Lesion Quantification Toolkit: A MATLAB software tool for estimating grey matter damage and white matter disconnections in patients with focal brain lesions

Abstract

Lesion studies are an important tool for cognitive neuroscientists and neurologists. However, while brain lesion studies have traditionally aimed to localize neurological symptoms to specific anatomical loci, a growing body of evidence indicates that neurological diseases such as stroke are best conceptualized as brain network disorders. While researchers in the fields of neuroscience and neurology are therefore increasingly interested in quantifying the effects of focal brain lesions on the white matter connections that form the brain’s structural connectome, few dedicated tools exist to facilitate this endeavor. Here, we present the Lesion Quantification Toolkit, a publicly available MATLAB software package for quantifying the structural impacts of focal brain lesions. The Lesion Quantification Toolkit uses atlas-based approaches to estimate parcel-level grey matter lesion loads and multiple measures of white matter disconnection severity that include tract-level disconnection measures, voxel-wise disconnection maps, and parcel-wise disconnection matrices. The toolkit also estimates lesion-induced increases in the lengths of the shortest structural paths between parcel pairs, which provide information about changes in higher-order structural network topology. We describe in detail each of the different measures produced by the toolkit, discuss their applications and considerations relevant to their use, and perform example analyses using real behavioral data collected from sub-acute stroke patients. We show that analyses performed using the different measures produced by the toolkit produce results that are highly consistent with results that have been reported in the prior literature, and we demonstrate the consistency of results obtained from analyses conducted using the different disconnection measures produced by the toolkit. We anticipate that the Lesion Quantification Toolkit will empower researchers to address research questions that would be difficult or impossible to address using traditional lesion analyses alone, and ultimately, lead to advances in our understanding of how white matter disconnections contribute to the cognitive, behavioral, and physiological consequences of focal brain lesions.

Validating atlas-based lesion disconnectomics in multiple sclerosis: a retrospective multi-centric study

“…In this work, we validated a novel approach to model structural disconnectomes without diffusion imaging in a large retrospective multi-centric study of multiple sclerosis. Similarly to what has been described in (Griffis et al, 2020), the HCP842 tractography atlas (Yeh et al, 2018) was used to extract brain connections transected by lesions, and the remaining connectivity was modelled as a brain graph. Contrary to "lesion network mapping", our method allows to quantify the impact of lesions on the whole structural connectome instead of focusing on specific sub-networks.…”

“…Similarly to the method described in (Griffis et al, 2020), we estimate disconnectomes without diffusion imaging data, using a population-averaged structural tractography atlas, namely the HCP842 tractography atlas (Yeh et al, 2018). The HCP842 atlas was built by averaging the Spin Distribution Function (SDF) in each voxel for 842 healthy subjects from the Human Connectome Project (Van Essen et al, 2012) (372 males, age range between 22 and 36 years old) whose diffusion data was reconstructed in MNI space using a q-space diffeomorphic strategy (Yeh and Tseng, 2011).…”

“…To overcome these limitations, previous studies proposed to model individual structural brain disconnectivity without requiring diffusion imaging (Griffis et al, 2020;Ravano et al, 2020Ravano et al, , 2019. Structural connectivity is approximated using the HCP842 atlas (Yeh et al, 2018), a population-averaged tractography atlas built using 842 healthy controls of the Human Connectome Project (Van Essen et al, 2012).…”

“…Structural connectivity is approximated using the HCP842 atlas (Yeh et al, 2018), a population-averaged tractography atlas built using 842 healthy controls of the Human Connectome Project (Van Essen et al, 2012). Similarly to what was proposed by (Griffis et al, 2020), we estimate individual connectivity loss resulting from white matter lesions by overlapping an automatically generated lesion mask with the tractography atlas and using brain graphs to model an individual structural disconnectome: a brain map representing the affected neuronal connections. In this retrospective multi-centric study, we evaluate the proposed approach, by comparing it as closely as possible with disconnectomes extracted from individual diffusion-based tractography in a subset of patients and we explore its clinical usefulness in reducing the clinico-radiological paradox in multiple sclerosis.…”

Sensorimotor, Attentional, and Neuroanatomical Predictors of Upper Limb Motor Deficits and Rehabilitation Outcome after Stroke

“…Despite univariate and multivariate brain-behavior mapping approaches have been shown to produce highly similar results [24], a bigger sample and the use of multivariate machine learning methods would have strengthened the generalization of our findings. Additionally, the severity of disconnections could be estimated more directly using other methods (e.g., [94]). Future studies should exploit a prospective design to collect information on a broader range of sensorimotor and cognitive skills, as well as multimodal neuroimaging data, to predict motor recovery in a large sample of patients.…”