4142

Can different etiologies provide converging evidence regarding the neural correlates of cognitive performance? Tumor versus stroke
Eva van Grinsven1, Anouk Smits1, Emma van Kessel1, Mathijs Raemaekers1, Edward de Haan2, Irene Huenges Wajer1,3, Veerle Ruijters1, Marielle Philippens4, Joost Verhoeff4, Pierre Robe1, Tom Snijders1, and Martine van Zandvoort1,3
1Department of Neurology & Neurosurgery, University Medical Center Utrecht Brain Center, Utrecht University, Utrecht, Netherlands, 2Department of Psychology, University of Amsterdam, Amsterdam, Netherlands, 3Department of Experimental Psychology and Helmholtz Institute, Utrecht University, Utrecht, Netherlands, 4Department of Radiotherapy, University Medical Center Utrecht, Utrecht, Netherlands

Synopsis

While lesion-symptom mapping can inform on which brain regions are crucial for a given behavior, it is still unclear whether different lesion etiologies show comparable structure-function relationships. In this study, support-vector regression lesion-symptom maps were compared between a glioma and stroke population. As expected, pathology distinct coverage patterns in the brain were found and there were more and larger significant voxel clusters in the tumor group. Our preliminary conclusion is that despite some differences in lesion-symptom associations in comparing a tumor and stroke population, these two populations can provide complementary information regarding involvement of brain regions for given cognitive tasks.

Introduction

Lesion-symptom mapping is a key tool in understanding the relationship between brain structures and behavior and is applied by both researchers and clinicians on a daily basis to assess and predict the functional outcomes after a brain lesion.1 However, the behavioral consequences of lesions from different etiologies may vary as a result of how they affect the brain. For example, ischemic stroke causes relatively immediate cell death within the affected brain area, whereas neural activity can persist after infiltration by a tumor.2 Additionally, lesion distribution differs between etiologies; stroke most often occurs in subcortical areas in the territory of the middle cerebral artery,3 while brain tumors often involve both subcortical and cortical structures in the frontotemporoinsular areas.4 This a-priori location bias is also incorporated in most of the current lesion-symptom knowledge, since this is most often based on a stroke population. Possibly, combining two populations in lesion-symptom mapping can partly circumvent this bias and lead to new insights. To assess whether lesion-symptom maps derived from different populations can provide converging evidence, the current study directly compared lesion-symptom maps between a stroke and a tumor population.

Methods

Data from two different studies were combined. The tumor data were collected in a single-center consecutive study cohort of treatment-naive diffuse glioma (WHO grade 2-4) patients who underwent awake brain surgery between January 2010 and July 2019. The stroke data was gathered as part of a multi-center prospective cohort study in adult patients administered to the hospital between September 2015 and December 2019. Both patient populations had neuropsychological testing and a MRI performed, pre-operatively for the tumor population or within three months after stroke. The administered neuropsychological tests are internationally widely used, standardized psychometric tasks for assessing neurocognitive deficits in the major neurocognitive domains. Raw scores on each test were transformed into z-scores based on published normative data. If applicable, corrections were made based on sex, age group and/or educational level. The tumor and infarct lesions (defined as hyperintense signal abnormalities) were delineated on T2 FLAIR images using the Smartbrush implemented in the iPlan v3.0 software5 or the ITK-snap software,6 respectively. Each individual’s FLAIR and binary lesion mask was then normalized to the MNI template using the unified segmentation-normalization algorithm implemented in SPM12.7 For each cognitive task multivariate lesion-symptom mapping analyses were performed using the updated version of the support-vector regression lesion-symptom mapping toolbox running under Matlab2019a,8 which is a multivariate regression algorithm based on machine learning.9,10 A lesion threshold of three patients was applied to restrict the analysis to those voxels with reasonable statistical power.1 To test the significance of the beta values permutation testing was used with 1.000 permutations and a voxelwise threshold of p<.005. Lesion volume was corrected for by regressing it on both the lesion and behavioral data. The AALCAT atlas was superimposed on the results to relate significant voxels to brain regions.

Preliminary Results & Discussion

196 tumor patients and 146 first-ever cerebral stroke patients were included for the analysis in the current study. As expected, the areas covered by the tumor and stroke lesions showed overlap, but also significant differences in coverage (Figure 1). Despite this difference in lesion distribution, the cognitive pattern did not vary to a great extent (Figure 2). When looking at the lesion-symptom maps, distinct cognitive tasks (i.c. Figure 3 and 4) are associated with distinct neuroanatomical patterns in both groups and follow the expected lateralization based on previous literature. Nonetheless, lesion-symptom maps from the tumor group are not identical to those from the stroke group, with some areas implicated in the tumor, but not in the stroke group or vice versa. Additionally, larger clusters of voxels covering more brain areas are significant in the tumor compared to the stroke group for all eleven cognitive tasks (max. 40 brain areas versus 23). These differences could be explained by several factors. Firstly, not all regions implicated by tumor were part of the lesion coverage in the stroke group, which hinders comparison. Secondly, in the tumor population both tumor and tumor surrounding edema were part of the lesion. Also, diffuse gliomas, specifically lower grade gliomas, infiltrate or displace healthy brain parenchyma without necessarily damaging this tissue. Thus, this parenchyma could still be functional even though it is adversely affected by the tumor presence. Alternatively, tumors can release toxins that may act at a distance and alter functions in areas not directly altered by their growth. Speculatively, the areas indicated in the stroke maps might represent crucial brain areas for that function, while for the tumor the areas indicate involvement.

Conclusion

Lesion-symptom maps based on different etiologies indicated divergent structure-function patterns, with more and larger significant clusters in the tumor group. This could partly be explained by a difference in lesion coverage of the brain and how the brain is affected by either a tumor or stroke lesion. Despite these considerable differences, for both populations cognitively distinct tasks were associated with distinct neuroanatomical brain regions, with the brain regions corroborating those found by previous lesion studies in these populations. Therefore, our preliminary conclusion is that these two populations can provide complementary information regarding involvement of brain regions in different cognitive processes.

Acknowledgements

No acknowledgement found.

References

1. Karnath HO, Sperber C, Rorden C. Mapping human brain lesions and their functional consequences. Neuroimage. 2018;165:180-189.

2. Krainik A, Lehéricy S, Duffau H, et al. Postoperative speech disorder after medial frontal surgery. Neurology. 2003;60(4):587-594.

3. Sperber C, Karnath HO. Topography of acute stroke in a sample of 439 right brain damaged patients. NeuroImage Clin. 2016;10:124-128.

4. Kleihues P, Burger PC, Aldape K, et al. WHO Classification of Tumors of the Central Nervous System. World Heal Organ Classif Tumours. Published online 2007:33-49.

5. BrainLab AG, Feldkirchen, Germany.

6. Yushkevich PA, Piven J, Hazlett HC, et al. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. Neuroimage. 2006;31(3):1116-1128.

7. Rorden C, Bonilha L, Fridriksson J, Bender B, Karnath HO. Age-specific CT and MRI templates for spatial normalization. Neuroimage. 2012;61(4):957-965.

8. The MathWorks, Inc., Natick, Massachusetts, United States.

9. Zhang Y, Kimberg DY, Coslett HB, Schwartz MF, Wang Z. Multivariate lesion‐symptom mapping using support vector regression. Hum Brain Mapp. 2014;35:5861-5876.

10. DeMarco AT, Turkeltaub PE. A multivariate lesion symptom mapping toolbox and examination of lesion-volume biases and correction methods in lesion-symptom mapping. Hum Brain Mapp. 2018;39(11):4169-4182.

Figures

Lesion prevalence maps for the tumor (panel A) and stroke group (panel B) are shown superimposed on the MNI brain in radiological view. The colors refer to the number of patients with a lesion at that voxel, with red indicating a higher number of patients. The maximum overlap is 48 and 17 for the tumor and stroke group, respectively. The MNI brain on the right indicates the location of the slices shown in the figure.

Percentage of patients with a Z-score of -1.5 or lower in the tumor (blue) and stroke group (green) for the eleven cognitive tasks. Asterisks indicate significant differences between the groups, based on the chi-square test. Abbreviations: BNT, Boston Naming Task; RAVLT-IR, Rey Auditory Verbal Learning Task Immediate Recall; DR, Delayed Recall; Recog, Delayed Recognition; ROCFT, Rey Osterrieth Complex Figure Task; DS, Digit Span; TMT, Trail Making Test.

Lesion symptom results for the Boston Naming Task. Lesion overlap indicating those regions in which at least 3 patients had a lesion for the tumor group in blue (N=172), the stroke group in pink (N=103), and overlapping regions shown in the green outlined purple area (Panel A). Voxels significantly associated with performance on this task for the tumor group (red) and stroke group (blue) with the green outline indicating overlapping lesion coverage, as shown in the above panel (Panel B). Maps are shown on the MNI standard brain in radiological view.

Lesion symptom results for the direct copy of the Rey Osterrieth Complex Figure. Lesion overlap indicating those regions in which at least 3 patients had a lesion for the tumor group in blue (N=167), the stroke group in pink (N=68), and overlapping regions shown in the green outlined purple area (Panel A). Voxels significantly associated with performance on this task for the tumor group (red) and stroke group (blue) with the green outline indicating overlapping lesion coverage, as shown in the above panel (Panel B). Maps are shown on the MNI standard brain in radiological view.

Proc. Intl. Soc. Mag. Reson. Med. 29 (2021)
4142