Clinical neuroanatomyDisconnection somewhere down the line: Multivariate lesion-symptom mapping of the line bisection error
Introduction
The line bisection task is a widely used test in the diagnosis of spatial perception deficits after stroke. Originally, this task was introduced to assess visual field defects (for a review see Kerkhoff & Bucher, 2008) and later adopted in the diagnosis of spatial attention deficits (Schenkenberg et al., 1980), where it quickly became established as a routine test in neuropsychological test batteries (e.g., Halligan et al., 1991; Vaes et al., 2015). In the line bisection task, the patient is asked to manually mark the midpoint of a horizontally presented line. A deviation from the true midpoint to the ipsilesional side is typically seen as a sign of post-stroke deficits in spatial perception.
The neural correlates of line bisection errors (LBE) have been the subject of several studies that utilised either statistical lesion behaviour mapping (Kenzie et al., 2015; Molenberghs & Sale, 2011; Thiebaut de Schotten et al., 2014; Toba et al., 2018, 2017; Verdon et al., 2010) or descriptive topographical methods (Binder et al., 1992; Golay et al., 2008; Rorden et al., 2006). Most often, LBEs have been associated with damage to the posterior parietal lobe (Binder et al., 1992; Kenzie et al., 2015; Molenberghs & Sale, 2011; Rorden et al., 2006; Thiebaut de Schotten et al., 2014; Toba et al., 2017, 2018; Verdon et al., 2010). Other critical regions were found in the posterior part of the temporal lobe or the temporo-parietal junction (TPJ) (Kenzie et al., 2015; Rorden et al., 2006), frontal lobe (Thiebaut de Schotten et al., 2014), and parts of the occipital lobe (Binder et al., 1992; Kenzie et al., 2015; Rorden et al., 2006; Toba et al., 2018, 2017). Further, several studies have suggested a critical role of white matter damage (Golay et al., 2008; Thiebaut de Schotten et al., 2014; Toba et al., 2017, 2018; Verdon et al., 2010). The relevance of white matter tracts has also been highlighted by studies using either fibre tracking (Vaessen et al., 2016) or region of interest-based multivariate lesion analysis in left hemisphere stroke patients (Malherbe et al., 2018). Especially damage to the superior longitudinal fasciculus (SLF) (Malherbe et al., 2018; Thiebaut de Schotten et al., 2014; 2005; Toba et al., 2017, 2018; Vaessen et al., 2016) and the arcuate fasciculus (Malherbe et al., 2018; Thiebaut de Schotten et al., 2014) was found to underlie LBEs. Subcomponents of the SLF connect frontal areas, such as the middle frontal gyrus and pars opercularis, with parietal areas, such as the angular gyrus and the supramarginal gyrus (Thiebaut de Schotten et al., 2011a; Wang et al., 2016).
In conclusion, while there is considerable correspondence between findings in previous studies, it is not yet possible to find a unifying theory. A likely explanation for the different findings is that rather than damage to a single anatomical module, damage to a network underlies LBEs. The relevance of brain connectivity for most cognitive functions is well known (e.g., Godefroy et al., 1998; Catani & Ffytche, 2005) and has also been postulated to be relevant for spatial attention (Bartolomeo, 2006; Bartolomeo et al., 2007; Karnath, 2009; Karnath & Rorden, 2012). However, two exceptions aside (Malherbe et al., 2018; Toba et al., 2017), previous investigations used univariate topographical mapping approaches to investigate the neural correlates of line bisection. These come with methodological caveats in the identification of complex neural correlates of behavioural functions, especially with respect to brain networks (see Sperber, 2020; Sperber et al., 2019b). The two multivariate topographical studies both used an analysis approach based on game theory and included only a few brain regions of interest at once. Such a priori feature reduction by testing only a few possibly crucial hubs can be necessary due to computational or statistical limitations. However, the selected parcellation can differ from the relevant functional parcellation of the brain as well as the typical anatomy of stroke lesions. In contrast, voxel-wise analysis approaches are able to maximize an analysis' ability to identify neural correlates that are not expected, or that do not fully correspond to the brain parcellation provided by an anatomical atlas.
In order to integrate findings of previous investigations into a bigger, coherent picture, the present study thus aimed to identify possible networks underlying the line bisection task, using machine learning-based, multivariate voxel-wise analysis approaches. First, we used structural lesion data and conducted a multivariate lesion-behaviour mapping analysis to detect areas where focal damage might directly induce LBEs. This approach is powerful in identifying complex configurations of neural correlates such as in brain networks (Mah et al., 2014; Sperber et al., 2019b; Zhang et al., 2014). Second, by quantifying virtual white matter disconnection related to lesion location, i.e., depicting white matter connections in healthy brains running through patients' lesion areas, we further aimed to investigate remote pathological processes.
Section snippets
Patient recruitment
The sample consisted of 172 neurological patients admitted to the Centre of Neurology at Tuebingen University, which were screened for a first ever right-hemisphere stroke. The sample size was determined following recommendations from Sperber et al., 2019a, Sperber et al., 2019b, concluding that sample sizes of at least 100–120 subjects are required to optimally model voxel-wise lesion location in SVR-LSM. We excluded patients with diffuse or bilateral brain lesions, patients with tumours, as
Parameter optimization
The parameter optimization routine for the structural lesion maps revealed an optimum C = 10 and γ = 2 which resulted in an average cross-validation prediction accuracy r = .25 and Reproducibility = .85. For the disconnection maps we achieved a similar model performance as for the lesion maps of prediction accuracy r = .30 and Reproducibility = .91, by using C = 30 and γ = 9.
Prediction accuracy appeared to be smaller than in previous publications using the technique (Sperber et al., 2019a;
Discussion
The present study investigated the neural underpinnings of ipsilesional rightward line bisection deviation in acute right hemispheric stroke. We used both multivariate mapping of lesion maps to characterise direct structural damage as well as multivariate mapping of disconnection metrics to reveal additional remote effects of right-hemispheric lesions in both hemispheres.
Conclusion and perspective
Our findings underline the importance of a network including several cortical nodes and intra-as well as interhemispheric connections in the emergence of the line bisection error. The use of support vector regression based lesion-symptom disconnection mapping revealed that we might miss relevant structures and connections when we only focus on focal damage. However, according to recent findings by McIntosh et al. (2017), the traditional interpretation of the line bisection task might produce a
Data and code availability statement
The datasets generated and analyzed during the current study are not publicly available due to the data protection agreement of the Centre of Neurology at Tübingen University, as approved by the local ethics committee and signed by the participants. We provide the scripts of the main analyses, as well as the statistical topographies and overlap maps, available at https://doi.org/10.17632/2hyhk44zrj.2.
CRediT authorship contribution statement
Daniel Wiesen: Conceptualization, Writing – original draft, Formal analysis, Methodology, Investigation. Christoph Sperber: Conceptualization, Writing – original draft, Methodology, Investigation. Hans-Otto Karnath: Conceptualization, Writing - review & editing.
Acknowledgments
This work was supported by the Deutsche Forschungsgemeinschaft (KA 1258/23-1). Daniel Wiesen was supported by the Luxembourg National Research Fund (FNR/11601161). The authors have no competing interests to declare.
References (78)
- et al.
A reproducible evaluation of ANTs similarity metric performance in brain image registration
Neuroimage
(2011) - et al.
Finding centre: Ocular and fMRI investigations of bisection and landmark task performance
Brain Research
(2012) - et al.
Horizontal space misrepresentation in unilateral brain damage. I. Visual and proprioceptive- motor influences in left unilateral neglect
Neuropsychologia
(2002) - et al.
Subjective experience of inner speech in aphasia: Preliminary behavioral relationships and neural correlates
Brain and Language
(2017) - et al.
Damage to white matter bottlenecks contributes to language impairments after left hemispheric stroke
NeuroImage Clinical
(2017) - et al.
The anatomy of spatial neglect
Neuropsychologia
(2012) - et al.
Mapping human brain lesions and their functional consequences
Neuroimage
(2018) - et al.
Line bisection as an early method to assess homonymous hemianopia
Cortex
(2008) - et al.
Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration
Neuroimage
(2009) - et al.
Anatomical predictors of successful prism adaptation in chronic visual neglect
Cortex
(2019)
Neural correlates of visuospatial bias in patients with left hemisphere stroke: A causal functional contribution analysis based on game theory
Neuropsychologia
Weight and see: Line bisection in neglect reliably measures the allocation of attention, but not the perception of length
Neuropsychologia
Weights and measures: a new look at bisection behaviour in neglect
Cognitive Brain Research
The ins and outs of meaning: Behavioral and neuroanatomical dissociation of semantically-driven word retrieval and multimodal semantic recognition in aphasia
Neuropsychologia
Age-specific CT and MRI templates for spatial normalization
Neuroimage
Disturbed line bisection is associated with posterior brain lesions
Brain Research
A simple measure of neglect severity
Neuropsychologia
Rethinking causality and data complexity in brain lesion-behaviour inference and its implications for lesion-behaviour modelling
Cortex
Diagnostic validity of line bisection in the acute phase of stroke
Neuropsychologia
Topography of acute stroke in a sample of 439 right brain damaged patients
NeuroImage Clinical
A network underlying human higher-order motor control: Insights from machine learning-based lesion-behaviour mapping in apraxia of pantomime
Cortex
Atlasing location, asymmetry and inter-subject variability of white matter tracts in the human brain with MR diffusion tractography
Neuroimage
Common brain networks for distinct deficits in visual neglect. A combined structural and tractography MRI approach
Neuropsychologia
Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain
Neuroimage
Structural white-matter connections mediating distinct behavioral components of spatial neglect in right brain-damaged patients
Cortex
Using machine learning-based lesion behavior mapping to identify anatomical networks of cognitive dysfunction: Spatial neglect and attention
Neuroimage
Sensitivity of clinical and behavioural tests of spatial neglect after right hemisphere stroke
Journal of Neurology, Neurosurgery, and Psychiatry
A parietofrontal network for spatial aware- ness in the right hemisphere of the human brain
Archives of Neurology
Left unilateral neglect as a disconnection syndrome
Cerebral Cortex
The control of the false discovery rate in multiple testing under dependency
Annals of Statistics
Distinct syndromes of hemineglect
Archives of Neurology
The rises and falls of dis- connection syndromes
Brain
LIBSVM : A library for support vector machines
ACM Transactions Intelligence System and Technology
Neural basis and recovery of spatial attention deficits in spatial neglect
Nature Neuroscience
Support-vector networks
Machine Learning
Exploring the syndrome of spatial unilateral neglect through an illusion of length
Experimental Brain Research
A multivariate lesion symptom mapping toolbox and examination of lesion-volume biases and correction methods in lesion-symptom mapping
Human Brain Mapping
Misrepresentation of horizontal space in left unilateral neglect: Role of hemianopia
Neurology
Causes of cross-over in unilateral neglect:between-group comparisons,within-patient dissociations and eye movements
Brain
Cited by (16)
Structural disconnections associated with language impairments in chronic post-stroke aphasia using disconnectome maps
2022, CortexCitation Excerpt :Disconnectome-symptom mapping analyses were performed using support vector regression (SVR-DSM) to investigate the relationship between individuals’ language scores and the probability of disconnection. Support vector regression has been used and validated as a multivariate method to model lesion-symptom associations in multiple lesion-symptom mapping studies (DeMarco & Turkeltaub, 2018; Fama, Hayward, Snider, Friedman, & Turkeltaub, 2017; Griffis, Nenert, Allendorfer, & Szaflarski, 2017; Mirman, Kraft, Harvey, Brecher, & Schwartz, 2019; Mirman, Zhang, et al., 2015; Wiesen, Karnath, & Sperber, 2020). Instead of investigating brain-behavior relationships at the voxel-level such as in traditional mass-univariate voxel-based lesion-symptom mapping (VLSM) analyses, this multivariate method uses a high dimensional feature space to evaluate the entire brain-behavior association simultaneously (Zhang, Kimberg, Coslett, Schwartz, & Wang, 2014).
Voxelwise structural disconnection mapping: Methodological validation and recommendations
2022, NeuroImage: ClinicalCitation Excerpt :The first methodological proposal for voxelwise disconnection mapping has been made by Foulon and colleagues (Foulon et al., 2018). This approach is available within the BCBtoolkit (https://www.toolkit.bcblab.com) and has been increasingly used in the recent years (e.g. Alves et al., 2021; Pacella et al., 2019; Salvalaggio et al., 2020; Wiesen et al., 2020). It relies on deterministic fiber tracking seeding from individual lesions (Thiebaut de Schotten et al., 2011).