MRI Score Ability to Detect Abnormalities in Mild Hypoxic-Ischemic Encephalopathy

Abstract

Background:

Magnetic resonance imaging (MRI) scores have been well validated in moderate/severe hypoxic-ischemic encephalopathy (HIE). Infants with mild HIE can have different patterns of injury, yet different scores have not been compared in this group of infants. Our objective was to compare the ability of three = MRI scores to detect abnormalities in infants with mild HIE, and infants with moderate/severe HIE were included for comparison.

Methods:

This is a single-center prospective cohort study of infants ≥36 weeks’ gestation with HIE born at a level III neonatal intensive care unit from 2017 to 2019. All infants with HIE underwent an MRI, but only infants with moderate/severe HIE underwent therapeutic hypothermia. At least two experienced MRI readers who were unaware of all clinical variables independently assigned three scores (Barkovich, NICHD NRN, and Weeke).

Results:

A total of 42 newborns with varying HIE severity underwent MRI on day five of life. In the overall cohort, abnormalities were reported in three (7%) infants using the Barkovich, in 10 (24%) using the NICHD NRN, and in 24 (57%) using the Weeke score. Agreement was excellent for each score: Barkovich score (k = 1.0), NICHD NRN (k = 0.92), and Weeke score (k = 0.9).

Conclusions:

Subtle injury due to mild HIE was detected with the highest frequency using the Weeke score, while inter-rater reliability was excellent for all three scores. These findings suggest that infants with mild HIE and subtle MRI abnormalities may benefit from detailed scoring systems, which is important for studies investigating the benefit of hypothermia in mild HIE.

Introduction

Neonatal hypoxic-ischemic encephalopathy (HIE) remains a major cause of neonatal morbidity and mortality with a global burden of 1 to 8 per 1000 live births. Infants with mild HIE were initially excluded from early neuroprotective trials given that they were previously thought to be at lower risk.¹ Recent studies have demonstrated that infants with mild HIE are also at risk for abnormal neurodevelopmental outcomes.^2–8 Brain magnetic resonance imaging (MRI) is the most studied biomarker to provide accurate prognostication for infants with HIE.^9–14 A key knowledge gap is that the multiple existing MRI scoring systems were designed and validated largely in infants with moderate to severe HIE.^12,15-17

This study aims to compare two established and more commonly used MRI scores with one more recent detailed MRI score. Each MRI score has limitations in reporting and detection based on the specific sequences and items the score includes in its evaluation.

The two most commonly used MRI scores in neonatal trials of moderate to severe HIE are the Barkovich score¹⁵ and the National Institute of Child Health and Human Development Neonatal Research Network (NICHD NRN) score.^16,18 Both these scores were developed and validated in cohorts of term infants with evidence of perinatal asphyxia. These scores group brain injury into specific recognizable patterns such as injury to the basal ganglia and thalamus (BGT) or basal ganglia/watershed (BG/W) pattern of injury. The NICHD NRN score adds the additional evaluation of focal cerebral lesions that are not accounted for by the Barkovich score. These scores are easy to apply given the grouping of patterns of injury. The Weeke score¹⁹ in comparison is a newer detailed score that includes the evaluation of diffusion-weighted imaging (DWI) and magnetic resonance spectroscopy (MRS) and where each specific component of the MRI is assessed separately. For example, the thalamus is assessed separately from the basal ganglia, and bilateral injury is graded higher in severity than unilateral injury for each component. The Weeke score overall is more time intensive and requires some degree of training.

The ability to provide accurate prognostication for infants with mild HIE is limited because we do not know how the existing MRI scores apply to this group of infants or which of the existing MRI scores might best detect abnormalities in infants with mild HIE.²⁰ This question is timely as new therapeutic interventions are being proposed to target infants with mild HIE. The aims of this prospective cohort study are specifically (1) to assess the frequency and severity of MRI abnormalities in infants with mild HIE using different MRI scoring systems and (2) to measure the inter-rater reliability of the three MRI scores.

Patients and Methods

Study design and patient eligibility

This prospective cohort study enrolled inborn infants greater than 36 0/7 weeks gestational age who were admitted to the neonatal intensive care unit (NICU) of the Parkland Health & Hospital System from June 2017 to May 2019. Pertinent demographic, clinical, laboratory, and neuroimaging data were obtained from the electronic health records of enrolled infants and their mothers. The study was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center and Parkland Health & Hospital System. Parental consent was obtained for voluntary participation. Infants were included in the study if they had a clinical diagnosis of HIE as determined by the presence of perinatal acidemia and by standardized modified Sarnat neurological examination by a certified examiner within the first six hours of life classifying the degree of encephalopathy as mild, moderate, or severe according to the NICHD scoring.^16,18

Perinatal acidemia was defined as an umbilical cord arterial pH ≤ 7.0 with base deficit ≥ −16 mmol/L or an umbilical cord arterial pH ≤ 7.15 with base deficit ≥ −10 mmol/L if in the presence of an acute perinatal event, an Apgar score ≤ 5 at 10 minutes, or a ventilation requirement for at least 10 minutes after birth. If the umbilical cord gas was not available, then a blood gas within the first hour of life was utilized. All infants had evidence of perinatal acidosis and encephalopathy on the neurological examination using the NICHD modified Sarnat score^16,18 in the first six hours of life.

Moderate and severe HIE were defined by published ACOG/AAP²¹ criteria, which include the presence of perinatal academia and encephalopathy affecting at least three of the six categories on the modified Sarnat examination. There is no standard or widely accepted definition for mild HIE. For this study, mild HIE was defined by previously published criteria from the prospective research in infants with mild encephalopathy (PRIME)^3,6 study, with presence of perinatal acidemia, and encephalopathy affecting ≥ 1 category on the modified Sarnat examination, but not meeting NICHD criteria for cooling.

Newborns identified with fetal acidemia and mild encephalopathy by the standardized NICHD neurological scoring system were classified as mild HIE and did not undergo therapeutic hypothermia. Infants with moderate or severe HIE underwent therapeutic hypothermia per standard protocols with whole-body cooling to 33.5°C for 72 hours. MRI of the brain was obtained in all study patients at day of life five to seven or before hospital discharge.

MR imaging protocol

Brain imaging was completed on a 3T MR Scanner (Siemens Skyra) housed within the Parkland Hospital NICU and dedicated only to neonates. The infants were “bundled and fed” to avoid sedation.²² The magnetic resonance examination, performed at 3T, consisted of harmonized protocols and sequences including T1- and T2-weighted sequences, DWI (b-1000), apparent diffusion coefficient (ADC) mapping, and single-voxel ¹H-MRS with one voxel in the left thalamus/basal ganglia and one voxel in the left parieto-occipital white matter. The ¹H-MRS was completed at an echo time (TE) of 30 ms.

MRI scoring systems

Barkovich score¹⁵

The specific score used is the “BG/W” score, which was created as a combination score to include both basal ganglia and watershed cortex patterns of injury. This score has been validated in multiple studies of infants with varying severity of HIE as a good predictor of neuromotor and cognitive outcomes. This score focuses on pattern recognition and includes evaluation of three MR sequences including T1-weighted, first-echo T2-weighted (typically referred to as proton density), and second echo T2-weighted sequences. The BG/W severity is graded on a scale from 0 to 4, with a score of 0 indicating normal; a score of 1 or 2 indicating abnormal signal in the basal ganglia/thalamus or cortex, respectively; and a score of 3 indicating abnormal signal in both the basal ganglia/thalamus and the cortex. A score of 4 indicates the maximum in terms of injury severity and correlates to global injury involving the entire cortex and basal nuclei, with basal nuclei involvement highlighted in Fig 1A and B.

FIGURE 1.

Magnetic resonance imaging (MRI) patterns of brain injury in HIE. MRI examples of abnormalities (marked by white arrows) as detected by the different scores are highlighted. (A and B) Extensive abnormal signal intensity in the bilateral basal ganglia and bilateral thalami on axial T1 and T2. This pattern indicates high severity of injury identified on all three scores. (C and D) Multiple unilateral punctate white matter lesions on axial diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC). Punctate white matter lesions are not specifically included in the Barkovich score. (E and F) Multiple bilateral punctate white matter lesions with involvement of the corpus callosum on axial DWI and ADC. The punctate white matter lesions are identified by both the NICHD NRN score and the Weeke score, whereas only the Weeke score includes additional severity scoring specific for injury to the corpus callosum. (G-I) Focal cerebellar hemorrhage on axial T1- and T2-weighted imaging and susceptibility-weighted imaging (SWI). Hemorrhage is only included by the Weeke score.

NICHD Neonatal Research Network score¹⁶

The NICHD NRN score is used in infants with moderate and severe HIE and has been correlated to neurodevelopmental outcomes out to age six to seven years.¹² This score was validated in randomized controlled trials of hypothermia and has a pragmatic escalation from a score of 0 indicating a normal MRI to a score of 3 signifying cerebral hemispheric devastation. Specifically, the identification of focal white matter injury, although not specifically included in the Barkovich score, is identified in the NICHD NRN by a score of 1a or 1b as highlighted in Fig 1C-F.

The Weeke score¹⁹

This score was developed in a validation and confirmation cohort of infants with moderate to severe HIE; this is a detailed item-based scoring system that ranges from a score of 0 (normal) to a maximum severity score of 57 and includes additional evaluation of ¹H-MRS. The Weeke score is comprehensive of the other scores but additionally allows for reporting of other types of injuries such as hemorrhage (Figs 1G-I) and focal or diffuse abnormal signal intensity that are not included in the prior two scores, and also includes the use of DWI sequence. The Weeke score includes 19 items of interest. Items are scored based on whether the injury to the region of interest is focal or extensive, and additionally on whether the injury is unilateral or bilateral. For example, a focal and unilateral injury to the thalamus would result in 2 points for that item, a focal and bilateral injury to the thalamus would result in 3 points for that item, and an extensive and bilateral injury to the thalamus would result in 4 points injury score for that item. The spectroscopy evaluation allows for 1 point if basal ganglia N-acetylaspartate is reduced and an additional 1 point if basal ganglia lactate is increased. This score is the most time intensive of the three scores to apply.

Statistical analysis of MRI readings

The primary outcome was to compare the frequency of injury detected on MRI for each score being used in infants with mild HIE, with infants with moderate/severe HIE being included for a comparison. Comparisons were completed using one-way ANOVA. The calculation of inter-rater agreement between readers for each score type was done using kappa (k) intraclass correlation. To test the reliability between readers for each of the three scores, the interrater agreement (kappa, k) was calculated between each score’s expert reader and the study investigator M.M. who underwent intensive training before applying each score. All readers were unaware of the clinical data and clinical outcomes, and each reader reviewed the MRIs in random order. Patient identifiers were removed from imaging before review. The examiners included N.R., an experienced pediatric neuroradiologist who reported both the Barkovich and NICHD NRN scores, as well as L.W./L.S.d.V. who reported the Weeke score. M.M. reported all three scores, and interrater agreement was calculated between M.M. and N.R. for the Barkovich and NICHD NRN scores and between M.M. and L.W./L.S.d.V. for the Weeke score. Adjudicated MRI scores in case of discrepant readings were used for the remaining statistical analyses. A linear regression with scatter plots was used to compare the three scores with the Pearson correlation coefficient included.

Results

The study population was a convenience group to capture all infants with mild HIE during the study period, and infants with moderate/severe HIE were included for comparison. A total of 44 patients with variable HIE severity were enrolled in this study, two of whom died before MRI completion. MRI was completed at median (interquartile range) day of life five^3,7 for the total cohort. Of the 42 infants enrolled during the study period, 16 (38%) infants had mild HIE, 24 (57%) infants had moderate HIE, and two (5%) infants had severe HIE. One infant with severe HIE subsequently died after completing the MRI.

The maternal and neonatal characteristics of this cohort are described in Supplemental Table 1. The infants with moderate/severe HIE had significantly lower Apgar scores than the infants with mild HIE (P < 0.05). The infants with moderate/severe HIE also had higher base deficits at birth on umbilical cord arterial gas, and there were three infant deaths. Placental chorioamnionitis was reported in 24 (57%) of the overall cohort, and maternal chorioamnionitis in 12 (29%) of the overall cohort, with no significant differences related to the severity of HIE.

Comparison of MRI scores

Of the total cohort described in Table 1, abnormalities were detected in three (7%) infants using the Barkovich score, increasing to 10 (24%) infants using the NICHD NRN score, and increasing further to 24 (57%) infants using the Weeke score. The Weeke score detected significantly more abnormalities in the overall cohort than either the Barkovich score (P < 0.01) or the NICHD NRN score (P < 0.01). The comparative scoring parameters for each of these three scores is provided in Table 2.

TABLE 1.

Comparison of MRI Score Abnormalities

Abnormal MRI	Score 1 Barkovich	Score 2 NICHD NRN	Score 3 Weeke
Overall Cohort N = 42	3 (7%)*	10 (24%)†	24 (57%)‡
Mild HIE N = 16	1 (6%)*	4 (25%)*,†	8 (50%)†
Moderate/Severe HIE N = 26	2 (8%)*	6 (23%)*	16 (62%)†

Abbreviations:

HIE = Hypoxic-ischemic encephalopathy

MRI = Magnetic resonance imaging

NICHD NRN = National Institute of Child Health and Human Development Neonatal Research Network

Data are presented as N (row percent).

Data across the row were analyzed by one-way ANOVA, P < 0.05.

Different superscripts (*, ^†, ^‡) across rows designate significant differences after one-way ANOVA.

TABLE 2.

Comparative Summary of the Parameters of Each Scoring System

Barkovich Score15	NICHD NRN Score16	Weeke Score19,*
0 Normal	0 Normal	Gray matter subscore 6 items Points possible: 23	BG, T, PLIC, brainstem, perirolandic cortex, †hippocampus
1 Abnormal signal in BG or T	1A Minimal ≤ 3 cerebral lesions without BG, T, ALIC, PLIC, or watershed pattern involvement	1H-MRS 2 items Points possible: 2	BG NAA, BG lactate
2 Abnormal signal in cortex	1B More extensive cerebral lesions without BGT, PLIC, ALIC, or watershed pattern involvement	White matter subscore	Cortex excluding perirolandic, WM abnormal SI, PMWL, hemorrhage, optic radiation, ‡corpus callosum
3 Abnormal signal in cortex and BG or T	2A Abnormal BGT, ALIC, PLIC, or watershed injury without additional cerebral lesions	6 items Points possible: 21
4 Abnormal signal in entire cortex and BGT	2B Abnormal BGT, ALIC, PLIC, or watershed injury with cerebral lesions	Cerebellum subscore 2 items Points possible: 8	Cerebellum abnormal SI or diffusion restriction, Cerebellar hemorrhage
	3 Cerebral hemispheric devastation	Additional subscore 3 items Points possible: 3	‡IVH,‡SDH,‡CVST

Abbreviations:

ALIC = Anterior limb of the internal capsule

BG = Basal ganglia

BGT = Basal ganglia and thalamus

CVST = Cerebral venous sinus thrombosis

¹H-MRS = Proton magnetic resonance spectroscopy

IVH, = Intraventricular hemorrhage

NAA = N-Acetylaspartate

PLIC = Posterior limb of the internal capsule

PWML = Punctate white matter lesions

SDH = Subdural hemorrhage

SI = Signal intensity

T = Thalamus

WM = White matter

^*Weeke score: Each item receives 1 point if focal injury (<50%) or 2 points if injury is extensive (≥50%), AND each item also receives 1 point if there is unilateral injury or 2 points if the injury is bilateral. There are 4 maximum points per individual item. There are 57 total points possible for this score.

^†This item has a maximum point possible of 3.

^‡This item has a maximum point possible of 1.

Inter-rater reliability and correlation of MRI scores

Among 42 infants, the inter-rater agreement, or k value, was excellent for each of the three scores: Barkovich score k 1.0 (1.0, 1.0) between M.M. and N.R., NICHD NRN score k 0.92 (0.75, 1.0) between M.M. and N.R., and Weeke score k 0.90 (0.77, 1.0) between M.M. and L.W./L.S.d.V. As demonstrated in Fig 2, the adjudicated scores showed that the NICHD NRN score correlated strongly to the Barkovich score (r = 0.77, P < 0.01). The adjudicated Weeke and the Barkovich score showed a moderate correlation (r = 0.68, P < 0.01). The adjudicated Weeke and the NICHD NRN score showed a moderate correlation (r = 0.65, P < 0.01).

FIGURE 2.

Correlation of magnetic resonance imaging (MRI) scores. The cutoff line highlights that the Weeke score allowed for reporting of additional abnormalities on MRI, which were not scored by either the Barkovich or the National Institute of Child Health and Human Development Neonatal Research Network (NICHD NRN) score.

Discussion

Key study findings in this prospective evaluation of infants with varying severities of HIE were the increased ability to systematically assess subtle MRI abnormalities associated with a clinical diagnosis of mild HIE when using detailed scoring systems, as well as the excellent inter-rater agreement of different scoring systems among expert examiners. The Weeke score, which is comprehensive of many MRI findings of potential importance, allows for the inclusion and systematic scoring of subtle abnormalities that are more likely to occur in mild HIE.

Historically, the limited neuroimaging data acquired on infants with mild HIE reflect the initial assumptions that these infants had a good prognosis. As new evidence has emerged challenging this, MRI studies in infants with mild HIE have demonstrated a range of abnormalities in 20% to 60% infants.^3,23 This wide range in MRI brain injury reflects variance in the timing of the MRI examination, in the scoring and protocols (e.g., with or without ¹H-MRS), and in the inclusion of white matter injury and/or hemorrhage. Limitations include the lack of a widely accepted definition of mild HIE within the first six hours of life. A recent retrospective cohort study by Walsh² included 48 infants with mild HIE who were treated with therapeutic hypothermia; 54% (26 of 48) were found to have an abnormal MRI using the Barkovich score. There were no differences in the rate of overall MRI abnormalities when comparing the infants with mild HIE with those with moderate or severe HIE. This study albeit important had different inclusion criteria as the examination was not done within six hours and did not offer a noncooled cohort for direct comparison with these study findings.

Another study by Gagne et al.²⁴ evaluated 215 infants with varying degrees of neonatal encephalopathy including both cooled and noncooled cohorts. Of the noncooled infants with mild HIE who obtained MRI scored by the Barkovich system, 40% had brain injury on MRI when compared with a lower incidence of brain injury in 31% of the infants with mild HIE who were cooled. In a recent prospective noncooled cohort, the PRIME study³ found that 17% (9/54) infants with mild HIE had brain injury on MRI using the NICHD NRN scoring system. In contrast, the magnetic resonance biomarkers in neonatal encephalopathy prospective cohort study⁹ found that 61% (29/47) infants with mild HIE had abnormal white matter on their brain MRI using the Rutherford scoring system.¹⁴ A recent review by Walsh et al.²⁵ summarized the recent studies evaluating MRI in mild neonatal encephalopathy and found from the available data that about 50% to 60% infants with mild neonatal encephalopathy would have some abnormalities on their MRI, with 20% to 40% of those infants demonstrating abnormalities consistent with injury from HIE. Injury to the BGT pattern is associated with acute, near-total asphyxia, and injury to the parasagittal cortex and white matter is associated with partial, prolonged asphyxia. In line with the prevailing hypothesis that mild HIE arises from less-severe hypoxic-ischemic insults, the most common pattern of injury observed on MRI in infants with mild HIE is white matter injury including punctate white matter lesions.^3,23,25

Considering the aforementioned studies and the limited number of prospective investigations evaluating MRI findings in mild HIE, the wide range of differences in the reported incidence of brain injury could be related to the small sample sizes, differences in scoring systems utilized, timing of evaluations, and variability in treatment as some were treated with hypothermia and some did not receive any cooling. The discrepancy seen in the wide range in the reported occurrence of abnormal MRIs in infants with mild HIE could also in part be explained by the different patterns of injury seen in mild HIE when compared with moderate or severe HIE. The patterns of injury in mild HIE can be subtle and are more likely to involve white matter injury, such as high white matter signal intensity or punctate white matter lesions, rather than injury to the deep central gray matter. Each MRI scoring system has a variable level of inclusion of these subtle abnormalities, and currently, many different MRI scoring systems are being used to describe infants with mild HIE, without clear consensus on which if any of the existing scoring systems is apt to describe the most common and significant abnormalities.

Our study highlights the importance of the specific score that is used in its ability to assess abnormalities in infants with mild HIE. It is important to note that the study was neither aiming nor powered to detect whether one score is superior or inferior to any other. The Weeke score simply allows for the inclusion of additional potentially important observations when addressing the mild end of the HIE spectrum for research or clinical management. In comparison to the Weeke Score, a major benefit of the Barkovich system is that it only includes injuries that are presumed to be associated with an asphyxial event. We have used in this study the clinically most widely used version, which is the simplified BG/W score, and this limits the interpretation of results related to the detailed Barkovich scoring. For example, the study by Walsh² reported 54% abnormalities using the complete Barkovich scoring, and in this report, 38% had abnormal BGT and watershed Barkovich scoring, and the remainder of the injuries were not encompassed in that scale, including intraventricular hemorrhage, cerebellar injury, etc.

This study confirms that using more detailed scoring systems with increased items reviewed allows for a higher reporting of abnormalities that may be clinically significant in infants with mild HIE, as opposed to the use of more simplified scores, which may be clinically in wide use but which do not allow for broad reporting of abnormalities. This question is timely and essential as we consider the application of new therapies to treat mild HIE. Using a more detailed score may be important for studies investigating the use of therapeutic hypothermia in mild HIE to be able to show differences in MRI abnormalities between infants who are cooled and not cooled. However, one limitation to more detailed scoring systems is that it is not clear if abnormalities such as subdural hemorrhage, which are highly unlikely to be clinically related to hypoxic-ischemic injury, will have relevance to neurodevelopmental outcomes or will adversely alter results if included in a clinical trial.

Few prior studies have addressed the calculation of inter-rater reliability of several existing MRI scores in HIE using expert examiners, and all were in infants with moderate and severe HIE,^19,26 thus limiting comparisons with the current study, which is focused on mild encephalopathy. We found no differences in the inter-rater reliability between the subgroups of mild and moderate/severe HIE, and overall there was good correlation between scores and examiners.

We recognize that these findings may not be broadly replicable in any clinical setting, as the scores were assigned by expert examiners with specific experience in each score, and the study entailed education and training among readers. Other limitations include the inability to report at present the association of MRI abnormalities with long-term outcomes in mild HIE, as these evaluations are still ongoing. In the current study none of the infants with mild HIE were treated with hypothermia, so we are unable to comment whether these findings would generalize to cooled infants with mild HIE.

The strengths of this study are the prospective design that included varying severities of HIE, and that it incorporated a systematic definition of what constitutes mild HIE within six hours of life using the PRIME criteria.³ In addition, the MRI was standardized in terms of protocol, sequences, and timing of the study, and the MRI readings were evaluated by expert examiners who were unaware of the clinical outcomes and the stage of HIE.

Conclusion

Subtle injury in mild HIE was reported with highest frequency using the Weeke score when compared with the Barkovich or NICHD NRN scores. This study additionally emphasizes the importance of following systematic and standardized MRI acquisition protocols in infants with mild HIE that enable the reporting of atypical punctate and diffuse white matter lesions and subtle injury. These findings suggest that infants with mild HIE may benefit from detailed scoring systems that feature the inclusion of subtle abnormalities. More detailed scoring systems broaden our view regarding which MRI findings might be modified in future intervention studies or might help to improve MRI abnormality interpretation and classification. The MRI score chosen for a trial should be carefully considered and selected to best ascertain treatment effect. However, more investigation is necessary to determine which score is an optimal biomarker for the mild end of the HIE spectrum in future planned trials. This is the first study to compare MRI scores in mild HIE, and larger cohorts will be needed to validate these findings. Future and ongoing directions will include the long-term developmental follow-up of infants with mild HIE to determine the prognostic value of these imaging scores.