cni » projects » image analysis protocol development » new reference frames for characterizing spatial lesion prevalence
new reference frames for characterizing spatial lesion prevalence
Although at present magnetic resonance imaging (MRI) is used clinically to diagnose white matter diseases, assess their progression and activity, and monitor the efficacy of drug therapy, it offers considerable promise for elucidating disease pathogenesis, as well. In multiplre sclerosis (MS), the current state of the art for in vivo evaluation of disease burden is the quantification of lesion volume and whole brain tissue volume (as a measure of atrophy). These methods are being advanced to analyze the spatial distribution of lesions to better characterize their relationship to clinical symptoms [Benson 2002, Sperling 2001, Weiner 2000, Lee 1999, Narayanan 1997]. Despite longstanding histological evidence of the fundamental role of the vasculature in MS [Putnam 1935, Courville 1968, Lassmann 2003, Lucchinetti 2000, Lumsden 1970], minimal work has been done to evaluate this relationship in vivo [Kidd 1999, Tan 2000] and these small studies relied only on visual inspection. Quantitative analysis of the relationship between the vascular architecture of the brain and white matter pathology can potentially be used to discriminate putative subtypes of MS, aid in the differential diagnosis of white matter diseases, and aid in the characterization of lesion subtypes, several of which have been described histopathologically to have a hypoxic component [Lassmann 2002].
We have been developing techniques for visual and quantitative analysis of the geometrical relationship between white matter pathology (imaged with MRI techniques) and the vascular system of the brain (imaged using MR angiography). Although this method has many potential applications, we initially apply it to quantitative analysis of the relationship between lesions in multiple sclerosis and the cerebral blood vessels. After, the MRA and MRI data are registered, the lesion and vessel boundaries are extracted from MR images. The segmented boundaries are used to create surface models of vessels and lesions (Fig. 1). In this initial work, we focus on two issues: a) distances between MS lesions and vessels, and b) distribution of lesions with respect to caliber of vessels. For the former, we compute a distance map of the vessel model, such that each voxel stores its distance vector to the closest vessel (Fig 2). It is used to measure the Euclidean distances between lesions and their closest vessels (Fig 3). For the latter, we compute a radius map, in which each voxel stores the radius of its closest vessel (Fig. 4). It is used to measure the distribution of lesions with respect to the vessel caliber. We compute and analyze relationship between the basic geometrical characteristics of individual lesions and the closest vessels’ locations and calibers (Fig.5).
We intend to apply this technique to analyze a large MS patient population, statistically evaluate the relationship between the geometric parameters of the lesions and vessels, and further to examine the correlations of these derived metrics to clinical, cognitive, or other quantitative MRI variables. In addition to MS, this method can be useful to study vascular relationships to cerebral pathology in other diseases such as cerebrovascular disease and stroke. Deriving novel quantitative variables that describe the geometry of pathology with respect to the vascular architecture may improve our understanding of pathogenesis, diagnosis, and prognostic capabilities. New MRI based classification approaches for white matter disease could be a long-term extension of this work.
Attached media: Figure 1
Simultaneous presentation of lesions’ and vessels’ surface models. Green represents lesions and red represents vessels. This facilitates visual inspection of how lesions are distributed with respect to vessels.
Attached media: Figure 2a , Figure 2b
A gray scale representation of a distance map computed for vessels of one of our subjects. A distance map of an object is a 3D array of voxels, where each voxel stores a vector distance to the closest object voxel. Black represents the distance zero, i.e. the vessel boundary. As the distance from the vessel boundary increases the color gets brighter. a) 2D slice. b) Three 2D slices combined with 3D vessel model (red).
Attached media: Figure 3a , Figure 3b
A distance map combined with a surface model of lesions. a) 3D distance map (3 perpendicular slices) and lesion surface model. b) a 2D slice cut away from a). Green is the boundary of lesion intersecting this slice. Since a voxel color value is equal a distance to the closest vessel, distances from lesions can be read interactively (using 3D Slicer) by pointing to a lesion of interest.
Attached media: Figure 4a , Figure 4b , Figure 4c
A gray scale coded radius map. Black represents zero, white represents maximum radius in this volume (here 24 that is 9.4mm). a) 2D slice overlaid with vessels (yellow). b) 2D slices in 3D space combined with a vessel surface model (red) and lesion surface model (green) c) a 2D slice of a radius map and lesion boundaries intersecting with it. Using 3D Slicer one can read a color value of the pointed voxel that is equal to a radius of the closest vessel.
Attached media: Figure 5a , Figure 5b
Lesion characteristics. a) 3D view of exemplary MS lesions extracted and labeled. Arrows show the value of lesion label. b) Table with geometrical parameters of lesions shown in a). The lesion labels in a) correspond to lesion label values in the table.