage: ClinicalRelating multi-sequence longitudinal intensity profiles and clinical covariates in incident multiple sclerosis lesions
2016, NeuroImage: ClinicalCitation Excerpt :Previous work to characterize the longitudinal behavior of lesions on structural MRI and to further relate these changes to clinical information has only involved single structural MRI sequences. In the work of Meier and Guttmann (2003), Meier and Guttmann (2006) and Meier et al. (2007), longitudinal lesion behavior is characterized only on the intensity normalized proton density (PD) volume, using bi-weekly MRI studies. Although they did not relate these changes to clinical covariates, it was found that the maximal insult within a lesion occurred at the center of the lesion, that lower initial intensity within a lesion was predictive of repair, and that most lesion activity did not last beyond 10 weeks.
Magnetic resonance imaging of cells in experimental disease models
2009, Progress in Nuclear Magnetic Resonance Spectroscopy