William G. Bradley, Jr., M.D., Ph.D. Effect of Formation Paul G. Schmidt, Ph.D. on the MR Appearance of Subarachnoid Hemorrhage1

Subarachnoid hemorrhage has a much T HE EARLIEST reports of magnetic resonance (MR) images of in- higher intensity in magnetic resonance tracranial hemorrhage (1, 2) suggested a relatively intense ap- (MR) images with the passage of time. peanance of the hemorrhage relative to the surrounding brain. This Acute is diff i- was attributed to the short Ti relaxation time of the presumably cult to see; within 1 week its appearance paramagnetic, iron-containing . Figure 1 illustrates this has become intensified on Ti-weighted appearance in a patient imaged 1 week after rupture of an aneurysm images. Different concentrations of blood in an anterior communicating artery, with resulting intraparenchy- and lysed red blood cells in cerebrospinal mal hematoma and subarachnoid hemorrhage. The short Ti charac- fluid (CSF) were examined spectroscopi- ten of the injured area is enhanced relative to surrounding brain on cally but did not significantly alter Ti a Ti-weighted spin-echo image. and T2 relaxation of CSF acutely. Ultra- As experience was gained, subsequent reports (3, 4) indicated violet visible spectroscopy of bloody CSF that acute intracranial hemorrhage could be much more difficult to stored hypoxically for 3 days showed the detect on MR images. This was attributed by Sipponen et al. (3) to a presence of methemoglobin. The iron in lack of Ti shortening during the acute phase; however, they at- methemoglobin is paramagnetic; in com- tempted no explanation of this phenomenon. They reported that bination with water this facilitates Ti re- acute subarachnoid hemorrhage was difficult to detect by MR laxation. It is concluded that methemo- study, particularly in comparison with computed tomography (CT) globin formation with Ti shortening at (4). Although DeLaPaz et ai. (4) agree that acute intracranial hemor- least partially accounts for the increasing rhage is difficult to detect on MR images, they disagree as to the intensity of the MR appearance of subar- mechanism causing the intense appearance. While the data of Sip- achnoid hemorrhage over time in the cen- ponen et al. (3), ‘Bailes et al. (1), and Bydder et al. (2) suggest a Ti- tral nervous system and may also explain shortening process, the data of DeLaPaz et a!. (4) suggest no change the intense appearance of subacute hem- in Ti but rather a prolongation of T2, both of which would increase orrhage in MR images elsewhere in the the intensity on spin-echo images. Figure 2 illustrates the rather body. subtle increased intensity present in acute subarachnoid hemor- rhage 17 hours after ictus. One week after the acute subarachnoid Index terms: Brain, hemorrhage #{149}Hemorrhage, hemorrhage, the appearance is significantly more intense (Fig. 3). magnetic resonance studies #{149}Red blood cells To attempt an understanding of the variable appearance of subar- achnoid hemorrhage on MR images, the structure of hemoglobin Radiology 1985; 156:99-103 and its various breakdown products must be considered in some detail. In its circulating form, hemoglobin alternates between the oxy and deoxy forms as oxygen is exchanged during its transit through the high-oxygen environment of the lungs and low-oxy- gen environment of the capillary circulation. To bind oxygen re- vensibly, the iron in the hemoglobin (heme iron) must be main- tamed in the reduced ferrous (Fe2) state (5). To do this, the red blood cell maintains several metabolic pathways to prevent various oxidizing agents from converting its heme iron to the nonfunc- tional fernic (Fe3) state. When hemoglobin is removed from the circulation, these metabolic pathways fail and the hemoglobin mol- ecule begins to undergo oxidative denaturation. 1 From the MR Imaging Laboratory, Huntington The heme iron normally is suspended in a nonpolar crevice in Medical Research Institutes (W.G.B., P.G.S.) and the the center of the hemoglobin molecule. It is held in this position by Department of Radiology (W.G.B.), Huntington Me- morial Hospital, Pasadena, California. Presented at the a covalent bond with a histadine at the so-called F8 position of the 70th Scientific Assembly and Annual Meeting of the globin chain and by four planar hydrophobic van den Waals bonds Radiological Society of North America, Washington, with various nonpolar groups on the globin molecule. The group D.C., November 25-30, 1984. Received January 11, attache4 to the sixth coordination site of the heme iron varies. It is 1985; accepted and revision requested February 1 1; re- vision received March 11. occupied by molecular oxygen in oxyhemoglobin; it is vacant in C RSNA, 1985 deoxyhemoglobin (Fig. 4). As the oxidative denaturation of hemo-

99 globin proceeds, the ferrous heme ing the red blood cells; to do this, fresh Figure 1 iron is oxidized to the fernic state and venous blood was repeatedly passed methemoglobin is formed. The five through a 25-gauge needle before being mixed with CSF. The magnetic relaxation bonds to the globin molecule are un- times of oxy- and deoxyhemoglobin were changed; the sixth coordination site is compared by bubbling either oxygen or now occupied by either a water mole- nitrogen through fresh solutions of cule or a hydroxyl ion, depending on bloody CSF. Methemoglobin was pro- whether the methemoglobin is acidic duced by treatment of deoxyhemoglobin or basic. At physiologic pH, the acid with sodium nitrite (NaNO2). Before mea- form predominates. With continued surement, all samples were agitated to sus- oxidative denaturation, methemoglo- pend the red cells. All measurements were bin is converted to derivatives known performed at 38#{176}C. as hemichromes (5). While the iron in Quantitation of subarachnoid hemor- rhage generally is performed by means of these compounds remains in the fer- light (visible) spectroscopy (8). Such anal- nc state, alteration of the tertiary ysis provides the “xanthochromic” index structure of the globin molecule oc- that is used to quantitate the degree of curs, with the result that the sixth co- hemorrhage. The xanthochromic index is ordination site of the heme iron is oc- the sum of the absorption values at 415 nm cupied by a ligand from within the (oxyhemoglobin) and 460 nm (). Subacute subarachnoid hemorrhage. Intra- globin molecule (most likely the dis- All samples are prepared for spectropho- parenchymal hematoma (arrow) and subar- tal histadine at the E7 position). tometric analysis by centrifugation, with achnoid hemorrhage (arrowhead) are noted To consider the effect of subarach- examination of only the supernatant. By 1 week after rupture of aneurysm in anterior such analysis, little methemoglobin has noid hemorrhage on MR imaging communicating artery. The contrast between previously been found either in acute su- the lesions and the surrounding brain is en- over time, one must evaluate the in- barachnoid hemorrhage or within several hanced on this Ti-weighted spin-echo image tenaction between the iron-contain- weeks after the hemorrhage (8, 9). (TR = 0.5 sec; TE 28 msec). ing hemoglobin within the red blood cells and water protons in the cere- brospinal fluid (CSF). A trivial expla- nation for the Ti shortening apparent on images of subacute hemorrhage might be the lysis of red blood cells that can occur as a result of their expo- sure to phospholipases in the CSF. We can readily exclude this mechanism for enhanced proton relaxation be- cause water molecules are already al- lowed access to the hemoglobin by rapid transit across the red cell mem- brane (6, 7). Thus, changes in proton relaxation enhancement effects may be attributed directly to changes in the interactions between the CSF wa- ten protons and the heme iron that occur as a result of oxidative denatun- ation of hemoglobin. In view of the discrepant reports in the MR imaging literature and the lack of an obvious explanation from a. b. the literature on MR and ultraviolet- Acute subarachnoid hemorrhage 17 hours post ictus. a. Midline sagittal section shows minimally increased intensity in the pontine and medul- visible (UV) spectroscopy for the lary cisterns (arrowheads). The intensity of the CSF in these subarachnoid spaces should changing appearance of subarach- . be the same as that in the fourth ventricle (arrow). noid hemorrhage, the following stud- b. Coronal section demonstrates minimally increased intensity in the left sylvian cistern on ies were undertaken. this TR 1.5-sec. TE 28-msec image.

The solutions of oxy- and deoxyhemog- CSF) and compared with the known stan- lobin and methemoglobin were analyzed dards previously produced. MATERIALS AND METHODS using a Perkin-Elmer UV-visible spectro- photometer. Attention was directed to the Subarachnoid hemorrhage was mod- 630-nm region, which is specific for meth- RESULTS eled in vitro by adding fresh venous hu- emoglobin. man blood to artificial CSF, producing a To assess the change in magnetic relax- The effect of concentration of blood 10% (by volume) solution. The Ti and T2 ation times for bloody CSF stored hypoxi- in CSF is demonstrated in Figure 5, relaxation times of this solution were eval- cally, a 10% solution of lysed red blood which shows a small (10%) decrease uated using an IBM Minispec desktop cells in CSF was measured over several in Ti and T2 relaxation times as the spectrometer operating at 20 MHz. The ef- days. In a second experiment, a 20% solu- concentration is increased from 0% fect of red blood cell concentration on Ti tion of whole red cells in CSF was again (pure CSF) to 10% red blood cells. and T2 relaxation was evaluated by mea- stored hypoxically for several days with suring the relaxation times at concentra- sequential measurement of Ti and T2 re- Such Ti shortening cleanly is not the tions varying from 0% (pure CSF) to 10% laxation times. Ti shortening was demon- primary mechanism for the increased by volume. The effect of red blood cell strated. Spectrophotometric analysis was intensity observed in subarachnoid lysis was evaluated by mechanically lys- performed on the “unknown” (bloody hemorrhage clinically. Figure 5 also

100 #{149}Radiology July 1985 Figure 3 Figure 4 0 o

Oxyhemoglobin .> Fe K

11 > Fe <

Deoxyhemoglobin

H20 0H II <> :i+ <> Subacute subarachnoid hemorrhage 1 week post ictus. Intensity of the CSF in the left < ‘]J(F8) “JJ(F8) sylvian cistern is significantly greater than Methemoglobin observed during the first 24 hours post ictus. Acid Base (TR = 1.5 sec. TE = 28 msec.) N(E7)

Reversible1.1 Figure 5 Hemichromes

1.6 oxygen. Following oxidation to the paramagnetic (ferric) LEGEND form as methemoglobin, heme iron can no longer bind oxy- 0 T1-LYSED gen and is thus nonfunctional. Continued oxidative dena- 1.4 #{149}T1-WHOLE turation produces hemichromes, which are low-spin ferric

0 T2-LYSED compounds with the sixth coordination site occupied by a

U 12-WHOIE ligand from the now denaturated globin chain.

0 measurement. Figure 7 shows Ti i40 years ago (10); 90 years later Pau- shortening of a 20% solution of whole ling and Coryell (ii) considered the

WHOLE red blood cells in CSF stored hypoxi- magnetic properties of blood in the cally for 160 hours. The Ti value con- fluid state. By using a capillary tube

0.6 - tinues to shorten for 90 hours and filled with either oxy- or deoxyhe- then plateaus. moglobin suspended between the Spectrophotometny of both un- poles of an electromagnet, they were 0 2 4 6 8 10 12 known solutions of bloody CSF dem- able to determine that deoxyhemog- 61000 CONC. 1% V/V) onstrate strong absorption at 360 nm, lobin was paramagnetic (attracted to Change in magnetic relaxation times (20 called the Soret band (Fig. 8). Strong the stronger pant of the magnetic MHz) as a function of red blood cell concen- absorption is present for all forms of field), while oxyhemoglobin was dia- tration in CSF and red blood cell lysis. A iO% decrease in the Ti (above) and T2 (below) hemoglobin in this region; thus it is magnetic (repelled from the stronger times is noted as a function of increasing not specific. That portion of the spec- part of the magnetic field). These ob- concentration of red blood cells in CSF from tnum more specific for methemoglo- servations led them to describe the pure CSF to 10% by volume. There is no sig- bin is the 630-nm region, which is various electron spin states of oxy- nificant effect of red blood cell lysis on Ti or shown in Figure 9. Here temporal in- and deoxyhemoglobin. In deoxyhe- T2 relaxation. crease in a broad peak is observed in moglobin, the heme iron is in the the “unknown” 20% bloody CSF solu- “high spin” ferrous state character- demonstrates that the effect of red tion, which corresponds to increasing ized by six electrons in the outer (d) blood cell lysis is negligible. methemoglobin concentration. Al- shell, four of which are unpaired. As shown in Figure 6, Ti measure- though there is significant difference When oxygen is added, one of the ments of oxy- and deoxyhemoglobin in the height of the 630- and 360-nm electrons is partially transferred to are quite similar, confirming what peaks, this reflects differences in the the oxygen molecule, resulting in a has been reported previously (iO). extinction coefficients rather than low spin form with a single unpaired Methemoglobin in a iO% solution of differences in concentration. Thus, by electron in the outer shell (5). lysed red blood cells in CSF is seen to comparison with known standards, Although the static susceptibility have a significantly lower Ti. No sig- the peak observed at 92 hours at 631 test performed by Pauling and Con- nificant difference in T2 was found nm is shown to correspond approxi- yell demonstrates that deoxyhemog- among oxy-, deoxy-, and methemog- mately to 90% methemoglobin. lobin is paramagnetic, this does not lobin. Figure 6 also illustrates the se- ensure a proton paramagnetic en- quential decrease in Ti for the 10% DISCUSSION hancement effect in aqueous solution. “unknown” solution of lysed red The magnetic properties of dried Such an effect was originally de- blood cells in CSF oven 84 hours of blood were first evaluated by Faraday scnibed by Bloembengen et al. (6) and

Volume 156 Number 1 Radiology #{149}101 requires not only that a panamagnetic Figure 6 Figure 7 center be present but also that it be OXYGENATION AND HYPOXIC STORAGE, 38” accessible to surrounding water pro- OXIDATION tons. The quantitation of this effect 10% LYSED BLOOD IN CSF 10% LYSED BLOOD IN CSF requires consideration both of the 38#{176}

. DEOXY magnitude of the magnetic moment 2.0 OXY FRESH of the paramagnetic dipole (the num- E ben of unpaired electrons), the elec- 1.8 . 66HRS tron spin relaxation rate (in effect, the Ti of the electron), the concentration 1.6 of paramagnetic dipoles, the average distance from surrounding water pro- 1.4 tons, and the relative motion of the proton and panamagnetic centers. 1.2 - Changes in Ti relaxation time during mod- Such theories of proton relaxation by eled subarachnoid hemorrhage. Changing paramagnetic solute ions are based on 1.0 Iii Ti relaxation time for solution of 20% whole translational diffusion and the dis- blood in CSF is studied for 160 hours and is tance of closest approach of the pro- Relaxation times for blood in CSF at 20 MHz. seen to decrease to a plateau value at approxi- ton and panamagnetic ions, which de- The Ti times for oxy- and deoxyhemoglobin mately 90 hours. termines an “outer sphere” of are similar; the Ti time for methemoglobin (produced by treatment with NaNO2) is sig- influence (12). It has also been shown nificantly less than that of oxy- and deoxyhe- that there can be a contribution to ne- moglobin. The Ti time of the 10% solution of laxation from exchange between sol- lysed blood and CSF is studied over a period Figure 8 vent and water ligands in the first co- of 84 hours during hypoxic storage and is OXY HEMOGLOBIN ordination sphere of the paramag- seen to decrease. netic ion, that is, “inner sphere ef- fects” (i2). Thus, while the deoxy form of hemoglobin is considered panamagnetic on the basis of static intense, short Ti appearance of suba- susceptibility experiments (ii), the cute or chronic hemorrhage observed Ti relaxation times of aqueous solu- elsewhere in the brain and the nest of tions of oxy- and deoxyhemoglobin the body. do not show a difference in proton Although we believe this is com- panamagnetic relaxation effects (13). pelling evidence, it should be empha- Methemoglobin, on the other hand, sized that the experiment described causes significant Ti shortening in here may differ somewhat from the in aqueous solution due to a combina- vivo clinical situation. Various en- tion of both “inner sphere” and “out- zyme systems may affect the subar- en sphere” effects (12). Thus Ti relax- achnoid space that are not present ation by methemoglobin is due to a here (14). In addition, red blood cells combination of ligand-exchange ef- may be cleared from the subanach- fects (from the water molecule at the noid space more rapidly (15) than is sixth coordination site) and from out- required to produce Ti shortening by en sphere diffusional effects, perhaps methemoglobin formation. Thus, ex- 800 100 600 500 400 300 A(nm) because solvent protons have in- tension of this explanation to the Spectrophotometry of oxy- and methemoglo- creased access to the heme iron clinical situation, while strongly sug- bin and hypoxically stored blood in CSF. through the nonpolar crevice (12). gested, has not been fully verified. Strong absorption is noted near 400 nm Ti shortening has been demon- We have detected methemoglobin in (Soret band), which is nonspecific. The spe- strated during hypoxic storage of two a subacute subdunal hematoma and cific absorption for methemoglobin is barely solutions of red blood cells in CSF believe it to account (at least partially) perceptible at 630 nm in both the methemog- lobin and the unknown bloody CSF solution. that were intended to simulate subar- for the Ti-shortening observed clini- achnoid hemorrhage. The same Ti cally. Spectrophotometnic and MR shortening has been observed clini- analysis of clinical subanachnoid cally, as shown in Figures 1-3. In con- hemorrhage are in progress. or is adsorbed to the lysed red blood trast to DeLaPaz et al. (4), we did not The apparent disagreement be- cell membrane, both of which will be find prolongation of T2. tween our findings and previous carried to the bottom of the tube dun- Sequential spectrophotometnic spectrophotometnic studies that dem- ing centnifugation and not be present analysis of the solution shows a come- onstrated minimal methemoglobin in the supernatant to cause significant sponding increase in the concentra- might be resolved by comparison of absorption during UV spectroscopy. tion of methemoglobin during the pe- the differing methods of sample prep- Regardless of the exact mechanism nod of Ti shortening. Because oxy- anation. In the reported spectrophoto- of Ti shortening oven the first week and deoxyhemoglobin are known to metric analyses, particulate scattering following subarachnoid hemorrhage, have an insignificant effect on proton causes significant degradation, neces- it should be emphasized that the ap- relaxation enhancement in aqueous sitating centnifugation. With MR. pearance of acute subarachnoid hem- solutions (6), we presume that methe- such centnifugation is not necessary onnhage is much less striking on MR moglobin is responsible for both the and in fact results in a significant pro- images than on CT scans. For this nea- Ti shortening observed in the current longation of the measured Ti values. son, we fully agree with DeLaPaz et experiments and that seen clinically. Thus, it would appear that the pana- al. (4) that CT is more sensitive and Furthermore, we believe that methe- magnetic methemoglobin is either specific in this evaluation. CT should moglobin may be responsible for the held within the intact red blood cell thus be considered the examination of

102 #{149}Radiology July 1985 Figure 9

), nm) Spectrophotometry near 600 nm. Oxyhemoglobin is seen to have no absorption at 630 nm. During hypoxic storage of a solution of blood and CSF, a broad-based peak devel- ops at 630 nm. By comparison with known standards, the peak at 92 hours was shown to correspond to 90% methe- moglobin. Differences in peak heights reflect differ- ences in extinction coefficients.

choice for evaluating patients follow- 3. Sipponen JT, Sepponen RE, Sivula A. Nu- gens and related substances. Proc Nat Acad ing an acute intracerebral event that clear magnetic resonance (NMR) imaging Sci 1936; 2.2:159-163. of intracerebral hemorrhage in the acute 12. Koenig SH, Brown RD. Lindstrom TR. In- could include hemorrhage. However, and resolving phases. J Computer Assist teractions of solvent with the heme region patients evaluated in the subacute Technol 1983; 7:954-959. of methemoglobin and fluoro-methemog- stage (i week on more post ictus) may 4. DeLaPaz RL, New PFJ, Buonanno FS, et al. lobin. Biophys J 1981; 34:397-408. be better studied by MR imaging as, NMR imaging of intracranial hemorrhage. 13. Singer JR, Crooks LE. Some magnetic J Computer Assist Technol 1984; 8(4):599- studies of normal and leukemic blood. by that time, Ti shortening will oc- 607. Clin Eng 1978; 3:237-243. cur, allowing a specific diagnosis of 5. Wintrobe MM, Lee CR, Boggs DR. et al. 14. Roost KT, Pimstone NR, Diamond I, hemorrhage. #{149} Clinical hematology. Philadelphia: Lea & Schmid R. The formation of cerebrospinal Febiger, 1981; 88-102. fluid xanthochromia after subarachnoid Acknowledgment: We thank Steve Abram- 6. Bloembergen N, Purcell E, Pound. Relax- hemorrhage: enzymatic conversion of he- son, Ph.D., for preparation of the artificial CSF; ation effects in nuclear magnetic reso- moglobin to bilirubin by the arachnoid Leslee Watson, Jay Mericle, and Terry Andrues nance absorption. Phys Rev 1948; 73:679- and choroid plexus. Neurology 1972; for technical assistance; and Kaye Finley for 712. 22:973-977. manuscript preparation. 7. Brooks RA, Battocletti JH, Sances A, Larson 15. McQueen JD, Northrup BE, Leibrock LC. SJ, Bowman RL, Kudravcev V. Nuclear- Arachnoid clearance of red blood cells. Send correspondence and reprint requests to: magnetic relaxation in blood. IEEE Trans Neurol Neurosurg Psychiatr 1974; William G. Bradley, Jr., M.D., Ph.D., Hunting- Biomed Eng 1975; 1:12-18. 37:1316-1321. ton Medical Research Institutes, MR Imaging 8. Vermeulen M, van Gijn J, Blijenbeng BC. Laboratory, 10 Pico Street, Pasadena, CA 91105. Spectrophotometnic analysis of CSF after subarachnoid hemorrhage: limitations in the diagnosis of nebleeding. Neurology References 1983; 33:112-114. 9. Van den Meulen JP. 1. Bailes DR. Young IR, Thomas DJ, xanthochromia: an objective index. Neu- Stnaughan K, Bydder GM, Steiner RE. rology 1966; 16:170-178. NMR imaging of the brain using spin-echo 10. Pauling L, Coryell C. The magnetic prop- sequences. Clin Radiol 1982; 33:395-414. erties and structure of hemoglobin, oxyhe- 2. Bydder GM, Steiner RE, Young IR, et al. moglobmn and carbonmonoxyhemoglobin. Clinical NMR imaging of the brain: 140 Proc Nat Acad Sci 1936; 22:210-216. cases. AJR 1982; 139:215-236. 1 1. Pauling L, Coryell C. The magnetic prop- erties and structure of the hemochromo-

Volume 156 Number 1 Radiology #{149}103