Quantitative follow‐up of patients with multiple sclerosis using MRI: Technical aspects

Abstract

A highly reproducible automated procedure for quantitative analysis of serial brain magnetic resonance (MR) images was developed for use in patients with multiple sclerosis (MS). The intracranial cavity (ICC) was identified on standard dual-echo spin-echo brain MR images using a supervised automated procedure. MR images obtained from one MS patient at 24 time points in the course of a 1-year follow-up were aligned with the images of one of the time points. Next, the contents of the ICC in each MR exam were segmented into four tissues, using a self-adaptive statistical algorithm. Misclassifications due to partial voluming were corrected using a combination of morphologic operators and connectivity criteria. Index terms: sclerosis, multiple; magnetic resonance (MR), image processing; magnetic resonance (MR), volume measurement; magnetic resonance (MR), surface rendition; brain, volume; brain, white matter MULTIPLE SCLEROSIS (MS) is one of the most challenging diseases for image segmentation of brain structures and therefore provides a good model for the quantitative evaluation of brain disease and its progression. We have developed an image processing system that automatically identifies and quantifies individual MS lesions over time. This paper describes a software system that was developed for the analysis of serial MRI and gives results from the analysis of one MS patient's 24 MRI studies over a 1-year period. By integrating automatic registration and segmentation with partial voluming artifact correction, it became possible to estimate lesion volumes automatically from proton density-weighted (PDw) and T2-weighted (T2w) MR images. Serial brain MRI is important in the study of the natural course of MS and in the evaluation of potential therapeutic agents (1-7). Besides neuroradiologic interpretation, several image processing methods have been applied to quantify lesion burden from PDw and T2w spin-echo images (8,9). Manual outlining of lesions on MR was used initially, in spite of its limited reproducibility and high labor intensity (10-16). The first semi-automated methods to identify and measure the volumes were based on signal intensity properties or image gradients (edges) but still needed significant manual correction, entailing significant loss in measurement reproducibility (15-26).Our group proposed such approaches for the measurement of white matter lesions in MS (22,(27)(28)(29)(30) and started looking for solutions leading to fully automated segmentation algorithms. Edge-preserving noise reduction was introduced in an attempt to enhance the contrast-to-noise ratio between normal and abnormal white matter (31), image registration was proposed to identify individual lesions automatically (28,32), and the first three-dimensional (3D) movies of lesion evolution clearly demonstrated that lesions would swell and regress in a matter of weeks (33). The combination of these tools contributed largely to the assessment of large, well-defined structures (34-37) but lacked the consistency and sensitivity to monito...

Review of automatic segmentation methods of multiple sclerosis white matter lesions on conventional magnetic resonance imaging

“…However, the effects of each preprocessing step might affect each method differently making difficult any generalizations. A few authors (García-Lorenzo et al, 2008c;Kikinis et al, 1999;Wei et al, 2002) have evaluated their methods under different preprocessing pipelines. The comparison of preprocessing steps and their interaction is complex and is beyond the scope of the paper.…”

MRI‐based morphometric analysis of typical and atypical brain development

“…Large subject pools, by necessity, place large demands on the level automation necessary in the image analysis methods. A number of systems with very high levels of automation have been introduced [for example, Dale et al, 1999;Kikinis et al, 1999;Rosas et al, 2002;Watkins et al, 2001]. Calibration studies for these systems have been performed that indicate a high degree of volumetric similarity for results generated with these automated systems in comparison to manual segmentation [Rosas et al, 2002].…”

Landmark-Based Registration Using Features Identified Through Differential Geometry

“…Other artifacts such as chemical shift or susceptibility effects (see for instance [18]) may also account for the observed bias as they influence the detection of extremal points. Indeed, the two echoes are acquired with different receiver RF bandwidth to improve the signal/noise ratio [19]. Therefore, the chemical shift and the susceptibility effect are different in the two echoes.…”