Keywords

Introduction

Each year, approximately 800,000 people suffer a stroke in the United States. The causes of stroke are, in general, either hemorrhagic or nonhemorrhagic. Intracranial hemorrhage (ICH) is a common and often fatal stroke subtype [1, 2]. More than 30,000 patients die from spontaneous ICH annually. Iron has a major role in ICH-induced brain injury [1, 3, 4]. There is a progressive accumulation of iron in the cerebral tissue surrounding a hematoma from hemoglobin degradation within the hematoma as it resolves. The high level of non-heme iron remains in the brain for at least 1 month [5]. By enhanced Perls’ reaction, iron-positive cells are found in the perihematomal zone as early as the first day [5]. Our studies also have shown that free iron levels in cerebrospinal fluid (CSF) increase almost 14-fold after ICH on the third day and remain high for at least 28 days after experimental ICH [7]. Increases of brain iron levels cause brain edema, oxidative stress, brain atrophy, and neurological deficits following ICH [1, 6, 8].

If the patient survives the ictus, the resulting hematoma within brain parenchyma triggers a series of events leading to secondary insults and severe neurological deficits [9]. There is currently no proven therapy for ICH other than supportive care [10, 11]. Although the hematoma in humans gradually resolves, neurological deficits in ICH patients are usually permanent and disabling. Iron has a major role in brain damage following ICH [1, 11]. Brain injury after ICH appears to involve several phases [1]. These include an early phase involving the clotting cascade activation and thrombin production [10–16] and a later phase involving erythrocyte lysis and iron toxicity [2–4, 17–19]. After erythrocyte lysis within the hematoma, iron concentrations in surrounding brain can reach very high levels. Studies have shown that high levels of serum ferritin, an iron-binding protein, are independently associated with poor outcome and severe brain edema in ICH patients [16, 17].

It is well known that iron has a key role in brain edema formation following ICH [1, 19]. Perihematomal brain edema develops immediately after ICH and peaks several days later [6, 19]. Edema formation following ICH elevates intracranial pressure and may result in herniation [20]. In experimental ICH models, brain edema peaks around the third or fourth day after the hemorrhage then declines slowly [10, 21–23]. In species with significant white matter, perihematomal edema is mainly located within that tissue [22]. In humans, perihematomal edema develops within 3 h of symptom onset and peaks between 10 and 20 days after the ictus [23]. Several studies show that the degree of brain edema around the hematoma correlates with poor outcome in patients [20, 24, 25]. We have shown that desferoxamine, an iron chelate, reduces brain edema and hemorrhagic brain injury in aged rats and pigs [26–28]. Deferoxamine has been examined by a phase I trial in human subjects [29]. The promise it has shown in these trials has led to a further phase II trial, which was approved by the NIH.

Although there is a growing body of evidence for the role of iron in neuronal damage following parenchymal hemorrhage, there is a conspicuous lack of a reliable paradigm for accurate quantification of the tissue iron load. Interest in noninvasive quantification of tissue iron has resurged with development of effective iron chelate therapy in prevention of iron toxicity [30]. Existing noninvasive modalities such as magnetic resonance imaging (MRI) have been utilized to develop iron measurement algorithms in the heart [31]. Traumatic brain injury is another condition where MRI-based estimation of iron deposition in the cerebral tissue has been explored [32, 33]. Human physiology is incapable of excreting excess iron. Iron is deposited in various tissues in the body in clinical conditions that require repeated blood transfusions. Paramagnetic effects of tissue-deposited iron, causing signal inhomogeneity on MRI, have been exploited to estimate tissue iron levels in human visceral tissue [30, 31, 34]. Although some advances have been made in quantifying iron in the human liver by MRI, a robust technique has not been developed yet [35, 36]. Bilgic et al. [37] have demonstrated feasibility of detection of iron levels in brain tissue of normal aging population by quantitative susceptibility mapping (QSM). They demonstrated specificity of the field-dependent relaxation rate (FDRI) technique in detection of minute concentrations of iron in the brain tissue with age. Preliminary clinical studies have demonstrated the ability to detect cerebral tissue iron in traumatic brain injury by utilizing T2* magnetic field correlation on MRI [32]. Liu et al. [38], utilizing a porcine model, demonstrated the ability of R2 (1/T2) relaxivity maps in detecting superparamagnetic iron oxide (SPIO) nanoparticles in the brain tissue following disruption of blood brain barrier (BBB) by low-frequency ultrasound. Sammet et al. [39] showed the importance of multiple spin echo sequences on MRI for reliable correlation with iron levels and developed a mathematical model to differentiate between ferritin and hemosiderin on MRI using an agarose-phantom. In a porcine ICH model, utilizing 7 T MRI, Wu et al. [40] demonstrated the application of T2* sequences to correlate with hematoma size and histopathological evidence of tissue iron. A study from Wang et al. [41] showed the reliability of QSM on T2* MRI scan in patients with ICH. They utilized the application of R2* (1/T2*) relaxivity maps to obtain a reliable calculation of hematoma volume.

At the time of this writing, there have been no studies performed to ascertain accurate tissue iron levels at the periphery of the hematoma in patients with ICH. Our hypothesis is that MRI is sensitive enough to pick up field inhomogeneity due to the presence of excess brain tissue iron. A reliable MR-based iron quantification algorithm can be beneficial in management of patients with ICH. It can serve as a surrogate marker of severity of neurotoxicity. Once validated, it can serve to monitor efficacy of the iron-chelating agents like deferoxamine currently being tested in this patient population. Above all, it may possibly enable prediction of ICH in patients if correlation of brain tissue iron levels can be demonstrated with functional outcome in the future.

Materials and Methods

MRI Phantom Construction

Institutional Review Board approval was obtained to perform MRI scans on selected patients with ICH. Initially, we constructed a phantom with varying concentrations of iron and then scanned the phantom in a 3 T MRI. The phantom was constructed with a cylindrical container made out of material that would not cause signal aberration on MRI; a plastic cylindrical container was utilized. A liquid iron preparation, ferumoxytol (Feraheme®, AMAG Pharmaceuticals, Waltham, MA, USA), which is used for intravenous ferrous sulfate therapy, was selected as the appropriate liquid preparation. It is a non-stoichiometric magnetite (superparamagnetic iron oxide) coated with polyglucose sorbitol carboxymethylether.

We then prepared serial dilutions of the ferumoxytol in seven different reducing dilutions, commencing with 6 mg/ml and ending with 0.0047 mg/ml. Each subsequent concentration was 50 % less than the prior. The above concentrations were chosen as the iron concentrations in the published literature lie in the middle of the above distribution. Four milliliters of each concentration was placed in a 4-ml glass vial, avoiding any air bubbles. Seven glass vials were then attached to the undersurface of the lid of the plastic container selected for the phantom (Fig 1a), which was then filled with sterile water to replace any air bubbles in the container. The phantom was scanned in a 3 T MRI with the following susceptibility weighted sequences: 3D TR = 40 ms, TE = 6.5, 11, 15.5, 20, 24.5, 29, 33.5, 38 ms, 1.5 mm slice-to-slice, acquired as 3 mm, acquired resolution matrix = 240 × 240, FOV – 240 mm × 240 mm (Fig 1b). T2* signal magnitude measurements were then obtained corresponding to these concentrations (Table 1).

Fig. 1
figure 1

(a) Image of the lid of the phantom with seven 4-cc vials containing serial dilutions of iron concentration in sterile water stuck to the undersurface. (b) MRI of the phantom on a 3 T scanner with following sequence specifications: 3D TR = 40 ms, TE = 6.5, 11, 15.5, 20, 24.5, 29, 33.5, 38 ms, 1.5 mm slice-to-slice, acquired as 3 mm, acquired resolution matrix = 240 × 240, FOV – 240 mm × 240 mm

Full size image
Table 1 Seven serial dilutions scanned in the MRI phantom with T2* signal measurements corresponding to each concentration
Full size table

Human Subjects Scanned

ICH patients, age 18–85 years, with brain parenchymal hemorrhage, no previous ICH, and no evidence of physiological calcification on noncontrast CT of the head were included in the study. One control and two human subjects who met the inclusion criteria for the study were scanned with the same MRI protocol utilized on the phantom. The two human subjects had spontaneous basal ganglia hemorrhage and were scanned on day 7 of their hemorrhage. In the control human brain, MRI regions of interest (ROIs) were drawn in the basal ganglia region on both hemispheres. Calculations were performed on R2* maps generated from the MRI sequences (Fig. 2a). In the two human subjects, regions of interest were drawn on the periphery of the hematoma in the left basal ganglia (Fig. 2b, c). Region of interest measurements were also performed on the contralateral normal hemisphere in an identical anatomical location.

Fig. 2
figure 2

(a) A 70-year-old male patient with axial MRI brain showing left basal ganglia hemorrhage with regions of interest drawn for measurement of the R2* magnitude. (b) A 68-year-old male patient with axial MRI brain showing left basal ganglia hemorrhage with regions of interest drawn for measurement of the R2* magnitude. (c) A 45-year-old male subject utilized as control showing axial MRI with regions of interest drawn for baseline R2* magnitude measurement

Full size image

Results

Control Human Subject

In the right basal ganglia, the T2* average of three ROIs measured 44.3. In the left basal ganglia in the same control human subject, the average of three ROIs measured 53.2. Both of these measurements correspond to an iron concentration of 0.01 mg/ml.

First Human Subject with ICH

Perihematomal measurements performed with three ROIs in the left basal ganglia showed T2* values of 15.3. This corresponds to an iron concentration of 0.04 mg/ml. An average of three ROIs in the contralateral normal brain in an identical anatomical location measured 49.5. This corresponds to 0.01 mg/ml of iron concentration.

Second Human Subject with ICH

Perihematomal measurements performed with three ROIs in the left basal ganglia showed T2* values of 18.97. This corresponds to an iron concentration of 0.04 mg/ml. An average of three ROIs in the contralateral normal brain in an identical anatomical location measured 53.3. This corresponds to 0.01 mg/ml of iron concentration.

The above measurements demonstrate consistently that the detection of an iron concentration in the periphery of the hematoma of an ICH in the left basal ganglia on day 7 following the ictus are 4 times higher than the normal baseline concentration.

Conclusion

Our experiment demonstrates proof of principle of MRI being able to detect a 4 times increase in tissue iron levels in the periphery of the hematoma in comparison to baseline. The initial translation from bench to bedside of iron-chelating therapy with deferoxamine being investigated by a phase II trial holds promise. Our hypothesis, once validated in a larger study, can provide a surrogate marker of severity of neurotoxicity following an ICH. Moreover, an MRI-based brain tissue iron quantification may provide a more objective way of monitoring therapy with iron chelates. Furthermore, the correlation of tissue iron quantification with functional outcome following an ICH needs to be studied in a large-scale prospective analysis involving human subjects.