Mapping human brain lesions and their functional consequences

Abstract

Neuroscience has a long history of inferring brain function by examining the relationship between brain injury and subsequent behavioral impairments. The primary advantage of this method over correlative methods is that it can tell us if a certain brain region is necessary for a given cognitive function. In addition, lesion-based analyses provide unique insights into clinical deficits. In the last decade, statistical voxel-based lesion behavior mapping (VLBM) emerged as a powerful method for understanding the architecture of the human brain. This review illustrates how VLBM improves our knowledge of functional brain architecture, as well as how it is inherently limited by its mass-univariate approach. A wide array of recently developed methods appear to supplement traditional VLBM. This paper provides an overview of these new methods, including the use of specialized imaging modalities, the combination of structural imaging with normative connectome data, as well as multivariate analyses of structural imaging data. We see these new methods as complementing rather than replacing traditional VLBM, providing synergistic tools to answer related questions. Finally, we discuss the potential for these methods to become established in cognitive neuroscience and in clinical applications.

Keywords: Cognitive neurology; Human; Lesion analysis; MLBM; Machine learning; Mass-univariate; Multivariate lesion behavior mapping; Network; Neuroanatomy; Neuropsychology; Non-parametric mapping; Stroke; VLBM; VLSM; Voxel-based lesion symptom mapping.

Figure 1

Spatial distribution of voxel-wise collateral damage in a large, unselected sample of stroke patients. (A) Overlay of 274 unselected, normalized lesions used by Sperber and Karnath (2017). (B) In these 274 patients, all patients with damage to a certain voxel (white circle; voxel 123/94/88 in MNI space) have been identified; an overlay of their lesions (n = 47) is shown. (C) Same as in the panel before, but for another voxel (white circle; voxel 143/154/88 in MNI space; n = 25). Note that collateral damage to both voxels is not centered on the voxel itself, but spatially oriented towards the center of the territory of the middle cerebral artery.

Figure 2

Diffusion tensor tractography in a patient with a cortico-subcortical infarction in the territory of the middle cerebral artery (modified from Staudt, 2010; reprinted with permission from John Wiley and Sons). Left: Coronal T1-weighted image depicting the lesion, leaving only a small bridge of preserved white matter between the lesion and the enlarged lateral ventricle. Transcranial magnetic stimulation (red) indicated preserved crossed cortico-spinal motor projections. Right: Diffusion tensor tractography visualizes the extensive connectivity mediated by this small bridge of preserved tissue (seed area for fiber tracking).

Figure 3

Basic principles of the use of normative resting state functional magnetic resonance imaging (rsfMRI) data in lesion analysis. (A) Individual structural imaging is used to obtain normalized binary lesion maps. (B) Each lesion map is used as a region of interest in a seed-based rsfMRI analysis in a sample of healthy subjects. The resulting topography identifies voxels in which brain activity correlates positively (red) or negatively (blue) with brain activity in the seed region. In the next step, a threshold is applied on this topography to create a binary map (here depicted for positive correlations). (C) Binarized rsfMRI maps obtained from multiple lesions are used in a group analysis of patients with a certain behavioral deficit of interest to identify networks affected by the lesions.

Figure 4

A cartoon illustrating machine learning, offering a multivariate approach to lesion analysis. Consider a study where a portion of the stroke patients have spatial neglect (for example neglecting the petals on the left side when asked to copy a drawing of a flower). Independently, neither the proportion injury to the posterior nor anterior parts of temporal cortex are able to reliably classify the patients’ disorder. This demonstrates the limitation of the mass-univariate approach, where none of the features (e.g., brain areas) are independently strong predictors. However, knowing the injury to both regions can accurately classify patients, as demonstrated by the dotted line (the ‘hyperplane’) which is weighted by damage to both areas in order to predict impairment. In this case, this line accurately classifies most individuals. In reality, these tools are able to use hundreds of features (in the case of region of interest analyses) or thousands of features (in the case of voxelwise analyses), but the resulting multi-dimensional ‘hyperplanes’ are more difficult to visualize graphically.