Infarct topography and functional outcomes

Motor recovery

The characterization of motor recovery in AIS is well studied, which is likely a reflection of the number of quantitative assessment tools for measuring motor performance. As a result, our understanding of the role of infarct topography in motor outcomes is more advanced than other areas of stroke recovery. The involvement of specific brain regions in AIS clearly and logically has a strong influence on motor recovery (Figure 1). Comparing cortical versus mixed or subcortical lesions, there is a strong correlation with upper limb motor recovery after AIS. In a study of stroke rehabilitation patients with pure cortical compared to subcortical strokes, 3 out of 4 patients in the cortical group versus 1 out of 17 in the subcortical group had recovery of upper limb motor function of it. ⁶⁰ Importantly, only 1 of 28 patients with radiographic evidence of involvement of the posterior limb of the internal capsule had recovery of upper limb movement compared to 5 out of 13 patients with sparing of it. ⁶⁰ One explanation for these findings is that the lesion load of the corticospinal tract portends motor recovery potential.^61,62

In addition to injury of the corticospinal tract, infarction of specific regions of the cortex may also strongly influence motor recovery. Acute involvement of the somatosensory cortex, intraparietal sulcus, and primary motor cortex, evidenced by restricted diffusion on DWI, are associated with worse motor recovery and functional outcomes after AIS.^63,64 Another study showed that CT perfusion within 9 h of stroke onset identified multiple brain regions including the insular cortex, superior temporal gyrus, postcentral gyrus, putamen, caudate, and internal capsule that were all independent predictors of motor recovery. ⁶⁵ Involvement of the putamen has also been shown to increase the likelihood of a residual gait disorder following AIS. ⁶⁶

Voxel-based lesion-symptom mapping (VLSM) approaches have also shown value in understanding the role of infarct topography on motor recovery. VLSM compares neurobehavioral scores between patients with and without lesions on a voxel-wise basis. ⁶⁷ VLSM analysis showed that the areas most associated with worsening motor performance are situated at the junction of the corona radiata and corticospinal tract. ⁶⁸ In a study of 50 patients with recent ischemic or hemorrhagic stroke, while corticospinal tract lesion load volume on FLAIR MRI sequences performed three to eight weeks post-event was predictive of performance in two functional mobility scales, using VLSM, walk speed response to gait rehabilitation was predicted by damage to the putamen, external capsule, and insula. ⁶⁹

Not surprisingly, the extent of ischemic injury to the components of the motor pathway will affect motor recovery. These regions include the eloquent portions of the cortex involved in motor function as well as the corticospinal tract as it converges within the corona radiata and descends in the internal capsule. Using clinical variables alone to prognosticate motor recovery post-stroke is only moderately effective. ⁷⁰ Incorporating region-specific or tract-specific involvement, however, is likely to improve predictions of motor recovery.

Recovery of language

In AIS, aphasia is frequently the result of ischemic injury to the language centers of the dominant hemisphere. Post-stroke aphasia is exceedingly common, as 20–40% of AIS patients will present with aphasia as an initial symptom.^71–73 While the potential for motor recovery is heavily weighted by the extent of ischemic injury to the corticospinal tracts, the potential for recovery of language after AIS is dependent on the cortical involvement of the language centers. AIS infarct burden in language regions is an important predictor of aphasia recovery potential. ⁷⁴

Multiple studies have highlighted the importance of the superior temporal gyrus (STG) and, in particular, the posterior portion, in aphasia recovery.^75–78 In one study, the severity of auditory comprehension deficits was strongly correlated with the extent of injury to the posterior STG: patients with damage to less than half of the STG, detected on CT scans acquired six months post-stroke, were more likely to have good comprehension at six months. ⁷⁸ Lesions of the dominant hemisphere STG also appear more likely to cause a persistent global aphasia as opposed to involvement of the inferior frontal gyrus or pre- and postcentral gyrus. ⁷⁶ These studies suggest that the degree of sparing of the STG is critical for functional aphasia recovery. In addition, in a study of 97 patients with aphasia secondary to left hemisphere AIS, patients with exclusively subcortical involvement (e.g. basal ganglia) had less severe aphasias than those patients with cortical lesions involving Broca’s and Wernicke’s areas. ⁷⁹

Perfusion imaging and, in particular, perfusion-diffusion mismatch (PDM) has also been shown to have prognostic utility in aphasia recovery. The presence of PDM in Brodmann area 37 (left posterior inferior temporal cortex) on MRI 24 h post-stroke, for example, independently predicted the degree of acute improvement in naming performance in a population of patients with acute left hemisphere ischemic strokes. ⁵¹ In another population of 58 patients with aphasia secondary to AIS, CT perfusion imaging within 9 h of symptom onset demonstrated near-normal to hyperemic relative cerebral blood flow values in the left angular gyrus and insular cortex were independent predictors of aphasia improvement by hospital discharge. ⁸⁰ Moreover, when these variables were incorporated into a model also containing the admission NIHSS aphasia score and presence of a proximal cerebral artery occlusion, there was 91% accuracy in the prediction of aphasia outcomes. These findings would suggest that preserved or recovered perfusion in specific brain regions inform on the presence of salvageable tissue and as a result, could be useful in predicting functional aphasia recovery.

In summary, the potential for language recovery after AIS is influenced by the involvement of different cortical regions including the STG as well as potentially the angular gyrus and insular cortex of the dominant hemisphere. Incorporating the presence or absence of injury of these brain regions into prediction models with standard clinical variables could offer improved prognostication for post-stroke recovery of language deficits.

Cognitive outcomes

Following an ischemic stroke, 10% of patients will develop secondary dementia. ⁸¹ Understanding the factors that influence the development of post-stroke dementia is of major value in counseling patients and families on recovery expectations. The clinical determinants of post-stroke cognitive dysfunction seem to be heavily influenced by cortical injury or hemispheric involvement.^82,83 In one study of 190 patients with first stroke, 74% of patients with cortical stroke compared to 46% of patients with subcortical stroke had evidence of cognitive impairment. ⁸⁴ Another study of patients with AIS performed neuropsychiatric testing and observed that cognitive dysfunction was associated with the specific arterial territory involved and presence of white matter hyperintensity, but not the laterality of the ischemic stroke. ⁸⁵

VLSM have also been used to identify region-specific predictors of cognitive outcomes. One group used VLSM analysis to identify eloquent voxels in patients with AIS using DWI obtained between 24 and 72 h after onset, which were found predominantly in the left hemisphere prefrontal, cingulate, peri-insular, middle, superior temporal cortex, amygdala, hippocampus, and deep nuclei. ⁸⁶ Furthermore, they demonstrated that infarct location was the strongest predictor of good cognitive outcomes (MoCA> 25). ⁸⁶

White matter tract integrity may also be informative in prognosticating cognitive outcomes after ischemic stroke. In a study of 14 right hemispheric stroke patients and 18 healthy controls, diffusion tensor imaging maps were compared between patients with good and poor cognitive recovery, where good recovery was defined as improvement in neuropsychological test results administered at three months post-stroke compared to results from 72 h. ⁸⁷ The poor cognitive recovery group showed decreased contralesional axonal integrity at three months post-stroke in several left hemisphere regions when compared to healthy controls. ⁸⁷ This data highlight the contributions of infarct topography as well as white matter tract integrity on post-stroke recovery of cognitive function. Moving forward, the use of standardized advanced imaging techniques to characterize infarct topography combined with outcome measures representative of the NINDS CDEs could improve the accuracy of prognostication.

To further understand the specific effects of infarct topography on post-stroke outcomes, it is important to consider the potential downstream effects of the infarct on brain functional network activity. A novel approach that has recently been used to explore these points is lesion-based functional connectivity network analysis. ⁸⁸ By evaluating the blood-oxygen-level-dependent time course in regions of the brain exhibiting lesions in resting state functional MRI from healthy controls, common networks of brain regions were found to be involved in specific clinical syndromes such as peduncular hallucinosis, post-stroke pain and subcortical aphasias despite heterogeneous location of lesions. ⁸⁸ The promise of this approach is that the inclusion of network analysis into infarct topography analysis could augment prediction models of more sophisticated outcome measures.

Volumetric analyses

For intravenous (IV) recombinant tissue plasminogen activator (tPA) trials, acute lesion volumes on CT or MRI greater than 1/3 MCA territory are often used as exclusion criteria.^91–96 Other MRI approaches proposed for trials investigating IV tPA in extended windows are PDM, LVO site and FLAIR-diffusion mismatches. ⁹⁷ Additional details can be found in another review article. ⁹⁷ Although studies have shown that CDM and PDM are closely related, ⁹⁸ PDM was found to be more accurate for selecting patients likely to benefit from reperfusion therapy in the 3–6-h window, ⁹⁹ with CDM patients showing no increased benefit from thrombolysis.^99,100 Because of these findings and known limitations of the clinical score (i.e. NIHSS) from its weighting of left hemispheric infarcts and lesions involving the motor pathways as described above, selection of patients for extended time window treatment based on CDM, although easier to execute, is currently not clinically indicated. However, there is a clinical thrombectomy trial underway investigating the utility of CDM in patients with known MCA LVO. ¹⁰¹

In recent thrombectomy trials, infarct volume had a large effect on likelihood of good outcome (mRS< 3). In one study of 107 patients with anterior circulation LVO that underwent endovascular thrombectomy (EVT), follow-up infarct volume of 40 to 50 cm³ (at median time of 41.8 h from symptom onset and assessed with CT in 58.9% of the population) had the highest accuracy for predicting good outcome (sensitivity 74.1–81.5%; specificity 77.5–85.0%). ¹⁰² Aside from infarct volume, the early clinical response to EVT appears to be highly predictive of functional outcome. In one study, the authors looked at markers of early stroke severity, including infarct volume and baseline NIHSS, and showed that the trajectory of the two-day longitudinal NIHSS revealed subgroups of patients with large, minimal, and no improvement following EVT. ¹⁰³ Not surprisingly, the subgroup with large improvement was more likely to have good outcomes at 90 days and had an accuracy of 84.5% in predicting 90-day mRS. ¹⁰³ Infarct volume continued to have a role, as the “large improvement” subgroup was represented by younger patients with larger regions of hypodensity on acute CT. In a single center MRI study, ¹⁰⁴ patients classified as likely to benefit from EVT (DWI volume< 70 cm³, age< 80 and pre-stroke mRS< 2), were also found to have more favorable outcomes after treatment. In this study, DWI lesion volume was estimated at the MRI console using measurements from three perpendicular axes and an assumed ellipsoid geometry (ABC/2). ¹⁰⁵

In the EVT trials, ischemic core volume was one criteria; however, perfusion mismatch ¹⁰⁶ and collateral status ¹⁰⁷ were incorporated into the patient selection algorithm for several trials. In ESCAPE, delayed-phase CTA was employed to characterize collateral status and infarct core size as part of the trials inclusion criteria for potential EVT. ¹⁰⁷ EXTEND-IA and SWIFT PRIME used CT perfusion (CTP) for CT sites or PDM for MRI sites for patient selection. CTP was used to determine the presence of a target mismatch profile by defining the ischemic core as tissue with relative CBF less than 30% of normal tissue and the “penumbra” as tissue with Tmax greater than 6 s. ^106,108 The use of CTP to represent the ischemic core in these and other studies, however, is somewhat controversial as it is unclear whether reductions in relative perfusion metrics represent tissue infarction or mere hypoperfusion that can be reversed with revascularization. ¹⁰⁹ Nonetheless, in patients with a target mismatch profile, reperfusion therapy was strongly associated with a favorable clinical response defined as NIHSS score of 0 to 1 or a ≥ 8-point improvement on NIHSS at 90 days.^110,111 DEFUSE-3¹¹² is an ongoing prospective randomized Phase III multicenter controlled trial addressing whether a target mismatch profile can be used to select patients likely to benefit from endovascular treatment in an extended time window.

Aside from using anterior circulation LVO as an inclusion criterion, infarct topography was not utilized in the decision process of patient eligibility for EVT trials. Instead patient selection was based on lesion volumes facilitated by the use by the use of either automated software ¹¹³ or visual assessment (ABC/2). ¹⁰⁵ Future studies are needed to compare automated with manual approaches for patient selection for revascularization therapies.

Region-of-interest analyses

The Alberta Stroke Program Early CT Score (ASPECTS), ¹¹⁴ which was originally created to assess early acute ischemic injury and is incorporated into many decision models for pursuing endovascular therapy, is one approach that is used to link specific acute infarct locations to long-term stroke outcomes.^115–117 Multiple studies have shown that integrating ASPECTS score into prediction models might be clinically useful. In one study, ASPECTS score was evaluated in data from the original NINDS tPA study and lesions involving the primary motor cortex/parietal lobe (M6) and lentiform nucleus increased the likelihood of disability (mRS> 2). ¹¹⁶ A simplified ASPECTS (sASPECTS) score, which scored only the caudate, lentiform nucleus, insula, and M5 region lesions, produced similar results to the ASPECTS score, and was found to be an independent predictor of three-month mRS> 2 in patients with anterior circulation AIS. ¹¹⁸

Although ASPECTS is traditionally performed on non-contrast CT scans, ASPECTS has also been applied to DWI. One study compared the performance of CT-ASPECTS and DWI-ASPECTS in a large EVT cohort. ¹¹⁹ In 74 patients, the inter-rater agreement for CT-ASPECTS and DWI-ASPECTS was 0.58 and 0.87, respectively. DWI-ASPECTS correlated with functional outcome while CT-ASPECTS did not. Both CT- and MRI-based ASPECTS, however, correlated with DWI volume but DWI-ASPECTS was superior. ¹¹⁹ In a cohort of patients with AIS treated with intravenous tPA that underwent CT and MRI in the hyperacute stage <3 h), the accuracy (with 1 being the highest accuracy) for CT-ASPECTS was 0.62 and DWI-ASPECTS 0.64 for predicting mRS < 3 at 90 days. ¹²⁰ Comparing CT- to DWI-ASPECTS in AIS patients within 3 h of onset, the detection rate was significantly higher for DWI than CT (76.9% vs. 30.0%; p < 0.01). ¹²¹ These results suggest the sensitivity for detecting early ischemia is greater with MRI. However, this may be due to later acquisition of DWI than CT, since typically MRI is performed later than CT scans.

One group showed that in 213 patients with intracranial internal carotid artery, M1 or M2 middle cerebral artery occlusions who underwent EVT, on multivariable logistic regression, only M4 and M6 involvement on DWI ASPECTS, obtained < 72 h from symptom onset, were associated with poor outcomes (3 month mRS > 2). ¹¹⁷ Interestingly, in right hemispheric strokes, M6 involvement independently predicted poor outcome (odds ratio 5.8; 95% confidence interval 1.9–20.3), whereas in left hemispheric strokes, M4 involvement was a predictor of poor outcomes (odds ratio 4.3; 95% confidence interval 1.3–15.0). ¹¹⁷

The advantage of ASPECTS is that in trained professionals, it can quickly be calculated on CT or DWI performed in the acute stages and, appears to be associated with long-term functional outcomes. There are also efforts to automate ASPECTS calculation, ¹²² which will simplify the approach further. The ASPECTS score, either manual or automatically determined, has not yet been found to be able to select patients likely to benefit from endovascular therapy. ¹²³ The role of ASPECTS in most current trials is predominantly to exclude patients who present with large strokes (e.g. ASPECTS < 7) as opposed to focusing on particular regions that are involved (e.g. lentiform nuclei).

Voxel-based analyses

Because of the poor sensitivity of CT in the hyperacute stage, voxel-based approaches using CT have been rarely used and therefore discussion will be limited to MRI. As mentioned earlier, acute infarct volume only correlates moderately with AIS outcomes.^8–10 The integration of lesion location into models predicting functional outcomes after AIS offers promise for improving the accuracy of long-term prognostication models. An early example is a hazard atlas, which utilized discharge or chronic imaging lesion location combined with an “expert” atlas to predict NIHSS. ¹⁵ Another study used penalized logistic regression based on chronic imaging lesion location and extent to predict NIHSS. ¹²⁴ These preliminary proof-of-concept studies linking lesion location with degree of disability has motivated the development of new techniques aimed at characterizing the specific role of acute infarct topography on global functional outcome metrics beyond laterality or specific functional domains.

VLSM has also been used to assess long-term global functional outcomes using acute imaging. In 101 patients with middle cerebral artery infarcts and two to three day DWI, greater disability as measured on the one-month mRS were associated with acute injury to the corona radiata, internal capsule, and insula, with asymmetric impact patterns found with respect to injury to the right angular gyrus and left superior temporal gyrus. ¹²⁵ In this study, however, lesion volumes were not taken into consideration. Another study involving 490 AIS subjects who had DWI obtained within 48 h of last known well included lesion volume in VLSM analysis to investigate the role of acute lesion topography in AIS severity and long-term functional outcome. ¹²⁶ Figure 4 shows the T-score results, on a voxel-by-voxel basis, of the effect of injury to each voxel on three to six month mRS. Adjusting for age and gender (Figure 4(b)), larger regions of tissue in the right hemisphere were implicated with worse mRS, while the opposite was true for the left hemisphere compared to unadjusted results (Figure 4(a)). When also adjusting for lesion volume (Figure 4(c)), injury to smaller regions within only the left hemisphere (e.g. corona radiata, internal capsule, postcentral gyrus, putamen, and operculum) was found to be independently associated with poor mRS. That is, patients with acute infarcts in these regions (independent of age, sex and lesion volume) were likely to have more disability at three to six months, as measured by mRS, than patients who did not have lesions affecting those regions. Similar results were observed when the analysis was limited to only patients who were still alive at six months post-stroke (Figure 4(d)). Moreover, the authors showed that left hemisphere injury, especially to the posterior limb of the internal capsule and white matter tracts was also associated with higher admission NIHSS. ¹²⁶ VLSM has also been used to identify critical regions in functional outcomes after cerebellar ischemic strokes by examining injury at two weeks on 3D T1-weighted image ¹⁷ or within 72 h. ¹⁸ Both studies showed that in patients with impaired motor performance after cerebellar stroke lesions were more common in the paravermal lobules IV/V, deep cerebellar nuclei, and the middle cerebellar peduncle.^17,18 OPEN IN VIEWER

The use of the ASPECTS score and VLSM represents two approaches to incorporate acute lesion location into predictions of long-term outcomes at the acute stage. Early identification of patients at risk of poor recovery can potentially be used for selection of patients for clinical trials, aggressive intervention and focused post-stroke rehabilitation programs. Although ASPECTS on CT or MRI is becoming frequently used for selection of patients for thrombectomy in clinical trials, ¹⁰¹ anatomical territories identified by VLSM analyses as regions linked with poor long-term outcomes if infarcted are not currently being used. Indeed, these VLSM identified regions overlap the “clinically relevant penumbra” found by others to determine functional outcome after thrombolysis more so than volume of salvaged tissue.^127,128 Future prospective studies are needed to validate both techniques for their utility in patient selection for revascularization therapy.