There is a need to clarify the influence of lesion boundary definitions and of the inclusion of incidental infarcts on lesion-based brain-behaviour studies. In general, stroke neuroimaging analysts lack standardized guidelines for how to address incidental infarcts and define lesion borders when manually tracing lesions . While it appears that lesion-based studies to date have generally not accounted for tissue damage beyond the necrotic lesion core, no study to date has examined whether regions of additional injury could in fact have an impact on brain-behaviour relationships. Thus, we sought to investigate whether different lesion boundary definitions and the inclusion of silent infarcts influenced anatomical behavioural localization or lesion volume correlations with a clinical measure of motor impairment.
Our results showed that the volume of the necrotic lesion core (Region 1) was most closely associated with motor impairment scores. Although the volume of the peri-necrotic region showed a weak trend toward significant association with motor impairment scores, analysis of cumulative lesion volumes showed that the addition of the peri-necrotic tissue to the volume of the necrotic lesion core neither strengthened nor weakened the association between lesion volume and motor impairment scores for whole group or subgroup (n = 9) analyses. With regard to our subtraction lesion analysis, previous research suggests that the posterior putamen is involved in motor control [25–27]. Including the peri-necrotic tissue in our subtraction lesion analysis appeared to strengthen our results, as a larger region of more frequent putaminal damage was noted in our subtraction analysis once lesion tracings included the peri-necrotic region (seen in Figure 2).
In general, these results suggest that the correlation between lesion volume and post-stroke impairment may not be strongly influenced by the volume of the peri-necrotic region. However, it may be the location more so than the volume of the peri-necrotic tissue that is important in behavioural-anatomical relationships, since accounting for the volume of this region had an equivocal influence on lesion volume correlations with motor impairment, while accounting for peri-necrotic tissue notably strengthened the anatomical localization of motor deficits in our subtraction analysis. This finding regarding volume-impairment correlations may have been due to the fact that the average volume of the necrotic lesion core was 4 times larger than the volume of the peri-necrotic region, which could have dampened the influence of the peri-necrotic region on correlational analyses; however, including the peri-necrotic region may have impacted our lesion subtraction analysis because increasing the size of lesion tracings may increase the probability that overlap will be observed for lesions that fall within close proximity to one another.
A recommendation that peri-necrotic tissue be included in lesion tracings for lesion-based studies that involve procedures such as subtraction analysis is supported by previous literature, which suggests that this region may in fact remain dysfunctional in chronic stroke patients. For instance, in a functional MRI (fMRI) study of the activation observed in the cortical peri-infarct region during finger tapping, Cramer and colleagues  noted that even among well-recovered chronic stroke patients, activation in the area surrounding the core of the infarct was reduced compared to normal controls, implying that this region may remain dysfunctional post-stroke. Likewise, Seitz and colleagues  postulate that since persisting depression of cerebral metabolism and blood flow beyond the necrotic core of brain infarction has been noted, chronic structural brain lesions revealed on CT and MRI likely reflect only a portion of the total amount of brain tissue affected by ischemia. This evidence, which suggests that the peri-infarct region, and perhaps the region beyond the observable lesion boundaries, can remain dysfunctional in the chronic stage of stroke, lends support to the suggestion that it may be desirable to include the altered peri-necrotic tissue when tracing lesions for anatomical-behavioural localization studies. This view is supported by our subtraction lesion analysis findings.
The inclusion of pre-existing lacunes in the same hemisphere as the index stroke did not notably influence our subtraction lesion analysis - most likely because of their small volume - and the volume of lacunes was not significantly correlated to motor impairment scores, either alone or when added to cumulative lesion volumes. However, this latter finding may stem from the observations that only 29% of patients in our full sample (n = 41) exhibited lacunes, few patients in our sample had small index lesion volumes, and the average volume of the lesion core (Region 1) was substantially larger than Region 3, which may have dampened the potential association between lacunar lesion volumes and clinical measures. Previous investigations suggest that lacunes, which result from the occlusion of a penetrating cerebral artery [29, 30], can have a significant impact on the clinical expression of disease. For example, the seminal work of Snowden and colleagues in The Nun Study suggested that lacunes may play a role in determining the presence and severity of Alzheimer Disease symptoms . Also, the presence of lacunes is indicative of a higher total burden of cerebral small vessel disease, which is thought to be an important prognostic determinant after stroke . Although lacunes have been noted to be important and relevant findings by previous investigators, lacunes did not have a notable influence in our investigation on lesion volume correlations with FM scores nor on subtraction lesion analysis. The use of varying MRI sequences between studies that have examined the clinical implications of lacunes could also be contributing to differences in findings among groups. The influence of lacunes on brain-behaviour relationships could perhaps also depend on the clinical measure of interest; for instance, the burden of lacunes could be more tightly coupled with measures of cognitive impairment than with motor impairment. Consequently, taking into consideration the findings from this investigation and from previous authors, while the inclusion of lacunes may not have a notable impact on lesion volume correlations with clinical measures nor on behavioural-anatomical localization findings, it is likely advisable to at least note the presence of lacunes in patients involved in such studies. As previous literature indicates, the presence of lacunes could be an important variable in predictive statistical analyses.
Lastly, the volume of lesions in the hemisphere opposite to that of the index stroke (Region 4) was not significantly associated with motor impairment scores for whole-group (n = 41) analysis and showed no appreciable overlap for lesion location on subtraction lesion analysis. This latter finding may be due, in part, to the relatively small size of these lesions in patients with severe hemiparesis in our sample. This suggestion highlights a shortcoming of the subtraction lesion analysis technique used in this study. Since patients are dichotomized into 'affected' and 'unaffected' groups for this analysis rather than being evaluated along a behavioural continuum, which is possible when using analysis techniques such as voxel-based lesion symptom mapping (VLSM) , data regarding the potential impact of lesions in both hemispheres may have been lost. Future investigations in this realm with a larger sample size may allow for VLSM analysis. The observation that pre-existing lesions in the opposite hemisphere were not associated with more severe motor impairment may also have been the result of the location of these lesions. For example, fMRI studies have revealed bilateral activation in the cerebellum, basal ganglia, and sensorimotor cortices during active limb movements , which suggests that activity in a distributed bi-hemispheric network is involved in limb motor control. But, if the lesions in our patient population fell outside of these regions associated with active limb movement, then the presence of lesions in both hemispheres may not have notably influenced our results. On the other hand, the results from our sub-analysis of patients with small subcortical lesions lends support to the idea that pre-existing asymptomatic stroke lesions in the cerebral hemisphere opposite to the hemisphere of a clinically overt stroke may indeed be clinically important. These results showed that the association between cumulative lesion volumes and motor impairment scores in our subset of 9 patients with small index strokes was strengthened when lesions in the opposite hemisphere were accounted for. From this portion of our analysis, the most important finding to note is likely that there appears to be variability with respect to the influence of incidental lesions in the opposite hemisphere on brain-behaviour relationships depending on the size of the index stroke.