Longitudinal Monitoring of Microstructural Alterations in Cerebral Ischemia with in Vivo Diffusion-weighted MR Spectroscopy

Abstract

Background

The time course of cellular damage after acute ischemic stroke (IS) is currently not well known, and specific noninvasive markers of microstructural alterations linked to inflammation are lacking, which hinders the monitoring of anti-inflammatory treatment.

Purpose

To evaluate the temporal pattern of neuronal and glial microstructural changes after stroke using in vivo single-voxel diffusion-weighted MR spectroscopy.

Materials and Methods

In this prospective longitudinal study, participants with IS and healthy volunteers (HVs) underwent MRI at 3.0 T. In participants with IS, apparent diffusion coefficients (ADCs) and concentrations of total N-acetyl-aspartate (tNAA), total creatine (tCr), and total choline (tCho) were measured in volumes of interest (VOIs), including the lesion VOI (VOI_les) and the contralateral VOI (VOI_cl) at 2 weeks, 1 month, and 3 months after IS. HVs were examined once, with VOIs located in the same brain regions as participants with IS. Within- and between-group differences and longitudinal changes were examined using linear mixed-effects models.

Results

Twenty participants with IS (mean age, 61 years ± 13 [SD]; 12 women) and 20 HVs (mean age, 59 years ± 13; 12 women) were evaluated. No differences in ADCs or concentrations were observed in VOI_cl between HVs and participants with IS. In participants with IS, the ADC of tCr was higher in VOI_les than in VOI_cl at 1 month (+14.4%, P = .004) and 3 months after IS (+19.0%, P < .001), while the ADC of tCho was higher only at 1 month (+16.7%, P = .001). No difference in the ADC of tNAA was observed between the two VOIs at any time point. tNAA and tCr concentrations were lower in VOI_les than in VOI_cl and were stable over time (approximately −50% and −30%, respectively; P < .001).

Conclusion

High diffusivity of choline-containing compounds and total creatine (tCr) in the ischemic lesion 1 month after ischemic stroke (IS) indicates glial morphologic changes, suggesting that active inflammation is still ongoing at this time point. High tCr diffusivity up to 3 months after IS likely reflects the presence of astrogliosis at the chronic stage of cerebral ischemia.

Clinical trial registration no. NCT02833961

© RSNA, 2022

Online supplemental material is available for this article.

Summary

In ischemic stroke lesions, elevated choline and creatine diffusivities detected with diffusion-weighted MR spectroscopy suggest inflammation-related morphologic changes in glial cells, which are still active 1 month after stroke.

Key Results

• ■ In a longitudinal study of 20 participants with ischemic stroke, a higher apparent diffusion coefficient (ADC) of total creatine was seen in the ischemic lesion than in normal tissues at diffusion-weighted MR spectroscopy at 1 and 3 months after stroke (+14.4% [P = .004] and +19.0% [P < .001], respectively).

• ■ The ADC of total choline was higher in the ischemic lesion than in normal tissues only at 1 month after stroke (+16.7%, P = .001).

• ■ There was no evidence of a difference in the ADC of N-acetyl-aspartate between the lesion and normal tissues at any time point.

Introduction

Effective neuroprotective therapies for patients affected by ischemic stroke (IS) are still lacking. During ischemia, a blocked artery results in deprivation of glucose and oxygen in a brain region, triggering a cascade of pathologic mechanisms in the affected area, including apoptosis (or programmed cell death initiated by chemical signals), necrosis, and inflammation (1). Over the past decade, great attention has been given to the interplay between inflammation and repair in brain injury. Inflammatory processes are thought to be beneficial for tissue repair by clearing necrotic debris; however, by facilitating cell death, these processes may also have a detrimental effect on the surrounding ischemic tissue (2,3). Noninvasive markers of glial cell reactivity are highly desirable, notably for monitoring the effect of treatments targeting inflammation.

Diffusion-weighted MR spectroscopy noninvasively probes tissue microstructure by means of the diffusion of intracellular metabolites (4–6), enabling investigation of specific pathophysiologic processes after stroke. Most brain metabolites are confined preferentially within specific cell populations. N-acetyl-aspartate is located mostly in neurons, and its diffusivity has been used as a marker of intra-axonal integrity in multiple sclerosis (7). Choline-containing compounds and myo-inositol are preferentially located in glial cells, and abnormalities in their diffusion have been related to inflammation (8–10), while creatine- and phosphocreatine-altered diffusion has been linked to either glial pathologic processes (8) or changes in metabolic rates (11,12).

While most of the damage is known to occur in the first few hours after IS, evidence suggests that tissue injury may continue subacutely (13). In particular, postischemic inflammation is thought to still contribute to brain injury several weeks after stroke (1,3).

The goal of our study was to monitor the temporal changes of both neuronal and glial abnormalities occurring in the human brain after IS, up to the chronic stage of the disease. To this aim, the apparent diffusion coefficients (ADCs) and concentrations of total N-acetyl-aspartate (tNAA) (ie, N-acetyl-aspartate and N-acetyl-aspartyl-glutamate), total creatine (tCr) (ie, creatine and phosphocreatine), and total choline (tCho) (ie, choline-containing compounds) were measured in the ischemic lesion and contralateral tissue at three time points up to 3 months after IS.

Materials and Methods

Study Participants

Twenty-two participants with IS were prospectively recruited at the University Hospital Pitié-Salpêtrière (Paris, France) and examined at the Center for Neuroimaging Research (convenience series, July 2016 to March 2018). Data from 20 patients were included in the final analysis (Fig 1). The inclusion criteria were as follows: IS confirmed with diffusion-weighted MRI, time between stroke onset and study inclusion of less than 21 days, volume of the ischemic lesion greater than 8 cm³, and no history of neurologic or psychiatric disorders. The exclusion criteria were age younger than 18 years, contraindications for MRI, and life-threatening conditions occurring within the 6-month follow-up.

Figure 1:

Flowchart of patient inclusion. DW-MRS = diffusion-weighted MR spectroscopy.

All participants were evaluated with the National Institutes of Health Stroke Scale, the 3-month modified Rankin Scale, and the Barthel Index at each visit.

Twenty healthy volunteers (HVs) were recruited as controls. Each HV was individually matched by age (± 5 years) and sex to a participant with IS. The inclusion criteria for HVs were as follows: no history of neurologic or psychiatric disorders, Mini-Mental State Examination score higher than 27, age of 18 years or older, no contraindications to MRI, and no use of psychoactive medication or recreational drugs.

All study participants provided written informed consent according to local procedures before the study. The study was approved by the ethics committee. The trial was registered on ClinicalTrials.gov (identifier NCT02833961).

Data Acquisition and Postprocessing

All individuals were scanned using a 3.0-T whole-body Magnetom Prisma Fit MRI scanner (Siemens Healthineers). The body coil was used for transmission, and a 64-channel receive array head coil was used for reception.

Each participant with IS was scanned at three time points: 14 days ± 7, 30 days ± 10, and 90 days ± 30 after acute IS. HVs underwent only one MRI examination.

The acquisition protocol included a three-dimensional T2-weighted fluid-attenuated inversion-recovery sequence used for identification of the infarcted lesion, positioning of the spectroscopic volumes of interest (VOIs), and lesion segmentation. Three-dimensional T1-weighted images were also acquired for tissue segmentation.

A single-voxel semi–localization by adiabatic selective refocusing sequence (echo time, 100 msec; acquisition time, approximately 2 minutes 40 seconds per VOI) with diffusion gradients (14) (two b values of approximately 0 and 3.55 msec/μm²; diffusion time, 50 msec) was used. Diffusion-weighted water spectra were acquired for eddy current corrections. In participants with IS, data were collected in two VOIs, the infarcted lesion VOI (VOI_les) and the mirrored contralateral VOI (VOI_cl). VOI_les sizes were adapted to maximize the signal from infarcted tissues. For each HV, a VOI_les of the same size was positioned in the same location as that of the corresponding age- and sex-matched participant with IS. Similarly, VOI_cl was placed in the contralateral brain region (Fig 2A).

Figure 2:

Volumes of interest (VOIs) and diffusion-weighted MR spectra. (A) Examples of VOI locations in one healthy volunteer (HV) and one participant with ischemic stroke (IS) are shown on axial fluid-attenuated inversion-recovery images. The MR spectra measured in the ischemic lesion VOI (VOI_les) (red box) and in the contralateral VOI (VOI_cl) (green box) in normal-appearing tissue are shown for a 73-year-old woman hospitalized for IS (lesion volume, 13 cm³) with National Institutes of Health Stroke Scale score of 7. In the HV (72-year-old woman), the VOI_les and VOI_cl are positioned in the same locations as in the corresponding age- and sex-matched participant with IS. (B) Spectra acquired in VOI_cl of the HV and in VOI_cl and VOI_les of the participant with IS at 2 weeks after IS (S1). Spectra were acquired at b values of 0.1 msec/μm² and 3.55 msec/μm². Spectra are shown without line broadening and scaled to the water signal acquired at b value of 0.1 msec/μm² in VOI_cl. S2 = 1 month after IS, S3 = 3 months after IS, tCho = total choline, tCr = total creatine, tNAA = total N-acetyl-aspartate.

Spectra were postprocessed as previously described (14). LCModel version 6.3–0G (15) was used to quantify metabolite signal intensity from diffusion-weighted spectra. tNAA, tCr, and tCho ADCs were calculated in each VOI, assuming a monoexponential decay of the signal as a function of the b value. The ADCs measured in the lesions, or in the region matching a lesion for HVs, and in the contralateral regions were termed ADC_les and ADC_cl, respectively.

tNAA, tCr, and tCho concentrations were evaluated in both VOI_les and VOI_cl from the spectra with a b value of approximately 0 msec/μm² with use of the water spectra acquired in VOI_cl as reference for both VOIs. The tissue composition of VOI_cl and water T2 relaxation effects on quantification were taken into consideration.

Volumes of infarcted tissue were obtained with manual segmentation of fluid-attenuated inversion-recovery images with use of MRIcron (NeuroImaging Tools & Resources Collaboratory). The fractions of VOIs filled with infarcted tissue were estimated by overlaying the lesion masks with the corresponding VOIs.

Further details on acquisition and postprocessing are available in Appendix S1.

Statistical Analysis

Statistical analyses were conducted using R version 3.6.1 (R Development Core Team) and Matlab (MathWorks) by F.X.L. and L.Y.C. (with 15 and 16 years of experience, respectively).

Within- and between-group differences and longitudinal changes in participants with IS were examined using linear mixed-effects models (LMMs). For each analysis, one LMM was built for each of the six markers under investigation (tNAA, tCr, and tCho ADCs and concentrations). Data from three VOIs_les and three VOIs_cl were not acquired or were not usable due to low patient compliance or technical issues. These data were replaced using a multiple imputation technique (16). Briefly, for each missing value of a given variable, data distribution was evaluated from the acquired data set of that variable with use of a bootstrap procedure. From the data distribution, the most likely value of the missing value was fitted using an additive regression model.

Paired t tests were performed to compare ADCs and concentrations in the two regions to evaluate possible differences between VOI_cl and VOI_les in HVs.

Then, to evaluate differences between HVs and participants with IS in metabolite ADCs and in concentrations measured in VOI_cl at the three time points, the LMMs were fitted to the ADCs and concentrations measured in VOI_cl with use of the within-participant factor group (HVs and participants with IS at each time point) and age as well as an age by group interaction term as a fixed effect. The participant identifier was assigned as a random effect (intercept) to account for the repeated measurements across time points (including values from the matched HVs). Covariate adjustment for sex was considered in all models.

Next, as no differences in ADCs and concentrations in VOI_cl were found between HVs and participants with IS at any time point and for any metabolite, the statistical analysis was conducted within participants with IS (post hoc analysis) only to study the evolution over time of the relative differences of VOI_les compared with VOI_cl, expressed as percentage of the contralateral value: (ADC_les – ADC_cl)/ADC_cl × 100 (hereafter referred to as ADC_rel) and [metabolite]_rel = ([metabolite]_les – [metabolite]_cl)/[metabolite]_cl × 100 for each of the three metabolites. LMMs were fitted to the percentage changes calculated for each marker with use of the ordinal number of the visit (1, 2, and 3) as a fixed effect and including age, sex, and the fraction of VOI_les occupied by the lesion for covariate adjustment. LMMs were also fitted to percentage changes calculated for each marker in HVs, including sex and age for covariate adjustment.

Percentage changes were reported for HVs and for each visit of participants with IS as estimated marginal means (ie, mean values predicted by the LMMs) and standard errors provided by the emmeans package (version 1.4.5). All percentage changes significantly different from zero were identified using a t test based on the t ratio statistic with the Kenward-Roger approximation for degrees of freedom. All LMMs were fitted using the lmer function of the lme4 package (version 1.1–21), and details of the models are provided in Appendix S1.

The level of statistical significance was set at P < .05 or adjusted P for all tests. A sample size of 20 participants with IS and 20 HVs was considered sufficient based on our previous study on diffusion-weighted MR spectroscopy data reproducibility (14) and the expected changes in the lesion (17).

Results

Participant Characteristics

Two of the 22 participants with IS included in the study could not complete all MRI examinations and had partial diffusion-weighted MR spectroscopy data at two time points. Data from 20 participants with IS (mean age, 61 years ± 13 [SD]; 12 women) and 20 HVs (mean age, 59 years ± 13; 12 women) were analyzed (Fig 1). Demographic information and acquisition details are summarized in Tables 1 and 2.

Table 1:

Characteristics of HVs and Participants with IS

Table 2:

Characteristics of Participants with IS at the Three Time Points

All participants with IS had an ischemic infarct lesion located in the corona radiata, putamen, or occipital lobe. None of the patients experienced a recurrent stroke during follow-up.

MR Spectroscopy Data

The mean VOI size was 10 cm³ ± 3 (range, 8–18 cm³). ADCs and concentration mean values and SDs calculated for both participants with IS and HVs are summarized in Table 3. Figure 2B shows examples of diffusion-weighted spectra acquired in HVs and participants with IS. The Cramér-Rao lower bounds of the three metabolites in VOI_les in participants with IS ranged from 2% to 20% for both b values. The Cramér-Rao lower bounds in VOI_cl in participants with IS and from both VOIs in HVs ranged from 2% to 6% for all metabolites and both b values.

Table 3:

ADCs and Concentrations of Metabolites Measured in HVs and Participants with IS at the Three Time Points for Both VOIs

Metabolite ADCs and Concentrations in the Contralateral Hemisphere

No differences in metabolite ADCs or in concentrations were observed in HVs between VOI_cl and VOI_les (ie, between the regions matching lesions of participants with IS and their mirror region). No differences in ADCs or in concentrations were observed in VOI_cl between HVs and participants with IS at any time point (Table S1).

Metabolite ADCs and Concentrations in the Lesion

Metabolite ADCs.—In participants with IS, no differences in ADC_rel for tNAA from zero were observed at any time point, indicating no difference between the lesional and contralateral hemispheres (Fig 3, Table 4). In contrast, ADC_rel for tCr was higher than zero at both 1 and 3 months after IS, indicating increased ADC_rel for tCr at these two time points (+14.4% [P = .004] and +19.0% [P < .001], respectively), while ADC_rel for tCho was higher than zero only at 1 month after IS (+16.7%, P = .001) (Fig 3, Table 4). ADC_rel did not change over time for any of the metabolites (Table S2). ADC_rel did not correlate with the volume of the lesion for any of the metabolites. Metabolite ADCs obtained in participants with IS and HVs are reported for both VOI_cl and VOI_les in Figure S1.

Figure 3:

Box and whisker plots show apparent diffusion coefficients (ADCs) of metabolites. ADC percentage changes (ADC_rel) of (A) total N-acetyl-aspartate (tNAA), (B) total creatine (tCr), and (C) total choline (tCho) in the volume of interest (VOI) of the lesion with respect to the contralateral VOI (with ADC_rel = [ADC_les – ADC_cl]/ADC_cl, where ADC_les is the lesion ADC and ADC_cl the contralateral ADC) for participants with ischemic stroke (IS) at each time point (open circles) and for healthy volunteers (HVs) (filled circles). For participants with IS, the values are plotted versus the number of days after the symptom onset. For each box, the midline indicates the median, and the bottom and top edges indicate the 25th and 75th percentiles, respectively. The bottom and top whiskers indicate the minimum and the maximum values, respectively, after exclusion of outliers. The asterisks indicate significant differences in metabolite ADC percentage change with respect to zero at a given time point (** = P < .01, *** = P < .001).

Table 4:

Estimated Marginal Means of ADC_rel for tNAA, tCr, and tCho of HVs and Participants with IS

Metabolite concentrations.—In participants with IS, the relative difference for tNAA concentration was lower than zero at 2 weeks, 1 month, and 3 months after IS, indicating lower tNAA concentrations in VOI_les at all three time points (−45.7%, −54.0%, and −57.6%, respectively; P < .001) (Fig 4, Table 5). Similarly, the relative difference for tCr concentration was lower than zero at 2 weeks, 1 month, and 3 months after IS (–30.2%, −28.2%, and −33.6%, respectively; all P < .001). The relative difference for tCho concentration showed a nonsignificantly higher value only at 1 month after IS (17.6%, P = .07) (Fig 4, Table 5).

Figure 4:

Box and whisker plots show concentrations of metabolites. Percentage variation of (A) total N-acetyl-aspartate (tNAA), (B) total creatine (tCr), and (C) total choline (tCho) concentrations in the volume of interest (VOI) of the lesion with respect to the contralateral VOI for participants with ischemic stroke (IS) at each time point (open circles) and for healthy volunteers (HVs) (filled circles). Percentage variation in concentration was calculated as (concentration_les – concentration_cl)/concentration_c, where concentration_les is the concentration in the lesion and concentration_cl the concentration in the contralateral region. For participants with IS, the values are plotted versus the number of days after IS. For each box, the midline indicates the median, and the bottom and top edges indicate the 25th and 75th percentiles, respectively. The bottom and top whiskers indicate the minimum and the maximum values, respectively, after exclusion of outliers. The asterisks in plots A and B indicate significant differences in relative metabolite concentrations with respect to zero at a given time point (**** = P < .0001). The asterisk in C indicates a significant change between 1 and 3 months after IS (* = P < .05). [tCho]_rel = percentage variation in tCho, [tCr]_rel = percentage variation in tCr, [tNAA]_rel = percentage variation in tNAA.

Table 5:

Estimated Marginal Means of Concentration Percentage Changes for tNAA, tCr, and tCho of HVs and Participants with IS

There were no changes over time in the relative differences for tCr and tNAA. Only the relative difference for tCho changed over time (Table S3). Post hoc tests indicated a significant decrease in the relative difference for tCho at 3 months compared with that at 1 month after IS, after a nonsignificant increase at 1 month compared with 2 weeks (16.7%, P = .08). Metabolite concentrations were not correlated with the volume of the lesion. Metabolite concentrations obtained in participants with IS and HVs are reported for both VOI_cl and VOI_les in Figure S2.

Discussion

We investigated longitudinal cell-specific microstructural changes in cerebral ischemia by measuring the temporal evolution of metabolite apparent diffusion coefficients (ADCs) in the infarcted brain up to 3 months after ischemic stroke. The most salient findings of our study were that the ADC of choline-containing compounds was higher in the ischemic lesion than in the contralateral side at 1 month after stroke, and the ADC of creatine and phosphocreatine was higher at both 1 and 3 months after stroke, while no differences in the diffusivity of these metabolites were observed at 2 weeks after stroke. Conversely, the ADC of N-acetyl-aspartate and N-acetyl-aspartyl-glutamate did not show any significant change between lesion and contralateral tissue at any stage of the disease.

The concentrations of tNAA and tCr in the ischemic lesion were significantly lower than those in the contralateral region and were stable over the three time points.

The higher diffusivity of both tCho and tCr in the lesion 1 month after stroke may be attributed to reactivity-related morphologic changes in glial cells linked to an inflammatory response. Glial cell hypertrophy induced by pathologic reactivity (18) leads to an increase in the intracellular space and could be the cause of the elevated diffusivity of glial metabolites in the cytosol (8–10). tCho and tCr ADCs were previously found to be abnormally high in the white matter of patients with neuropsychiatric systemic lupus erythematosus, which is characterized by microglial proliferation and reactive astrocytosis (8). More recently, the ADC of tCho, but not the ADC of tCr, was found to be significantly higher in the thalamus of healthy individuals after intravenous administration of lipopolysaccharide compared with placebo (19). Low-dose lipopolysaccharide is a well-established experimental method for inducing systemic inflammation, and it was shown by means of microscopy to activate microglia in rodents (20). Similarly, the ADC of tCho was significantly higher in the white matter of mice after a 6-week cuprizone intoxication than in controls, and this difference was associated with microglia and astrocyte activation assessed by means of immunohistochemistry (9). In contrast, no changes were observed in tCho diffusivity in a mouse model of pure astrocytic reactivity (10). Taken together, these results suggest that tCho diffusivity could be more specific to microglial morphologic changes than to changes in other glial cells. Although this hypothesis requires further corroboration, the changes in the ADC of tCho over time observed in our study may indicate that inflammatory processes involving microglial activation are still ongoing at 1 month after stroke, while they tend to normalize at the chronic stage of the disease.

In contrast to tCho, which is predominantly located in glial cells, tCr can be found in all cell types, including neurons; therefore, this metabolite is usually not associated with cell-specific morphologic changes. Nevertheless, previous studies suggested that the concentration of tCr in astrocytes is twice that in neurons (21–23). In addition, the core of the infarcted lesion is affected by severe neuronal death, as suggested by the known pathophysiology of ischemic tissue, as well as by the concentration data of the neuronal metabolite tNAA. In fact, tNAA levels in the ischemic lesion were reduced by more than 45% compared with that in the contralateral side at all time points. tCr levels were also significantly reduced, but to a lesser extent (approximately 30%). These findings are in line with previous studies reporting metabolite concentrations in IS (24,25) and further suggest that the ADC of tCr in this disease might reflect microstructural changes in the glia (and especially astrocytes) rather than the neurons.

At approximately 3 months after IS, only the ADC of tCr was elevated (+19.0%, P < .001). The difference in the behavior of the ADC of tCho and the ADC of tCr at the third time point of our experiment could be consistent with a different compartmentalization of tCho and tCr in glial cells, namely with higher tCho levels in microglia and higher tCr levels in astrocytes. Activated microglia and macrophages invade the ischemic lesion 1–2 days after stroke and are thought to persist for at least 1 week, while an astrocytic response was observed up to 1 month after stroke (3). The elevated ADC of tCr up to 3 months after stroke in our study may reflect the presence of astrogliosis at the chronic stage of the disease. Our data seem to be in line with the known temporal evolution of inflammatory processes but also suggest that active inflammation may last longer than what was suggested previously (3). Unfortunately, due to the low signal-to-noise ratio at 3.0 T and the long echo time used in our study, we were not able to quantify the ADC of myo-inositol (a predominantly astrocytic metabolite), which has been suggested to be the most specific marker of astrocytic reactivity (10).

Although inflammatory processes have been shown to already be occurring a few hours after stroke (1), the ADCs of both tCho and of tCr showed only a nonsignificant increase between the ischemic lesion and the contralateral tissues at approximately 2 weeks after the infarct. This moderate increase, together with the use of a stringent statistical test with Tukey multiple comparisons correction, partially explains the lack of significant longitudinal changes for these two metabolites. Also consistent with findings from previous animal studies, this result may indicate the presence at 2 weeks of competing mechanisms in the infarcted area, such as glial reactivity and ongoing apoptotic processes that began in earlier stages. The latter was suggested by clinical and preclinical diffusion-weighted MR spectroscopy studies on hyperacute and acute stroke, which reported decreased metabolite ADCs in the ischemic lesion compared with healthy tissue (17,26–29). In view of these previous reports, the unchanged ADC of tCho and ADC of tCr at 2 weeks may indicate that at this time, metabolite ADCs could be increasing from abnormally low values typical of the hyperacute stage to increased values characterizing the subacute and chronic stages (17). The temporal change in tCho concentration observed in our study also points to a similar dynamic, showing an increase at 1 month after stroke, followed by a renormalization at 3 months. However, no tCho concentration changes were observed at any time point, indicating that this marker is less suitable than ADC to monitor inflammation in stroke (9,19).

tNAA concentration data suggested that a fraction of neurons were still viable in the VOI located in the ischemic lesion. Indeed, although reduced by approximately 50%, the tNAA signal was still detectable in all participants with IS, in agreement with previous studies, also suggesting that the tNAA concentration gradually decreases from 1 to approximately 12 days after infarction, when it reaches its minimum (24,30). Our data showed a constant tNAA concentration and ADC over time, suggesting, respectively, that no further neuronal loss occurred 2 weeks after the ischemic event and that the viable neurons were not characterized by morphologic changes. Nevertheless, due to possible metabolite longitudinal T2 changes and the proportion of healthy tissue that increased in VOI_les with time, these results should be interpreted with caution and may have biased the estimation of metabolic levels in the lesions (31).

In general, concentration and ADC measurements are independent, as they are driven by different physiologic processes (concentration changes being mainly driven by changes in cellular density or metabolic dysfunction, and ADC changes being associated primarily with cell morphologic changes). This is well reflected in our study by the observed N-acetyl-aspartate and tCr changes, showing a decrease in their concentrations associated with unchanged tNAA or increased ADCs of tCr, which further stresses the added value of metabolite diffusion measurements to complement standard MR spectroscopy experiments.

Our study had limitations. First, we were not able to acquire diffusion-weighted MR spectroscopy data at the acute stage of IS. The mean delay between stroke onset and participant inclusion in the study was 10 days, which falls in the early subacute stage of the disease. Because the measurements were performed at a research center, inclusion at an earlier stage was not feasible. Second, for three participants with IS, diffusion-weighted MR spectroscopy data were incomplete at one time point and were therefore replaced using a multiple imputation technique. Third, our conclusions are based on a small cohort. The sample size was too small to analyze subgroups of patients according to the lesion topography or cause. Fourth, higher magnetic field strength or longer acquisitions would be beneficial to quantify the diffusion of other metabolites, such as myo-inositol and lactate.

In conclusion, our data suggest that active inflammatory processes are still ongoing at 1 month after ischemic stroke, possibly contributing to prolonged tissue injury beyond the acute stage of the disease. These findings are in line with previous evidence suggesting that postischemic inflammation is a potential target for therapeutic intervention, indicating the need to develop new strategies for improving repair and preventing further damage that may occur subacutely. Further investigations that include larger patient cohorts and a longer temporal window are needed to assess whether different metabolite apparent diffusion coefficient longitudinal behaviors correspond to different tissue and clinical trajectories.