Assessing early erythrolysis and the relationship to perihematomal iron overload and white matter survival in human intracerebral hemorrhage

Abstract Aims Iron released from lysed red blood cells within the hematoma plays a role in intracerebral hemorrhage (ICH)‐related neurotoxicity. This study utilizes magnetic resonance imaging (MRI) to examine the time course, extent of erythrolysis, and its correlation with perihematomal iron accumulation and white matter loss. Methods The feasibility of assessing proportional erythrolysis using T2* MRI was examined using pig blood phantoms with specified degrees of erythrolysis. Fifteen prospectively enrolled ICH patients had MRIs (3‐Tesla) at days 1–3, 14, and 30 (termed early, subacute, and late periods, respectively). Measurement was performed on T2*, 1/T2*, and fractional anisotropy (FA) maps. Results Pig blood phantoms showed a linear relationship between 1/T2* signal and percent erythrolysis. MRI on patients showed an increase in erythrolysis within the hematoma between the early and subacute phases after ICH, almost completing by day 14. Although perihematomal iron overload (IO) correlated with the erythrolysis extent and hematoma volume at days 14 and 30, perihematomal white matter (WM) loss significantly correlated with both, only at day 14. Conclusion MRI may reliably assess the portion of the hematoma that lyses over time after ICH. Perihematomal IO and WM loss correlate with both the erythrolysis extent and hematoma volume in the early and subacute periods following ICH.

functional outcome with image-guided hematoma evacuation following ICH have so far failed, although there is evidence of reduced mortality. 7 More than two decades of basic science research in animal ICH models has identified iron as a major neurotoxin released from the hematoma after ICH. [8][9][10][11] However, those findings have yet to translate to patients. A lack of significant benefit in the deferoxamine phase II treatment trial in ICH 12 suggests that our understanding of iron-induced injury after ICH is still incomplete.
Neuroimaging has a potentially significant role in unraveling some of the missing links in our understanding of the natural history of human ICH, particularly in relation to iron, given its paramagnetic properties. Magnetic resonance imaging (MRI) has been used to assess perihematomal iron following ICH. [13][14][15][16][17] A pilot study using MRI demonstrated that hematoma volume impacts iron overload in the surrounding perihematomal tissue. 13 Another parameter that may impact perihematomal iron after ICH is the rate of erythrolysis within the hematoma. The phenomenon of ultra-early erythrolysis has been demonstrated in animal ICH models and in human ICH on MRI. 11,18,19 The aims of the current study were to determine whether MRI can be used to quantify erythrolysis in the hematoma in human ICH, evaluate the time course of erythrolysis, and examine whether the extent of erythrolysis correlates with perihematomal iron overload and white matter loss. No such analysis has been performed to date, and our study is part of a larger focus on developing novel imaging prognostic markers for ICH.

| Institutional approval and patient selection
The University of Michigan Institutional Review Board approved the study, and all patients consented to be included. Prospective patient recruitment commenced in early 2013 to acquire pilot data to translate the understanding of the ICH natural history from animal ICH models to human subjects. The pilot project with 3 patients and one control (normal human subject) led to successful application to the National Institutes of Health with secured funding in terms of two R21 grants in 2017 and 2018. Until the beginning of 2020, a total of 15 patients were recruited. All patients with basal ganglia hemorrhage were included, and MRI was performed on days <3, 14, and 30 following the hemorrhage. Exclusion criteria were previous hemorrhage or calcification in the basal ganglia region on noncontrast head CT and any contraindication for performance of MRI scans. Fifteen patients recruited to the study were analyzed. Not all patients had MRIs performed on all the prespecified data points, and not all screened and eligible patients were included in the study due to refusal to join.

| Magnetic resonance imaging
Brain MRI examinations were performed on a 3T MRI system using a 32-channel head coil for acquisition of standard 3D T1weighted, T2-weighted, and fluid-attenuated inversion recovery sequences. White matter integrity was probed using 32 direction b-values 0-800 s/mm 2 , single-shot echo planar diffusion tensor imaging for generation of fractional anisotropy (FA) maps along with other diffusion/anisotropy metrics at 1 × 1 × 2.3 mm 3 resolution. Relaxivity maps, sensitive to the presence and distribution of iron, were generated from a 3-dimensional (3D) eightecho gradient echo scan at following timings: TR = 40 ms and TE = 6.5 ms + (n-1) × 4.5 ms, where n = 0, 1,…, 7. Quantitative maps R2* (in Hertz units) and T2* (in millisecond units) at a pixel resolution 0.83 × 0.83 × 1.5 mm 3 were generated by monoexponential fit to pixel signal decay versus TE.

| Examination of erythrolysis within the hematoma
For each subject and time point, a volume of interest (VOI) sphere was manually defined on all slices along the hematoma boundary F I G U R E 1 Porcine blood phantom vials containing different percentages of lysed red blood cells (RBCs) and one vial with purified water were evaluated using T2* MRI. T2* maps were created, and 1/ T2* = R2* (Hz) values determined. A, An example of such analysis, with colorcoded image for 1/T2*. The percentage of cell lysis is shown against each vial. B, A plot of 1/T2* against the percentage of lysed RBCs shows a linear relationship (p = 0.0013, R 2 = 0.979). Values are represented as means ± standard deviation, n = 6 on 1/T2* maps. A custom analysis routine written in MATLAB (version R2015b; MathWorks ® ) was used to calculate the "nonhypointensity" volume within the hematoma as follows. A display of the slice containing the greatest hematoma cross-sectional area based on the VOI mask was used to prompt manual definition of a circular contralateral region (nominal area ~425 mm 2 ) in the normal brain on the first-echo (TE = 6.5 ms) image. A circle of interest at each slice was automatically collated into a sphere of interest including all the slices of interest showing the hematoma.
An identical sphere of interest was applied to the contralateral anatomically identical location. Calculation of (mean -standard deviation) for pixels within the contralateral region on the first-echo image was used as a "reference" signal. The measurement on the contralateral basal ganglia region, considered to be normal, was deemed a value of 100% signal. The threshold was chosen as 1 standard deviation below the basal measurement on the contralateral side. Anything at or above this threshold was deemed to be a non-hypointense signal within the measured hematoma volume. Pixels on the first-echo 3D volume within the hematoma VOI having a value greater than the reference signal were summed to estimate "non-hypointensity" volume for the hematoma. A porcine blood phantom was utilized to test the hypothesis of lack of susceptibility from lysed red blood cells (RBCs). Fresh porcine blood from a euthanized swine was collected. Packed RBCs were obtained by centrifuging unclotted blood. The plasma and buffy coat were discarded. The RBCs were washed with five volumes of saline three times. To prepare lysed RBCs, packed RBCs were frozen in liquid nitrogen for 5 min, followed by thawing at 37℃.
Various proportions of lysed and unlysed blood were prepared in 4-mL vials. These were then stuck on the undersurface of the lid of the water bath phantom. Pure nonionic water was utilized to prepare the water bath. This phantom was then scanned in a 3T MRI scanner with a protocol identical to the one applied for humans. Maps of 1/T2* were then created and analyzed ( Figure 1).
The authors hypothesize that the MRI neutral environment of the water bath mimics the brain parenchyma in human ICH. Moreover, the contralateral normal anatomy was used to normalize the data measured within the hematoma and its surrounding tissue on the side of the ICH.

| Perihematomal iron overload
Perihematomal iron overload was determined, as described previously. 17 In brief, on MRI phantoms, there is a linear relationship between R2* and iron concentration. 13

| Perihematomal white matter loss
White matter loss was assessed using FA maps of two spheres, one to estimate an expected volume of white matter in the perihematomal tissue. ICH-induced white matter loss (in milliliters) was determined as the difference between that number and the measured perihematomal white matter volume.

| Statistics
The relationships between erythrolysis and other imaging parameters were examined by regression analysis. Comparisons between means were conducted by using analysis of variance (ANOVA) if the data passed a test of normality (Kolmogorov-Smirnov test). For non-normal data, Mann-Whitney and Wilcoxon nonparametric tests were employed (as indicated in text). Statistical significance was taken as p < 0.05. This was a pilot study, and a power analysis was not performed.

| RE SULTS
MRIs on pig blood phantoms demonstrated a graded decrease in susceptibility (i.e., bright signal) as the blood sample in the vial ap-  Figure 2C (day 3, green) and Figure 2D (day 14, red) and quantified in Figure 2E. As seen in the latter, there is a marked increase in the percent erythrolysis between patients exam-  Figure 3A). In contrast, at day 14, there was a very tight correlation (p < 0.0001, R 2 = 0.9850; Figure 3B).  Figure 5A) and hematoma volume (p = 0.0006, R 2 = 0.834; Figure 5B).

| DISCUSS ION
This study has several major findings. (1) Pig blood phantoms show that T2* imaging can be used to quantify the degree of erythrolysis.
(2) In ICH patients with basal ganglia hemorrhage, the percentage erythrolysis within the hematoma progressively increases from signal ( Figure 1). We, therefore, suggest that regions of T2* nonhypointensity within the hematoma can be a surrogate marker for early erythrolysis. We propose that hemoglobin-bound iron loses its ability to create susceptibility in the electromagnetic field of the MRI after cell lysis (and hemoglobin dispersion). Thus, the MRI T2* maps show a bright to isointense signal and not a typical dark signal from iron and its ensuing susceptibility. Whether any phenomenon other than erythrolysis can cause such a change in the T2* signal within hematomas needs to be investigated. However, our study is the first of its kind to quantifiably translate the benchside understanding of ultra-early erythrolysis in small animal ICH models ("ghost RBC" shown on histology), demonstrating it in a porcine (large animal) blood phantom and then in ICH patients.
In the current study, the degree of erythrolysis in the hematoma increased markedly between the early (days 1 and 3) and subacute  (Table 1). Whether different rates of erythrolysis might contribute to such variation merits further investigation. However, it will require a much larger sample than that in the current study. shown to attenuate both early erythrolysis and brain injury in a rat ICH model. 11 It would be important to examine if such an approach would also reduce perihematomal white matter loss in a gyrocephalic species.
The current study used novel methods to assess the change in erythrolysis in the hematoma and its relationship with perihemato- Currently, CT-based assessment of hematoma size and surrounding edema lacks granularity. 22 We suggest that MRI-based assessment provides a more robust assessment of various markers of tissue toxicity (perihematomal iron overload and white matter loss) in the perihematomal region but also can be correlated to cellular events such as erythrolysis occurring within the hematoma and the hematoma size itself. It is encouraging to find the temporal trends toward peak effects of erythrolysis by day 14 of ICH can be reliably tracked.
Our study has several limitations, most important being the small

| CON CLUS ION
Our study shows that multiparametric assessment of the hematoma and the surrounding tissue by MRI could provide tissue injury markers for a more informed assessment of extent of ICH-induced damage. Assessment of hematoma volume and erythrolysis volume within the hematoma may indicate the severity of the toxic onslaught inflicted on the surrounding brain tissue. Tracking these parameters on MRI over a month might provide insights into peaks of erythrolysis within the hematoma and iron overload on the surrounding tissue and the surviving white matter tracts therein. A larger human ICH cohort analysis of the aforementioned parameters may provide robust objective surrogate markers of tissue injury.

ACK N OWLED G EM ENTS
This work was supported by grants NS099684 and NS104663 from the National Institutes of Health.

CO N FLI C T S O F I NTE R E S T
The authors declare that they have no conflict of interest.