fetal brain magnetic resonance imaging response acutely to hypoxia-ischemia predicts postnatal...

Fetal Brain Magnetic Resonance ImagingResponse Acutely to Hypoxia-Ischemia

Predicts Postnatal OutcomeAlexander Drobyshevsky, PhD,1 Matthew Derrick, MD,1 P. V. Prasad, PhD,2 Xinhai Ji, MD,1 Ila Englof, BS,1

and Sidhartha Tan, MD1

Objective: Cerebral palsy (CP) is caused by either hypoxia-ischemia (H-I) or long-standing causative factors such as inflam-mation or genetics. Multiple pathophysiological events over time are thought to contribute eventually to cerebral palsy. Ourobjective was to examine whether the immediate response of the fetus to an acute H-I event determined the motor deficitsassociated with cerebral palsy.Methods: Serial diffusion-weighted imaging were performed on 79% gestation New Zealand white rabbits using a 3-Teslamagnetic resonance scanner during 40 minutes of uterine ischemia, 20 minutes of reperfusion, and at 4, 24, and 72 hours.Individual fetuses were tracked to near term, and the delivered kits were divided into hypertonic H-I (n � 18), nonhypertonicH-I (n � 9), stillbirth H-I (n � 4), and control groups (n � 16).Results: The hypertonia group had significantly less of a nadir in apparent diffusion coefficient (ADC) during H-I (71.6 �23.8% vs 84.5 � 9.3% baseline) and slower and incomplete recovery of ADC during reperfusion compared with the nonhy-pertonic group. All fetuses in the hypertonic and stillbirth groups had an ADC nadir of less than 0.83�m2/msec (70.3%decrease from baseline), whereas 94% of control animals had an ADC nadir greater than this value. The difference betweenoutcome groups was the largest at 4 hours reperfusion and persisted for 24 hours.Interpretation: Serial fetal brain scans indicate that the immediate response of a fetus to H-I is crucial to the development ofhypertonia. If the fetal brain can be scanned at the time of insult, ADC changes can predict which fetuses will have anunfavorable outcome.

Ann Neurol 2007;61:307–314

Despite improvements in perinatal practice during thepast several decades, the incidence of cerebral palsy(CP) has remained essentially unchanged.1 The causeof CP is thought to be multifactorial, includinghypoxia-ischemia (H-I), inflammation, genetic causes,and environmental risks.2 In approximately 70% to80% of cases, CP is antepartum in origin, and mostcases of antenatal H-I are considered idiopathic.3 Oneof the leading causes of CP from antenatal H-I is ab-ruptio placentae.4 In H-I, the combined effects of cel-lular energy failure, acidosis, glutamate release, intracel-lular calcium accumulation, lipid peroxidation, andnitric oxide neurotoxicity, among other factors, disruptessential processes in the cell, which ultimately lead tocell death.3 It is unclear what factors most contributeto the eventual neurological outcome: the inability ofthe brain to maintain homeostasis during the acute

phase of H-I or the inability to recover from the re-sultant cascade of adverse reactions, such as inflamma-tory and proapoptotic factors triggered by the initialH-I episode. If the immediate brain reaction to the ini-tial insult is an important determinant of the develop-ing motor and sensory deficits in CP, then neuroimag-ing could be useful in identifying infants at risk and,consequently, treatment strategies implemented.

Imaging studies have been conducted in the firstdays of life to diagnose H-I injury around the time ofdelivery,4 but currently, there is no way to diagnosefetal brain injury before delivery. Diffusion-weightedimaging (DWI) is a promising diagnostic imaging mo-dality because it has been shown to be a sensitive andearly indicator of H-I brain injury in both human ne-onates5 and neonatal animal models.6 An important is-sue that needs to be addressed to improve the diagnos-

From the Departments of 1Pediatrics and 2Radiology, EvanstonNorthwestern Healthcare and Northwestern University, Evanston,IL.

Received Oct 26, 2007, and in revised form Dec 6. Accepted forpublication Jan 16, 2007.

This article includes supplementary materials available via the Inter-net at http://www.interscience.wiley.com/jpages/0364-5134/supp-mat

Published online Apr 24, 2007 in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/ana.21095

Address correspondence to Dr Drobyshevsky, Department of Pedi-atrics, Evanston Northwestern Healthcare Research Institute, 2650Ridge Avenue, Evanston, IL 60201. E-mail: [email protected]

© 2007 American Neurological Association 307Published by Wiley-Liss, Inc., through Wiley Subscription Services

tic value of DWI is the relation between the timingand the magnitude of apparent diffusion coefficient(ADC, an absolute index derived from DWI) changesand the severity of brain injury. The ADC may be in-creased or decreased, depending on timing, severity,and site of injury in the early days of life.7 After theADC decline during acute cerebral H-I, the transientADC normalization to the pre-H-I level may occur,which, however, does not necessarily correspond to his-tological normalization8 and may obscure ongoing in-jury. Measured at a single time point, the ADC at 1 to2 hours after H-I underestimated final infarct volumein neonatal rats.9 However, measured serially, the pat-tern of ADC changes in a juvenile 3-week-old ratmodel of cerebral H-I was prognostic of the severity ofthe H-I insult by histological criteria.10 Furthermore,persistence of low ADC beyond H-I period was anominous sign of more massive cerebral necrosis.

Prediction of tissue fate based on DWI parametershas been investigated previously in focal H-I insult inhumans11 and animal models9; however, there hasbeen a paucity of studies correlating severity of globalH-I brain injury, assessed by quantitative DWI, andfunctional outcome. We hypothesized that the magni-tude of acute ADC reduction and persistence of thedecreased ADC during recovery after H-I are prognos-tic to the degree of fetal brain injury and postnatal out-come. We recently developed a clinically relevant fetalrabbit model12 that exhibits a CP phenotype in new-born kits after fetal H-I at preterm gestation. In thisstudy, H-I was induced at a later gestation age (79%)so that we could perform imaging with better qualityon larger fetuses. The postnatal outcome was similar tothe published model. This model allowed us to exam-ine the time course of ADC during H-I and recoveryperiod with serial imaging and to correlate it with post-natal neurological outcome of survivors.

Materials and MethodsHypoxia-Ischemia in the Magnetic ResonanceImaging ScannerThe surgical procedure has been described previously.12 Invivo, global H-I of fetuses was induced by sustained uterineischemia at 25 days gestation (79% term, E25) in timedpregnant New Zealand white rabbits (n � 5; Myrtle’s Rab-bits, Thompson Station, TN). This procedure models acuteplacental insufficiency at a premature gestation. Survivingkits exhibited a spectrum of sensory and motor deficits, in-cluding muscular hypertonia and characteristic posture, re-sembling human CP, as well as impaired locomotion, suck,swallow, and righting reflexes. In brief, dams were anesthe-tized with intravenous fentanyl (75�g/kg/hr) and droperidol(3.75mg/kg/hr), followed by spinal anesthesia using 0.75%bupivacaine. A balloon catheter was introduced into the leftfemoral artery and advanced into the descending aorta toabove the uterine and below the renal arteries. The catheter-ized animal was placed inside a magnetic resonance (MR)

scanner. Body core temperature was maintained at 37°Cwith a water blanket wrapped around the dam’s abdomenand connected to a temperature-controlled heating pump.After the dam was positioned in the magnet, the balloon wasinflated for 40 minutes, causing uterine ischemia and subse-quent fetal H-I. At the end of H-I, the balloon was deflated,resulting in uterine reperfusion. After the imaging session,the dam was removed from the magnet. Sterile field aroundthe femoral artery incision was maintained during the mag-netic resonance imaging (MRI) session. The catheter wasthen removed, femoral artery repaired, and the dam was al-lowed to recover. Six days after H-I, at E31, and 0.5 daybefore term, all fetuses were delivered by hysterotomy. De-livery by hysterotomy instead of natural birth was done toidentify the position of each fetus within the uterus. Twocontrol dams underwent imaging with the same surgical pro-cedure as experimental dams, but without inflation of theballoon.

GroupsSurviving kits underwent neurological assessment shortly af-ter delivery to determine presence of motor deficits and hy-pertonia as described previously.12 Hypertonia was deter-mined as an increased resistance to passive joint movement.The kits were assigned to the following groups: (1) controlkits (n � 16; born from nonischemic dams; all the kits werewithout hypertonia); (2) nonhypertonic kits (n � 9); (3) hy-pertonic kits (n � 18); and (4) stillborn kits (n � 4). Thekits born from H-I dams were divided into the latter threegroups. The latter two groups were also labeled as the groupswith an adverse outcome. Because the exact time of deathand presence of motor deficits could not be established onstillborn kits, the stillbirth group was also analyzed sepa-rately.

Magnetic Resonance ImagingFor each imaging session the dams were sedated with intra-venous infusion of fentanyl (75�g/kg/hr) and droperidol(3.75mg/kg/hr). MRI was performed in 3-Tesla (GE Health-care, Milwaukee, WI) clinical magnet using a quadrature ex-tremity coil. Single-shot fast spin-echo T2-weighted imageswere taken for anatomic reference in axial, coronal, and sag-ittal planes of the trunk of the rabbit dam, with 25 to 32axial slices covering all fetuses inside the dam. Slice thicknesswas 4mm, matrix 256 � 192, and field of view was 16cm.Anatomic scans were followed by diffusion-weighted echo-planar images (DWI) with b � 0 and 0.8msec/�m2, TR/TE � 7,400/70 milliseconds, 2 averages, and the same slicegeometry in 3 planes, as in reference anatomic images. Tomonitor the time course of H-I, we performed continuousDWI in axial plane with b � 0 and 0.8msec/�m2, 2 aver-ages, during 10 minutes of the baseline before H-I, 40 min-utes of H-I, and 20 minutes of reperfusion. Each DWI ac-quisition took around 2 minutes, depending on the numberof slices. Data from the baseline before H-I were combinedwith data from the control animals to obtain control ADCvalue at E25. Three more imaging sessions were performedat the next 4, 24, and 72 hours for each animal.

Surviving kits, or fixed-brain specimens of stillborn orpostnatal deaths, were imaged on 4.7-Tesla scanner (Bruker,

308 Annals of Neurology Vol 61 No 4 April 2007

Billerica, MA), using a surface 28mm coil and multislice T2-weighted RARE sequence (TE/TR 80/4,000 milliseconds) todetermine the presence of gross anatomic abnormalities.Trace-weighted diffusion images were obtained on the livesurviving kits with the following parameters: field of view2cm, matrix 128 � 64, 10 axial slices 1mm thick, 6 aver-ages, TR/TE 2,000/21 milliseconds, b � 0 and 1.09msec/�m2.

Tracking of Individual Fetus’s Fate and BrainResponse on Magnetic Resonance ImagingNoninvasive fetal imaging in small animals, carrying multiplefetuses, presented several challenges. The normal rabbituterus exhibited a slow, peristaltic motion with frequency 0.5to 3 contractions/min at E25. Respiratory and intestinal mo-tion of the dam, and individual fetal movement in the am-niotic sac, caused fetal head displacement up to 10mm in 30seconds, as seen on a time-lapsed image sequence of a singleslice, taken every 10 seconds (see Supplementary Movie 1).To monitor the time course of ADC, we found optimal re-sults using two averages to acquire a pair of images with andwithout diffusion weighting within 1 to 1.5 minutes. DuringH-I, uterine and fetal motion was noticeably reduced (seeSupplementary Movie 2).

In addition, a litter sometimes contained up to 12 fetuses(average, 9), and the frequent change in fetal position com-plicated tracking of individual fetuses within and across dif-ferent sessions. To identify each individual fetus, we used fastsingle-shot fast spin-echo scans that effectively minimizedimage blurring caused by fetal and uterine motion. The im-ages had sufficient T2 contrast to distinguish details of anindividual fetus’s gross anatomy, such as fetal brain, eyes, spi-nal cord, liver, and placenta (Fig 1A). We were able to trackthe position of each fetal brain within each uterine horn be-tween imaging sessions and subsequently link the MRI scansto the delivered kits after hysterotomy. In the described man-ner, evolution of an individual fetus’s brain response was fol-lowed from the initiation of H-I through fetal to the post-natal period, when neurological outcome was evaluated.

Data AnalysisADC maps were calculated off-line using in-house software,written in Matlab (Natick, MA). Position of each individualfetus within and between imaging sessions was determinedon orthogonal single-shot fast spin-echo anatomic images us-ing in-house software aiding in navigation between two-dimensional anatomic data sets. In the DWI series, fetalbrains were identified on images without diffusion weighting(see Fig 1A). Polygon ROIs were placed on the whole fetalbrains, avoiding ventricles, and transferred to the correspond-ing ADC maps (see Fig 1B). If the brain was covered byseveral slices, ROI was placed on the middle slice to mini-mize partial volume effect. To mitigate dependence of ADCmeasurements on the scanning plane position, we obtainedADC values as an average of ROIs for each fetal brain inthree orthogonal planes.

The position and shape of ROI placement was correctedmanually for each time point to account for fetal head move-ment. Most of the corrections for linear displacement werein the 0.5 to 2mm range. In this manner, the time course ofADC for each fetal brain was obtained during acute H-I;immediate reperfusion; at 4, 24, and 72 hours after H-I; andat E31 after birth.

The baseline was obtained by averaging five data points(taken in 10 minutes) immediately before induction of H-Ifor each fetus individually. Area under the ADC curve wascalculated as a sum of trapezoids between the baseline andthe ADC curve, as shown on Figure 2. Areas under the ADCcurves, relative to the baseline before H-I, were calculated forH-I phase (0–40 minutes of H-I), reperfusion phase (0–20minutes of reperfusion), and different intervals of recovery: 4to 24 hours and 24 to 72 hours after H-I.

Statistical AnalysisDifferences between groups were tested using one-way anal-ysis of variance (ANOVA), followed by post hoc multiplecomparison with Tukey’s test. The predictive value of ADC,measured at different time points, was tested with stepwiselinear regression using statistical package SPSS 8.0 (SPSS,

Fig 1. Fetal brain was outlined (magenta, arrow) on a B0 image without diffusion weighting (A) and the corresponding apparentdiffusion coefficient (ADC) map (B). Anatomic features aiding in location of a fetal head were the eye (e) and the nasal cavity (n).Color legend of ADC values is measured in �m2/msec.

Drobyshevsky et al: Diffusion Imaging of Fetal Hypoxia 309

Chicago, IL). Data are presented as means � standard devi-ation.

ResultsNeurological Outcome after Fetal Hypoxia-Ischemiaat 25 Days GestationIn the H-I group, 27 live and 4 stillborn kits weredelivered by hysterotomy from 4 dams. Fetal demisewas 13%. Of 27 live kits, 16 kits (60%) were severelyaffected and exhibited hypertonia of all limbs, inabilityto right themselves, postural, suck and swallow deficits,and were without locomotion. Two kits (7%) had mildhypertonia, and nine (33%) kits did not have hyperto-nia. A fifth H-I dam miscarried at E30; all the kits werefound to be dead and were excluded from analysis.

Acute Apparent Diffusion Coefficient Decrease IsLarger in Fetuses with Adverse OutcomeAfter the onset of H-I (Fig 3), ADC decline consistedof an initial 15- to 25-minute period of slow decrease(to 90–95% of the baseline), followed by a faster de-crease to the nadir point (65–80% of the baseline).After cessation of H-I and beginning of reperfusion,ADC began to recover and returned to 75 to 90% ofthe baseline after 30 to 40 minutes of reperfusion. Sim-ilar patterns of ADC decline and recovery to the base-line in H-I have been reported for neonatal rabbits13

and rats,14 although the magnitude of ADC decrease atthe nadir, 75.9 � 16.5%, was less than in these studies.

We examined whether the magnitude of ADCchange during acute H-I and immediate recovery of anindividual fetal brain was predictive of the neurologicaloutcome. Figure 3 shows the time course of ADCchanges relative to the ADC baseline before H-I for thefetuses, grouped by postnatal outcome. The magnitude

of ADC decrease in the fetuses from the hypertoniagroup was significantly larger than in nonhypertoniagroup (p � 0.05, ANOVA, post hoc comparison withTukey’s test), starting from 30 minutes of H-I and upto 15 minutes of reperfusion (see Fig 3). At the nadir,the ADC was significantly less in the hypertonia(71.6 � 23.8%) and the stillbirth (64.9 � 14.4%)groups than in the nonhypertonic group (84.5 �9.3%). The ADC values in the stillbirth group wereslightly lower (but not significantly) than in the hyper-tonia group.

The ADC in most fetuses began to return to thebaseline level immediately after the start of reperfusion.In 37.8% of fetuses, only in the groups with adverseoutcome (50% of hypertonic and 75% of stillbirthgroup), the ADC continued to decline or did not re-cover for 5 to 10 minutes after the onset of the reper-fusion. By 20 minutes of reperfusion, the ADC in fetalbrains in the nonhypertonic group returned to 94.2 �6.6% of ADC in the control group at the correspond-ing time point. In contrast, the recovery of ADC atthis time point was incomplete in the hypertonic group(85.6 � 18.2%). The stillbirth group also showed in-complete recovery (83.0 � 16.5%), suggesting thatsome of the animals may have survived the initial insult.

Looking at the distribution of values at the nadir ofthe ADC curve among the groups (Fig 4), there was asignificant difference between the hypertonic and thenonhypertonic groups. All fetuses in the hypertonicand stillbirth groups had an ADC nadir of less than0.83�m2/msec (70.3% decrease from baseline),

Fig 3. Time course of apparent diffusion coefficient (ADC),expressed as a percentage from the baseline ADC, in fetusesgrouped by postnatal outcome. * indicate time points withsignificant difference in ADC between hypertonic and nonhy-pertonic kits (p � 0.05, analysis of variance, post hoc com-parison with Tukey’s test). Note continuous decrease of ADCduring reperfusion in fetuses that died before birth. Opencircles denote control group; squares denote no hypertonia;triangles denote hypertonia; filled circles denote stillbirth.

Fig 2. Areas under the apparent diffusion coefficient (ADC)curves, relative to the baseline before hypoxia-ischemia (H-I),were calculated for H-I phase (0–40 minutes of H-I), reperfu-sion phase (0–20 minutes of reperfusion), and different inter-vals of recovery: 4 to 24 hours and 24 to72 hours after H-I.


whereas 94% of control animals had an ADC greaterthan this value. The data suggest that there may be athreshold value of ADC at the nadir that could be pre-dictive of eventual neurological outcome.

To further examine the difference in ADC changesbetween groups with different neurological outcomes,we calculated areas under the curve of ADC deviationfrom the baseline during H-I and reperfusion phases.The latter parameter was assumed to be indicative ofreperfusion injury. The magnitude of ADC decreasewas proportionally larger within groups with adversepostnatal outcome (hypertonic or stillborn) duringboth H-I and reperfusion phases (Fig 5), and the dif-ference in reperfusion phase between hypertonic andnonhypertonic kits was significant (p � 0.05,ANOVA, post hoc comparison with Tukey’s test). Thearea under the ADC curve during reperfusion periodsignificantly correlated with low ADC at 24 hours afterH-I (p � 0.05).

Time Course of Apparent Diffusion CoefficientChanges after Hypoxia-IschemiaFigure 6 presents longitudinal time course of the rabbitfetal brain ADC in control animals and those after H-Iat E25, grouped by postnatal neurological outcome.The ADC in control animals completely recovered by4 hours after sham surgery to the baseline and slowly,but significantly (p � 0.05, ANOVA) decreased withmaturation from E25 to postnatal E31. Starting from20 minutes after reperfusion, the time course of ADCin the nonhypertonic group was not significantly dif-ferent from control animals. The ADC of kits in thehypertonic group was significantly and persistently

lower than in control and nonhypertonic kits. TheADC in this group was significantly depressed even at4 and 24 hours after reperfusion. The difference inADC between outcome groups disappeared at 72 hoursafter H-I.

Fig 4. Box-and-whiskers plots of the apparent diffusion coeffi-cient (ADC) values at nadir of hypoxia-ischemia (H-I) andreperfusion time course. All fetuses with ADC at nadir less0.83�m2/msec had hypertonia suggesting possible thresholdvalue of ADC, predictive of eventual neurological outcome.Asterisks indicate significant difference in mean value fromthe control group (p � 0.05, analysis of variance).

Fig 6. Time course of perinatal development in brain appar-ent diffusion coefficient (ADC) in control rabbit kittens andthose after 40 minutes of sustained fetal hypoxia-ischemia(H-I) at 25 days gestation (E25), grouped by postnatal neuro-logical outcome at E31. Brain ADC in control kits progres-sively decreased with maturation. ADC in hypertonia group(squares) was persistently lower than in control (open circles)and nonhypertonic (gray circles) kits. The difference in out-come groups in ADC was most prominent at 4 hours afterH-I, decreased at 24 hours, and disappeared at 72 hours(*p � 0.05, analysis of variance, post hoc comparison withTukey’s test).

Fig 5. Area under the reperfusion phase of hypoxia-ischemiacurve were significantly larger in hypertonic (hatched bars)and stillborn (black bars) kits than in nonhypertonic kits(white bars) (mean � standard deviation). *p � 0.05, anal-ysis of variance, post hoc comparison with Tukey’s test.ADC � apparent diffusion coefficient.


Predictive Value of Apparent Diffusion CoefficientParametersTo find the best combination of fetal ADC time courseparameters that predicted postnatal outcome, we usedstepwise linear regression analysis. ADC values at allstudied time points and areas under ADC curve wereentered as predictors. The best fit (adjusted R2 � 0.96,ANOVA, p � 0.0001) was in a model with two pre-dictors: ADC nadir during H-I (standardized � ��0.664; p � 0.001) and area under the ADC curve at24 to 72 hours (standardized � � �0.361; p �0.007). Almost identical results were obtained whendata from stillborn kits were excluded from analysis.The best fit (adjusted R2 � 0.99, ANOVA, p �0.0001) was in the model with two predictors: ADCnadir during H-I (standardized � � �0.735; p �0.001) and area under the ADC curve at 24 to 72hours (standardized � � �0.304; p � 0.015).

Receiver operating characteristic curves for ADC na-dir and area under the ADC curve at 24 to 72 hoursare presented in Figure 7. Areas under the curve were0.978 for ADC nadir and 0.722 for the area 24 to 72hours. Linear combination of these parameters, esti-mated by the regression analysis, did not improve thearea under the curve just for nadir ADC only, possiblybecause the area of the nadir could not be improved.

Analyzing the results a little differently, all fetusesthat had substantial recovery of ADC to about 80% ofthe baseline within 20 minutes of observed reperfusionperiod and almost complete recovery by 4 hours had afavorable outcome without hypertonia. Most of the fe-tuses with adverse outcome (82%) did not recover by 4hours and had a lower ADC by 24 hours.

Fetal Ventriculomegaly Occurs between 24 and 72Hours after Hypoxia-IschemiaThis serial study also gave an opportunity to investigatethe onset of ventriculomegaly and other cystic lesions af-ter a known H-I insult. Most of the hypertonic kits(67%) showed lesions in cortex, midbrain (thalamus andpons), and striatum, with loss of tissue in hippocampus,cortex and striatum, and ventriculomegaly of varying de-gree. The findings on postnatal scans were similar towhat we had found after E22 H-I.15 No gross abnor-malities were detected in nonhypertonic and controlkits. There were no gross anatomic changes visible infetal brains on the scan after 24 hours after H-I, butseveral cases of ventriculomegaly (6/12 with ventriculo-megaly at P1) were detected on the scan at 72 hours(E28) after H-I (Fig 8), indicating that the onset of ven-triculomegaly or cystic lesions was as early as 24 to 72hours of H-I in the in utero rabbit fetus. There was alsoone dead fetus discovered at 72 hours.

DiscussionTo the best of our knowledge, this is the first study inan animal model that tracked the fate of each individ-ual fetus after H-I until birth. By obtaining the ADCtime course using serial DWI, we were able to associatethe individual fetal brain response to H-I and the neu-rological outcome. The fetuses with adverse outcome(hypertonia and stillbirths) had a significantly largerADC decrease during the acute episode of H-I and aslower and incomplete recovery of ADC comparedwith the nonhypertonic fetuses. The difference in ADCbetween outcome groups was the most prominent dur-ing acute and subacute phases after H-I and persistedfor the first 2 days after H-I.

Using a regression model, we sought the best com-bination of parameters that predicted hypertonia. Thebest combination was the ADC nadir during H-I andarea under the ADC curve at 24 to 72 hours. The datasuggested that both the magnitude of ADC responseduring H-I and slow recovery during the first 72 hourswere prognostic to neurological deficits. The receiveroperating characteristic curve areas were high for thenadir and could not be improved, suggesting that thenadir was the most important determinant of hyperto-nia. This is thus the first study demonstrating thatDWI changes in fetal brains may predict postnatalneurological outcome after global H-I.

The time course of MRI parameters during globalfetal H-I has not been described in the literature. Wefound that the pattern of ADC response to H-I in rab-bit fetuses was similar to what had been reported forneonatal rabbits13 and rats.14 Compared with adultstroke,16 it takes a proportionally longer time to reachmaximum decrease in ADC with younger animals.17

We observed a relatively slow and gradual ADC de-crease during H-I in our fetal model. At nadir, ADC

Fig 7. Receiver operating characteristic (ROC) curves for ap-parent diffusion coefficient (ADC) nadir and area under theADC curve at 24 to 72 hours. Areas under the curve were0.978 for ADC nadir and 0.722 for the area 24 to 72 hours.


was reduced by 75.9 � 16.5% and occurred only atthe end of 40 minutes of H-I. Such a slow ADC de-crease may be explained by the lower metabolic de-mands of the fetuses and young animals,18 and theglobal type of H-I in our fetal model, where oxygenreserves in the circulation mitigate initial hypoxia.

The time course of the acute ADC change in therabbit fetal model is similar to changes observed in hu-mans. There have been few serial studies of DWI inhumans. ADC continues to decline in the postnatal pe-riod after a suspicious episode before birth.5 Pseudo-normalization of depressed ADC occurs around 7thday after birth with subsequent increase of ADC aboveinitial values. In rabbits, there is a shorter ADC declineand a shorter period for ADC normalization (1–3days). Because these changes are also observed in ro-dents26 differences in these animals can be attributedto faster maturation and development compared withhumans. One limitation of human studies is that allthese studies involve postnatal MRI. The time of onset,duration, and extent of in utero injury cannot be dis-cerned with certainty because of limitations of currentbedside technology.

Another difference in the ADC response in ourmodel compared with stroke models is about the con-cept of “secondary energy failure.”19 A biphasic re-sponse of ADC is observed in animal models ofstroke,6,14,20 when a period of pseudonormalization isfollowed by a secondary ADC decrease about 24 hoursafter H-I. However, no secondary ADC decrease oc-curred in the nonhypertonic group in our study. Al-though in some of the fetuses from the hypertonicgroup, the ADC returned to 70 to 80% of the baselinewithin 20 minutes of the reperfusion and decreasedagain at 4 and 24 hours later, most of the hypertonicanimals showed no recovery of ADC at 4 and 24 hours

after H-I. Acquisition of MRI at additional time pointsmay be required to resolve possible, although unlikely,changes in the ADC between 4 and 24 hours. Thesestudies are expensive, unfortunately, and this study hadto compromise on performing serial MRI.

The persistent reduction of ADC after H-I in the hy-pertonia group within our study may indicate the failureof brain recovery after H-I. It has been reported thatthere are three patterns of ADC changes in the imma-ture rat model of cerebral H-I depending on the severityof the HI insult by histopathology.10 In a severe H-Iinsult, ADC decreases during the insult and does notrecover to normal after resuscitation. In a moderate H-Iinsult, ADC decreases during the insult, normalizes afterresuscitation, then decreases again 48 hours after the in-sult; this pattern was termed as a biphasic ADC reduc-tion pattern. In a mild H-I insult, ADC decreases dur-ing the insult, normalizes soon after resuscitation, andremains stable thereafter. Similarly, in an adult rat strokemodel, full recovery of ADC to preocclusion levels oc-curs if occlusion time is less than 30 minutes, but whenocclusion times exceeds 90 minutes, there is no recoveryin ADC until 2 to 7 days later.21,22 Delayed recoverycorrelates with severity of brain injury on histology.23

Thus, it has been shown on rodent models that brainswith the injury detected by histopathology also exhibitan abnormal pattern of ADC changes after H-I. Con-versely, we have demonstrated in this fetal H-I study atE25 that the pattern of ADC change is predictive of thepostnatal neurological outcome.

Interestingly, unlike in stroke models, the ADC of therabbit fetuses in the hypertonic and stillbirth groupscontinued to decline or did not start to recover for 5 to10 minutes after onset of reperfusion. The area underthe ADC curve during reperfusion also significantly cor-related with a low ADC at 24 hours after H-I. Delayed

Fig 8. There was one stillbirth after 40 minutes hypoxia-ischemia (H-I) at 25 days gestation (E25). The dead fetus was identifiedon anatomic scan only at E28, 72 hours after hypoxia (A), but not at 24 hours (E26). Arrow points to the dead fetal brain, dis-tinguished by loss of brain tissue contrast, compared with a live fetal brain (arrowhead). Ventriculomegaly can be observed in somefetuses at E28 (B). Arrow points to the enlarged lateral ventricles, filled with fluid (bright on T2-weighted image). No ventriculo-megaly was observed on adjacent brain (arrowhead).-


recovery of ADC in the early reperfusion phase after H-Icoincides with the burst of free radical production at thisperiod24 and, therefore, may be reflective of the oxidant-mediated injury during reperfusion. Fetal and neonatalbrains are known to be prone to free radical injury be-cause of their immature antioxidant system.25 Morestudies are necessary to determine whether the ADCchange during reperfusion/reoxygenation depends onproduction of free radicals. Furthermore, the predictivevalue of fetal MRI could be enhanced by observing thechanges in different brain regions, especially the vulner-able regions of the immature brain. However, currenttechnological aspects of the MRI system would need toadvance substantially before that becomes possible. Ourpreliminary data, obtained on imaging of a single iso-lated fetus after laparotomy, suggest that regional andage factors may significantly contribute to the diagnosisof perinatal brain injury.

In conclusion, the severity of fetal brain injury dur-ing hypoxia and immediate reoxygenation determineseventual postnatal neurological outcome. The patternof ADC changes during and after global fetal H-I is ofmore value than a single ADC determination. This hasprofound implications for the human situation. Inutero monitoring of the brain reaction to the H-I insultby serial DWI can potentially be applied to assess theonset, duration, and possibly extent of fetal brain in-jury in clinical practice. If the fetal brain can bescanned at the time of insult, ADC changes can predictwhich fetuses will have an unfavorable outcome.

This study was supported by the NIH (NINDS NS 43285, S.T.)and Department of Radiology, Evanston Northwestern Healthcare.

We thank R. Edelman for helpful suggestions and support.

References1. Perlman JM. Summary proceedings from the neurology group

on hypoxic-ischemic encephalopathy. Pediatrics 2006;117:S28–S33.

2. Nelson KB. Can we prevent cerebral palsy? N Engl J Med2003;349:1765–1769.

3. Volpe JJ. Neurology of the newborn. 4th ed. Philadelphia: WBSaunders, 2001.

4. Cowan F, Rutherford M, Groenendaal F, et al. Origin and tim-ing of brain lesions in term infants with neonatal encephalop-athy. Lancet 2003;361:736–742.

5. McKinstry RC, Miller JH, Snyder AZ, et al. A prospective, lon-gitudinal diffusion tensor imaging study of brain injury in new-borns. Neurology 2002;59:824–833.

6. Nedelcu J, Klein MA, Aguzzi A, et al. Biphasic edema afterhypoxic-ischemic brain injury in neonatal rats reflects early neu-ronal and late glial damage. Pediatr Res 1999;46:297–304.

7. Rutherford M, Counsell S, Allsop J, et al. Diffusion-weightedmagnetic resonance imaging in term perinatal brain injury: acomparison with site of lesion and time from birth. Pediatrics2004;114:1004–1014.

8. Miyasaka N, Kuroiwa T, Zhao FY, et al. Cerebral ischemichypoxia: discrepancy between apparent diffusion coefficientsand histologic changes in rats. Radiology 2000;215:199–204.

9. Wang Y, Cheung PT, Shen GX, et al. Hypoxic-ischemic braininjury in the neonatal rat model: relationship between lesionsize at early MR imaging and irreversible infarction. AJNRAm J Neuroradiol 2006;27:51–54.

10. Miyasaka N, Nagaoka T, Kuroiwa T, et al. Histopathologic cor-relates of temporal diffusion changes in a rat model of cerebralhypoxia/ischemia. AJNR Am J Neuroradiol 2000;21:60–66.

11. Nagesh V, Welch KM, Windham JP, et al. Time course ofADCw changes in ischemic stroke: beyond the human eye!Stroke 1998;29:1778–1782.

12. Derrick M, Luo NL, Bregman JC, et al. Preterm fetal hypoxia-ischemia causes hypertonia and motor deficits in the neonatalrabbit: a model for human cerebral palsy? J Neurosci 2004;24:24–34.

13. D’Arceuil HE, de Crespigny AJ, Rother J, et al. Diffusion andperfusion magnetic resonance imaging of the evolution of hy-poxic ischemic encephalopathy in the neonatal rabbit. J MagnReson Imaging 1998;8:820–828.

14. Qiao M, Malisza KL, Del Bigio MR, Tuor UI. Transienthypoxia-ischemia in rats: changes in diffusion-sensitive MR im-aging findings, extracellular space, and Na�-K�-adenosinetriphosphatase and cytochrome oxidase activity. Radiology2002;223:65–75.

15. Tan S, Drobyshevsky A, Jilling T, et al. Model of cerebral palsyin the perinatal rabbit. J Child Neurol 2005;20:972–979.

16. Moseley ME, Mintorovitch J, Cohen Y, et al. Early detection ofischemic injury: comparison of spectroscopy, diffusion-, T2-,and magnetic susceptibility-weighted MRI in cats. Acta Neuro-chir Suppl (Wien) 1990;51:207–209.

17. Vorisek I, Sykova E. Ischemia-induced changes in the extracel-lular space diffusion parameters, K�, and pH in the developingrat cortex and corpus callosum. J Cereb Blood Flow Metab1997;17:191–203.

18. Duffy TE, Kohle SJ, Vannucci RC. Carbohydrate and energymetabolism in perinatal rat brain: relation to survival in anoxia.J Neurochem 1975;24:271–276.

19. Thornton JS, Ordidge RJ, Penrice J, et al. Temporal and ana-tomical variations of brain water apparent diffusion coefficient inperinatal cerebral hypoxic-ischemic injury: relationships to cere-bral energy metabolism. Magn Reson Med 1998;39:920–927.

20. Thornton JS, Ordidge RJ, Penrice J, et al. Anisotropic waterdiffusion in white and gray matter of the neonatal piglet brainbefore and after transient hypoxia-ischaemia. Magn Reson Im-aging 1997;15:433–440.

21. Neumann-Haefelin T, Kastrup A, de Crespigny A, et al. SerialMRI after transient focal cerebral ischemia in rats: dynamics oftissue injury, blood-brain barrier damage, and edema formation.Stroke 2000;31:1965–1972.

22. Olah L, Wecker S, Hoehn M. Relation of apparent diffusioncoefficient changes and metabolic disturbances after 1 hour offocal cerebral ischemia and at different reperfusion phases inrats. J Cereb Blood Flow Metab 2001;21:430–439.

23. Wang L, Yushmanov VE, Liachenko SM, et al. Late reversal ofcerebral perfusion and water diffusion after transient focal isch-emia in rats. J Cereb Blood Flow Metab 2002;22:253–261.

24. Tan S, Zhou F, Nielsen VG, et al. Increased injury followingintermittent fetal hypoxia-reoxygenation is associated with in-creased free radical production in fetal rabbit brain. J Neuro-pathol Exp Neurol 1999;58:972–981.

25. Tan S, Zhou F, Nielsen VG, et al. Sustained hypoxia-ischemiaresults in reactive nitrogen and oxygen species production andinjury in the premature fetal rabbit brain. J Neuropathol ExpNeurol 1998;57:544–553.

26. Li F, Silva MD, Liu KF, et al. Secondary decline in apparentdiffusion coefficient and neurological outcomes after a short pe-riod of focal brain ischemia in rats. Ann Neurol 2000;48:236–244.


fetal brain magnetic resonance imaging response acutely to hypoxia-ischemia predicts postnatal...

Documents