The new insights into human brain imaging after stroke

3.1 Voxel-based lesion-symptom mapping

Voxel-based lesion-symptom mapping (VLSM) is an imaging technique aimed at evaluating the relationship between the symptom and the tissue damage, on a voxel-by-voxel basis (Bates et al., 2003). This approach has the advantage to determine which region is crucial for a given function, allows the assessment of all brain parenchyma without focusing on a specific region of interest, and allows the assessment of continuous data, without grouping patients by lesion side or clinical score cut-off. However, the results should be interpreted cautiously because the relationships between clinical performances and lesions are based on vascular territories in this context of stroke, whereas functional areas may not fully match vascular areas. In addition, there are some pitfalls to avoid at each step of the VLSM technique, such as lesion delineation, spatial normalization with correction for the lesion, choice of the right template, threshold determination for sufficient overlap, and correction for multiple comparisons (de Haan & Karnath, 2018).

The recent development of VLSM has contributed to identify the most relevant locations involved in different domains of disabilities. Using VLSM, spatial neglect has been associated with posterior parietal (supramarginal gyrus, angular gyrus, and temporal-parietal junction) and frontal cortices (Thiebaut de Schotten et al., 2014), together with deep temporal lobe (Verdon, Schwartz, Lovblad, Hauert, & Vuilleumier, 2010), in right-side hemispheric strokes, while motor deficit has been confirmed to be mainly related to lesions in close vicinity to the corticospinal tract (Dinomais et al., 2015). The global functional outcome assessed by the modified Rankin scale has also been associated with left-side hemispheric lesions and with lesions in the internal capsule interrupting the corticospinal tract (Ernst et al., 2018; Munsch et al., 2016). Regarding cognitive functions, Munsch et al. (2016) reported that global cognitive performances three months after the stroke were associated with left-side hemispheric lesions, mainly in frontal and temporal cortices, amygdala, hippocampus, and thalamus. Interestingly, while the stroke location remained a significant predictor of cognitive outcome after controlling for age, clinical severity, and stroke volume, the association with functional outcome did not persist, clinical severity remaining the only significant predictor. The association between domain-specific cognitive performances and stroke location is also a topic of interest, but has been less investigated. Assessing cognitive functions three months after stroke, we observed that the left frontal cortex was particularly associated with attention, language, and abstraction, left insula with naming, and left parahippocampal gyrus with delayed recall (Sagnier et al., 2019). Conversely, no association was observed with mood outcome, suggesting different pathophysiological mechanisms between poststroke cognitive and mood disorders. But further VLSM studies are still needed to confirm these results.

Besides the evaluation of motor and neuropsychological poststroke functions, VLSM also allows the investigation of brain areas involved in other fields of poststroke outcome such as poststroke pain. Using VLSM, Seifert et al. (2016) confirmed the role of insula and somatosensory cortex, which are part of the central pain matrix, in poststroke headache. However, they noted that patients with ischemic stroke and headache had a higher infarct volume. Though the difference was not significantly compared to patients without headache, adjustment for infarct volume or clinical severity is important in these type of studies. Stroke-related erectile dysfunction has also been analyzed using VLSM and was associated with lesions involving the right occipital cortex, thalamus, and the left parietal association area (Winder et al., 2017). However, the information concerning patients’ medications was lacking here, while it is known that some treatments often prescribed in stroke patients such as diuretics or beta-blockers can also cause erectile dysfunction. These observations strongly reinforce the need to perform these studies while adjusting for the most relevant determinants of the tested outcome measure, before drawing relevant conclusions on the role of brain lesion location.

Hence, though VLSM seems to be an accurate method to detect strategic lesion locations involved in prognosis, controlling for some relevant variables is important. Indeed, Laredo et al. (2018) reported an association between ischemic stroke located in the insula, superior temporal gyrus, rolandic operculum, superior longitudinal fasciculus, supramarginal gyrus, inferior fronto-occipital fasciculus, and a poor outcome or death at three months poststroke. However, the association with death was substantially reduced after controlling for stroke volume, suggesting that death was mainly related to the size of the lesion rather than its location. Moreover, covariables should be added carefully to avoid collinearity (i.e., clinical severity score and stroke volume, which are two highly collinear variables). Therefore, further large cohorts including strokes distributed widely over the brain parenchyma are needed to avoid bias related to large size lesions or over-represented locations.

Beyond the usual VSLM analyses mentioned above, a multiple-lesion symptom mapping approach has recently been developed, considering not only the comparison between the infarcted voxels and the noninfarcted voxels, but also the comparison between lesioned voxels due to different lesion types, such as ischemic stroke and white matter hyperintensities (Zhao et al., 2018). Being not sensitive to the multicollinearity of the specific lesioned-voxels, the approach allows to highlight the significant voxels associated with the outcome measure, according to the lesion type. After controlling for lesion volumes, age, sex, and educational level, Zhao et al. (2018) observed that ischemic stroke involving the left basal ganglia, frontal, temporal, and occipital cortexes was relevant to global cognitive outcome in the first year following the stroke. White matter hyperintensities in the corpus callosum, corona radiata, and the posterior thalamic radiations were also relevant in the cognitive outcome. Some significant voxels in the corpus callosum, left corona radiata, left posterior thalamic radiations, and left longitudinal fasciculus were shared by ischemic stroke and white matter hyperintensities. Thus, these results showed that not only the location of the lesion, but also its nature (necrosis, gliosis, edema, axonal, and myelin loss) might be significant determinants of poststroke recovery.

3.2 Perfusion-weighted imaging

Different techniques of brain perfusion analysis exist (Wintermark et al., 2005), including positron emission tomography, single-photon emission computed tomography, xenon-enhanced computed tomography, perfusion computed tomography, and using MRI: dynamic susceptibility contrast (DSC) MR perfusion and arterial spin labeling. DSC MR perfusion uses gadolinium chelate as a paramagnetic marker and it is based on a mathematical model considering arterial enhancement into the capillary system input. In contrast, arterial spin labeling does not need exogenous contrast agent and uses the own spins of the blood magnetically labeled as an endogenous marker (Wolf & Detre, 2007). Perfusion imaging, in the acute stage of an ischemic stroke, allows the identification of the salvageable penumbra corresponding to the parenchymal area where cerebral blood flow is sufficient to maintain ion homeostasis, but too low to sustain electrical activity (Chen & Ni, 2012). The discrepancy between the volume of the ischemic core and the larger volume of abnormal perfusion defines the notion of mismatch that has been used in several trials to improve the selection of patients eligible for revascularization strategies within the therapeutic time window or after (i.e., beyond the 4.5 hr delays for intravenous thrombolysis or 6 hr for mechanical thrombectomy). However, until now, no study has demonstrated the benefit of perfusion imaging in the acute therapeutic management of stroke patients in the very early stroke phase. Therefore, perfusion should not be used to identify patients eligible or not to intravenous thrombolysis within the 4.5 hr time window or the 6 hr for endovascular thrombectomy (Bang et al., 2010; Olivot et al., 2008). Nonetheless, comparison between endovascular trials in acute ischemic stroke demonstrated a lower number needed to treat the ratio in studies that used advanced imaging for patient selection, including a target mismatch (Campbell et al., 2015; Saver et al., 2015). More recently, the DEFUSE 3 trial (Albers et al., 2018) demonstrated a benefit of mechanical thrombectomy up to 16 hr after the stroke onset, with a proximal vessel occlusion, in a selected population of patients (Albers et al., 2018). The brain imaging criteria included an ischemic core volume < 70 ml, a target mismatch volume ≥ 15 ml, and a ratio of the volume of mismatch to ischemic core ≥ 1.8. Similarly, the EXTEND trial (Ma et al., 2019) and recent metanalysis (Campbell et al., 2019) demonstrated the benefit of perfusion imaging to select patients eligible to intravenous thrombolysis in the 4.5-9 hr time window, using slightly different parameters (an ischemic core volume < 70 ml, a target mismatch volume ≥ 10 ml, and a ratio of volume of mismatch to ischemic core ≥ 1.2). These studies emphasized the relevance of perfusion-weighted imaging at the acute phase of ischemic stroke for the identification of patients that will better respond to revascularization therapy even in an extended delay and that will have better 90-day functional outcome.

Interestingly, perfusion-weighted imaging acquired in the acute phase can also be normal, while diffusion-weighted imaging lesion demonstrates a recent infarct, indicating spontaneous reperfusion. This profile has been associated with reversible diffusion lesions, which are regions probably part of the ischemic penumbra and are usually of good prognosis (Olivot et al., 2009). Conversely, sustained hypoperfusion long after the acute event has also been reported and might contribute to a poor cognitive outcome (Brumm et al., 2010). This result suggests the potential role of brain perfusion in the efficiency of cognitive processes, but also suggests a potential role in the modulation of poststroke brain plasticity that will have to be investigated in further studies.

In conclusion, perfusion-weighted imaging is mainly used in the acute stroke setting to help clinicians in the identification of patients still eligible for a revascularization strategy, despite prolonged ischemia. Its role in the early assessment of poststroke prognosis following reperfusion therapies is still uncertain.

4.1 Magnetization transfer imaging

Magnetization transfer imaging is an MRI sequence allowing the quantification of the membrane cells’ density of a given tissue, based on the interactions between mobile protons contained in the free water and macromolecule-bounded protons contained in the membrane layer of cells. The magnetic transfer ratio (MTR) calculated from these sequences reflects the microstructural damage of a tissue. A low MTR indicates reduced proton interactions and brain damage, with myelin and axonal loss. While applied in various neurological diseases, it has been shown that lower MTR measured in the ischemic core one month after symptoms onset was associated with worse functional recovery (Sibon et al., 2015). Interestingly, higher MTR standard deviation and coefficient variation were reported in patients with good functional outcome, suggesting a nonuniformly injury of the tissue, including both complete infarcted and partially saved tissues. As well, Tourdias et al. (2007) observed one month after stroke, decreased MTR in normal-appearing brain parenchyma that presented perfusion or diffusion abnormalities at baseline (i.e., the “salvaged penumbra”). The authors suggested a partial decrease in the cellular content with neuronal loss due to incomplete infarction and short duration arterial occlusion, to explain this persistent MTR decrease in reversible perfusion and diffusion lesions. Thus, magnetization transfer imaging might be a useful tool to identify incomplete saved tissue in normal-appearing brain parenchyma and might constitute an additional therapeutic target in trials aiming at evaluating the efficacy of neuroprotective agents in stroke.

4.2 MR spectroscopy

MR spectroscopy allows the identification of different metabolites that reflects the brain structure and function. Bivard, Stanwell, Levi, and Parsons (2013), Bivard et al. (2014) raised the issue of the metabolic and perfusion status of the penumbra region in patients with good poststroke recovery. They first demonstrated that patients with a hyperperfusion status on a brain MRI realized 24 hr after symptoms onset had a better early clinical improvement and 90-day functional recovery compared to patients with persistent hypoperfusion (Bivard et al., 2013). Subsequently, using spectroscopy, they showed that these hyperperfused areas had a specific metabolic signature with increased glutamate, N-acetylaspartate, and lactate (Bivard et al., 2014). These results were probably explained by an increased astrocytes activity in the neurovascular unit, inducing increased blood flow, and by a rebound of protein synthesis in the cell, following the end of oxygen deprivation. Thus, increased glutamate and lactate after reperfusion were rather a marker of metabolic cellular activity upturn than a marker of excitotoxicity and anaerobic glycolysis usually seen in the first hours after a vessel occlusion. Moreover, N-acetylaspartate was also increased in the contralesional hemisphere of the patients with a hyperperfusion pattern, reflecting neuronal density and probably, a better neuroplasticity.

Nevertheless, while these techniques are useful to address poststroke pathophysiological mechanisms at the group level, their reliability to help clinicians in the early prediction of the poststroke outcome at the individual level is more limited.

5.1 Cortical morphology

Together with the extent of leukoencephalopathy, the most frequently reported radiological biomarker associated with poststroke cognitive impairment is cerebral atrophy (Dufouil et al., 2009; Pendlebury & Rothwell, 2009; Yatawara, Ng, Chander, & Kandiah, 2018). Indeed, many studies showed an association between global cerebral atrophy or medial temporal atrophy, and poststroke cognitive impairment (Cordoliani-Mackowiak, Hénon, Pruvo, Pasquier, & Leys, 2003; Firbank et al., 2007; Gemmell et al., 2012; Hénon, Pasquier, Durieu, Pruvo, & Leys, 1998; Jokinen et al., 2004; Kebets et al., 2015; Pohjasvaara et al., 2000; Tang et al., 2004). More specifically, Firbank et al. (2007) found that medial temporal atrophy was a predictor of memory loss two years after stroke, whereas the volume of white matter hyperintensities was not. Also, it has been suggested that medial temporal atrophy was associated with prestroke cognitive impairment (Hénon et al., 1998), supporting a possible neurodegenerative contribution in the occurrence of poststroke cognitive impairment. This hypothesis was emphasized by the results of Tuladhar et al. (2015) who observed in 426 subjects with cerebral small vessel disease, an association between white matter hyperintensities and cortical thickness in specific regions, an association between cortical thickness in frontotemporal regions and cognitive performances (global cognition, processing speed, flexibility, verbal fluency, and attention), but no relationship between white matter hyperintensities and cognitive performances after controlling for cortical thickness. Thus, these results argued for a more relevant role of cortical atrophy on top of white matter hyperintensities in poststroke cognitive impairment, but, however, did not allow to conclude concerning the pathophysiological mechanism leading to cortical atrophy. Indeed, cortical thinning might reflect a primary neurodegenerative process or might be a secondary consequence of white matter vascular disruption and related anterograde Wallerian degeneration. A strong interaction between white matter and cortical changes was also supported by Jang et al. (2017) using a combination of cortical thickness and white matter diffusion tensor imaging (DTI) analysis, that white matter hyperintensities could be a cause but also a consequence of cortical atrophy.

The studies cited above evaluated cortical atrophy using the measurement of cortical thickness, which is to date, the more accurate method to assess cortical atrophy using surface-based or voxel-based approaches (Li, Pardoe, et al., 2015). Although the two methods are valid for estimating cortical thickness, it is worth noting that the precision of white matter and pial boundaries that are used to estimate cortical thickness might be affected by voxel resolution. The voxel-based method is also sensitive to partial volume effects, which can lead to inaccuracies in the measurement of cortical thickness, mainly in high convoluted areas. The use of cortical thickness for the evaluation of structural plasticity in the specific context of stroke is varied (Brodtmann et al., 2012; Cheng et al., 2015; Duering et al., 2015; Ferris, Peters, Brown, Tourigny, & Boyd, 2018; Pundik, Scoco, Skelly, McCabe, & Daly, 2018; Schaechter, Moore, Connell, Rosen, & Dijkhuizen, 2006; Sterr et al., 2013; Thong et al., 2013; Tuladhar et al., 2015). For instance, Ferris et al. (2018) observed an association between type 2 diabetes (a known risk factor of cerebral small vessel disease) and thinning of sensorimotor cortices after chronic stroke. Combining microstructural connectivity analyses (Cheng et al., 2015; Duering et al., 2015), it has been demonstrated that the cortical thickness of areas connected to a subcortical ischemic stroke significantly decreased three to six months after stroke. In subjects with silent lacunar infarcts, it has been shown that the cortical thickness was lower in widespread regions and was associated with poorer performances in attention, memory, and language (Thong et al., 2013). All these studies emphasized the influence of chronic cerebral small vessel disease in the lowering of cortical thickness. Conversely, Pundik et al. (2018) showed that the sensory function improvement after rehabilitation was associated with higher cortical thickness in the temporo-occipital cortex of the ipsilesional hemisphere and in the parietal cortex of the contralesional hemisphere. Similarly, Sterr et al. (2013) observed an increase of cortical thickness in the contralesional somatosensory cortex of chronic subcortical stroke patients after 15 days of constraint-induced therapy. Hence, though the severity of cerebrovascular disease can lead to secondary cortical atrophy, rehabilitation might have an inverse effect in some strategic areas, suggesting recuperation and vicariance phenomena.

The other quantitative method currently used to assess cortical atrophy is the measure of cortical volume with a voxel-based morphometry approach (Ashburner & Friston, 2000). Studies including stroke patients showed that subjects with lower grey matter volumes had lower cognitive performances in executive, attentional, and global cognitive functions after stroke (Muller et al., 2011; Sagnier et al., 2017). It has also been suggested that chronic stroke patients had altered dynamic cerebral autoregulation bilaterally and that better autoregulation was associated with less ipsilesional temporal grey matter atrophy (Aoi et al., 2012), which reminds the concept of brain plasticity emerging after stroke.

Besides radiological biomarkers of cortical atrophy, the shape of the central sulcus is another original variation of cortical morphology that has been recently reported (Jouvent et al., 2016). It has been shown that its vertical position and the size of the hand knob were associated with functional recovery in patients with a history of small subcortical ischemic stroke. The hypothesis of a “motor reserve” was suggested, represented by developmental variations in the shape of central sulcus, leading to differences in poststroke recovery, regardless of the underpinning cerebrovascular disease. Hence, similar to the concept of brain reserve suggesting that people with bigger brains were likely to cope better against brain injury, this new concept of “motor reserve” suggested that people with an upper position of the hand knobs and smaller hand knobs were more likely to have a worse motor prognosis after brain injury. The shape of the central sulcus might, therefore, be a marker of underlying motor connections.

5.2 Chronic cortical cerebral microinfarcts

Chronic cortical cerebral microinfarcts (CMIs) have been recently associated with cognitive outcome after ischemic stroke and transient ischemic attack (Wang et al., 2016). They are defined as strictly intracortical lesion measuring under 4 mm, appearing in at least two planes hyperintense on T2-weighted imaging or FLAIR sequences, hypointense on T1-weighted imaging, and isointense on T2*-weighted imaging, gradient echo or susceptibility-weighted imaging (van Veluw et al., 2017). Cortical CMIs are distinct from enlarged perivascular spaces extending into the cortex or gyrus curvature and should be remote from larger ischemic stroke. They differ from acute CMIs by their signal on diffusion-weighted imaging that is hyperintense, and iso or hypointense on apparent coefficient diffusion for acute lesions. Chronic cortical CMIs have been originally described in pathological studies and associated with vascular dementia, Alzheimer's disease, cerebral amyloid angiopathy, together with cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (Arvanitakis, Leurgans, Barnes, Bennett, & Schneider, 2011; Jouvent et al., 2011; Smith, Schneider, Wardlaw, & Greenberg, 2012). In the last few years, van Veluw et al. (2013) showed that their detection of high-resolution brain MRI (7 Tesla) was possible. Of 23 lesions identified on 7T MRI with available histology, 12 were histologically confirmed to be cortical CMIs and the last 11 lesions were perivascular spaces, always in a juxtacortical location. Subsequently, it has been shown that the detection of cortical CMIs was possible on 3T MRI, with a detection of 27% of the lesions identified on 7T MRI (van Veluw et al., 2013).

The presence of cortical CMIs has been associated with advanced age, history of stroke (cortical or lacunar), white matter hyperintensities volume, and intracranial stenosis (Hilal et al., 2016). In patients with ischemic stroke or transient ischemic attack, chronic cortical CMIs have been identified in 9% of the patients (Wang et al., 2016), whereas in patients with stroke and intracranial atheroma, they have been identified in 52% (Leng et al., 2017). An association with cognitive impairment has been reported, including impairment in visuospatial functions, attention, executive functions, and processing speed (Ferro et al., 2017; Hilal et al., 2016; Wang et al., 2016). Nevertheless, the strength of the association with clinical performances is probably underestimated, as the load of cortical CMIs detected on MRI could be considered as the “tip of the iceberg,” most of the lesions being invisible on 3T MRI. In a pathological study, Coban, Tung, Yoo, Vinters, and Hinman (2017) suggested that an axonal disorganization adjacent to the cortical CMIs, together with a loss of axons and axoglial contacts resulting in neural transmission impairment, might underlie their role in impaired cognition.

In brief, chronic cortical CMIs appear to be a new marker of cerebrovascular disease, including cerebral small vessel disease and intracranial atheroma, and a risk factor of cognitive impairment. Their detection by the visual assessment on a brain 3T MRI, more available in routine, is possible, albeit underestimated and time-consuming.

5.3 Diffusion tensor imaging

DTI is an imaging technique allowing the identification of extracellular water along anisotropic structures (i.e., dependent on a preferred direction), as white matter tracts. The DTI parameters include fractional anisotropy (FA; the higher the FA, the higher the direction of the structure is preferred); mean diffusivity (MD), which quantifies the average amplitude of diffusion movements; axial diffusivity, a marker of axonal degeneration; and radial diffusivity, a marker of myelin loss and axonal degeneration. In addition, statistical algorithms enable to perform tractography, aiming at rebuilding the white matter tracts. However, it should be noted that, though FA and MD are the more widely DTI metrics used to explore white matter microstructural abnormalities, it has been described that these parameters could be contaminated by the extracellular water content relatively unrestricted by their local microenvironment, named “free water” (Maillard et al., 2019). The measurement of the free water content has shown to improve the specificity of DTI parameters and reproducibility, and it has been suggested that FA and MD should be corrected for free water content.

The integrity of white matter and its impact on poststroke prognosis have been the purpose of an important literature. In the area of motor recovery, meta-analyses (Kumar, Kathuria, et al., 2016; Kumar, Yadav, et al., 2016) suggested that FA measured in the corticospinal tract of subacute ischemic or hemorrhagic stroke patients was correlated with motor recovery. Other studies evaluated the integrity of the corticospinal tract by quantifying the overlap of stroke voxels on a corticospinal tract derived from healthy controls (Feng et al., 2015; Puig et al., 2017; Zhu, Lindenberg, Alexander, & Schlaug, 2010), but raising the problem of normalization of lesioned brains onto healthy brains. Another technique of corticospinal tract analysis consisted of quantifying the number of fibers constituting the corticospinal tract in the lesional hemisphere, compared to the contralateral side. Bigourdan et al. (2016) evaluated the corticospinal tract fiber number ratio by normalizing the number of fibers in the affected tract to the number of fibers in the contralateral tract. They demonstrated, in a sample of 117 subacute ischemic stroke patients, that the fiber number ratio at a baseline was an independent predictor of motor recovery one year after stroke, after controlling for age, stroke volume, and baseline motor impairment. However, though these results argued for a relationship between the alteration of the corticospinal tract integrity and motor outcome, it remained unknown whether this alteration was related to the stroke lesion itself, to white matter damages represented by hyperintensities, or microstructural abnormalities invisible on standard MRI sequences.

Thereby, other DTI studies addressed the issue of the impact of normal-appearing white matter on the poststroke outcome. In elderly population, it has been suggested that the alteration of normal-appearing white matter integrity was the first step toward the emergence of cerebrovacular disease, being a marker of increased risk of conversion from normal-appearing white matter to white matter hyperintensities (Maillard et al., 2013). The alteration of DTI parameters in normal-appearing white matter of healthy elderly subjects has been associated with worse performances in global cognition, information processing speed, and executive functions (Vernooij et al., 2009). Similarly, following a stroke, Kliper et al. (2014) reported a direct effect of normal-appearing white matter integrity on cognitive performances (global cognitive functions, memory, visuospatial functions, executive functions, and attention) assessed one year after stroke, while stroke volume did not have a significant effect. In addition to its impact on cognitive functions, median FA measured in the normal-appearing white matter of the contralesional hemisphere has been independently associated with 90-day functional outcome (Etherton et al., 2017). These results are in accordance with a growing literature, suggesting that white matter integrity is critical in brain organization and functioning of specific neuronal networks. For example, the relevance of corpus callosum integrity in poststroke motor recovery has been reported in many studies (Li, Wu, et al., 2015; Lindenberg, Zhu, Rüber, & Schlaug, 2012; Schaechter et al., 2009; Wang et al., 2012). Thus, normal-appearing white matter integrity appears to be a new radiological biomarker to quantify in poststroke recovery studies.

5.4 Brain iron accumulation

Brain iron accumulation increases oxidative stress that can produce negative effects. In animal models of stroke, it has been shown that iron accumulated first in activated microglia, a marker of neuroinflammation, then in parenchyma, and around amyloid plaque deposits, a marker of neurodegeneration (Justicia, Ramos-Cabrer, & Hoehn, 2008). In human, data are scarce, but brain iron accumulation has been observed in the thalami of ipsilesional hemispheres after stroke (Kuchcinski et al., 2017). Iron accumulation was measured on a brain MRI performed one year after stroke, using the transverse relaxation rate R2*, which was linearly correlated with brain iron concentration (Langkammer et al., 2010). More specifically, thalamic iron accumulation in mediodorsal nucleus was associated with ischemic strokes involving the frontal, temporal, and insular cortices, whereas iron accumulation in the pulvinar was associated with ischemic strokes involving the temporal and parietal cortices, suggesting a participation of the postlesional thalamo-cortical pathways interruption. In addition, thalamic iron accumulation was independently associated with poor one-year functional and mood outcome, emphasizing the influence of neurodegeneration in a poor outcome. Finally, these results offer new insights into the role of iron in brain functions and new therapeutic target to improve stroke outcome.