Establishing Cohort-Specific Reference Values for Term-Equivalent Age Brain Magnetic Resonance Imaging Abnormality Scores in a Korean Cohort

Article information

Neonatal Med. 2026;33(1):49-55

Publication date (electronic) : 2026 May 31

doi : https://doi.org/10.5385/nm.26001

Dabin Kim, MD¹, Hae Min Kang, MD¹, Ji Hye Yoon, MD¹, Jieun Jeong, MD¹, Gyeong Eun Yeom, MD¹, Baek Sup Shin, MD¹, Seh Hyun Kim, MD, PhD^,1,2, Ju Sun Heo, MD, PhD^1,2, Seung Han Shin, MD, PhD^1,2, Ee-Kyung Kim, MD, PhD^1,2, Han-Suk Kim, MD, PhD^1,2

¹Division of Neonatology, Department of Pediatrics, Seoul National University Children’s Hospital, Seoul, Korea

²Department of Pediatrics, Seoul National University College of Medicine, Seoul, Korea

Correspondence to: Seh Hyun Kim, MD, PhD Division of Neonatology, Department of Pediatrics, Seoul National University Children’s Hospital, 101 Daehak-ro, Jongno-gu, Seoul 03080, Korea Tel: +82-2-2072-0937 E-mail: neobrain@snu.ac.kr

Received 2026 March 11; Revised 2026 April 16; Accepted 2026 April 28.

Abstract

INTRODUCTION

Preterm infants are well documented to be at high risk for long-term cognitive, behavioral, and neurodevelopmental impairments (NDIs) that persist into later life [1,2]. Evidence from meta-analyses indicates that preterm-born children score significantly lower on cognitive tests and exhibit a two-fold risk of developing attention deficit and hyperactivity disorder compared to term-born infants [3]. Despite clinical advancements in neonatal survival, preterm infants remain at a high risk of adverse neurodevelopmental outcomes, ranging from cognitive delays to sensory and motor loss. This persistent risk underscores the critical importance of implementing vigilant and long-term surveillance, starting from the earliest stages of development [1]. As therapeutic interventions have become increasingly multifaceted, the early identification of infants at risk for neurodevelopmental delays has become paramount for optimizing clinical outcomes.

The global brain abnormality score by Kidokoro et al. [4], which uses term-equivalent age (TEA) brain magnetic resonance imaging (MRI) to predict NDI by the metric measurement of biparietal width (BPW), transcerebellar diameter (TCD), and deep gray matter area (DGMA) has been widely accepted. This scoring system exhibits excellent inter-rater reliability, with white matter and deep gray matter components showing significant correlations with the Bayley Scales of Infant and Toddler Development, 3rd Edition (Bayley-III) motor scores at 2 years of age [5].

However, population-specific standards are required. To address this, we aimed to propose new Korean-specific cutoff values for brain biometric measurements and examine their efficacy in predicting NDI among extremely preterm infants in Korea.

MATERIALS AND METHODS

RESULTS

1. Clinical characteristics and neurodevelopmental outcomes

A total of 99 extremely preterm infants were included in this study, of which 33 (33.3%) were diagnosed with NDI and 66 (66.7%) served as controls (Table 1). The median gestational age was 26.0 weeks (interquartile range [IQR], 24.6 to 26.4) in the NDI group and 26.3 weeks (IQR, 25.4 to 27.3) in the control group (P=0.08) (Table 2). Birth weight was 792.1±178.9 g in the NDI group and 857.1±168.3 g in the control group (P=0.08). The proportion of female infants was identical between the groups (42.4% in both groups). In the NDI group, eight infants (24.0%) had cerebral palsy, one (3.0%) had blindness, and no infants had hearing loss. The Bayley-III composite scores were significantly lower in the NDI group across all domains: cognitive (median 75.0 vs. 100.0, P<0.01), language (69.9±14.6 vs. 94.6±11.1, P<0.01), and motor (70.5±11.7 vs. 95.8±7.6, P<0.01). A developmental delay was observed in 72.7% of the infants for the cognitive and language domains, and in 87.9% of the infants for the motor domain in the NDI group.

Table 1.

Demographic and Clinical Characteristics of Study Population (n=99)

Table 2.

Brain Sizes according to the Neurodevelopmental Impairment Status

2. Brain size measurements and association with NDI

Brain MRI was performed at a median PMA of 37.7 weeks (IQR, 37.0 to 39.6) in the NDI group and 37.0 weeks (IQR, 36.3 to 37.4) in the control group (P<0.01). To account for differences in PMA at the time of MRI, the measured values (BPW, DGMA, and TCD) were corrected to a standard of 40 weeks’ PMA using the linear regression equations previously established by Kidokoro et al. [4]. All three brain size measurements were smaller in the NDI group compared to the control group: cBPW (67.4±6.4 mm vs. 69.8±4.5 mm, P=0.05), cTCD (median 43.5 mm vs. 46.6 mm, P=0.01), and cDGMA (median 10.6 cm² vs. 11.7 cm², P<0.01). Multivariate logistic regression analysis revealed that reduced cTCD was significantly associated with NDI (adjusted odds ratio [aOR], 0.96; 95% confidence interval [CI], 0.94 to 0.98), indicating that each 1-mm decrease in the cTCD was associated with a 4% increased risk of NDI. Similarly, reduced cDGMA was independently associated with NDI (aOR, 0.86; 95% CI, 0.79 to 0.93), suggesting that each 1-cm² decrease in the cDGMA was associated with a 14% increased risk of NDI (Table 3).

Table 3.

Univariate and Multivariate Logistic Regression Analyses for Factors Related to Neurodevelopmental Impairment

3. Comparison of the original and Korean-adapted scoring systems

The Korean-specific cutoff values for the KAS differed substantially from the original global brain abnormality score. For cBPW, the KAS cutoffs were >70 mm (score 0), 65–70 mm (score 1), 61–65 mm (score 2), and <61 mm (score 3) compared to the original global brain abnormality score cutoffs of >77, 72–77, 67–72, and <67 mm, respectively. For cTCD, the KAS cutoffs were >46 mm (score 0), 43–46 mm (score 1), 39–43 mm (score 2), and <39 mm (score 3) compared to the original global brain abnormality score cutoffs of >50, 47–50, 44–47, and <44 mm. For cDGMA, the KAS cutoffs were >11.6 cm² (score 0), 10.7–11.6 cm² (score 1), 9.9–10.7 cm² (score 2), and <9.9 cm² (score 3), compared to the original global brain abnormality score cutoffs of >9.5, 8.5–9.5, 7.5–8.5, and <7.5 cm² (Tables 4, 5). ROC curve analysis demonstrated that the KAS incorporating Korean-specific cutoff values showed slightly improved predictive performance for NDI compared to the original global brain abnormality score (Figure 1). Furthermore, the KAS explained a greater proportion of variance (r²) in the Bayley-III cognitive, language, and motor composite scores than the original Kidokoro scoring system (Figure 2).

Table 4.

Cutoff Values for cBPW, cTCD, cDGMA of Original Global Brain Abnormality Score

Table 5.

Cutoff Values for cBPW, cTCD, cDGMA of Korean-Adapted Scoring System

Figure 1.

Receiver operating characteristic (ROC) curves for predicting neurodevelopmental impairment (NDI): original global brain abnormality score vs the Korean-adapted scoring system. Abbreviation: AUC, area under the curve.

Figure 2.

Explained variance (r²) of the original global brain abnormality score and the Korean-adapted scoring system for Bayley-III composite scores.

DISCUSSION

This study demonstrated that, compared to the original global brain abnormality score normal values, the medians for cBPW and cTCD were shorter and cDGMA was larger in our population. The new scoring system with Korean-specific cutoffs achieved a higher area under the curve (AUC) than the original global brain abnormality score for predicting NDI. These results suggest that existing cutoff values may not adequately reflect the specific brain size distribution of Korean infants.

Although the global brain abnormality score proposed by Kidokoro et al. [4] has been widely used for brain growth assessment, it was primarily developed based on North American cohorts. The global brain abnormality scoring system evaluates both brain injury and impaired brain growth across four regions, including the cerebral white matter, cortical gray matter, deep gray matter, and cerebellum, using a combination of signal abnormalities on conventional TEA brain MRI and objective biometric measurements [4]. In a longitudinal cohort of very preterm children, the global brain abnormality score was independently associated with cognitive, motor, and behavioral outcomes at 2 years of age [6]. The association between the global brain abnormality score and motor outcomes persisted at both 2 and 10 years of age, supporting its role as a reliable predictor of long-term motor development in preterm infants [6]. Furthermore, Martini et al. [7] demonstrated that distinctive MRI abnormalities in the global brain abnormality scoring system, particularly white matter volume reduction, delayed myelination, and deep gray matter volume reduction, were independently associated with specific neurodevelopmental trajectories over the first 2 years of life in extremely low birth weight infants.

A previous study of Asian neonates demonstrated significant differences in the brain morphological shape and white matter microstructure among Chinese, Malay, and Indian populations, particularly in the cortical striatal thalamic circuit [8]. Despite the similar total brain volumes across the three groups, these findings suggest that population-specific anatomical variations are present at birth [8]. These findings suggest that brain biometric reference values derived from one population may not be universally applicable to other ethnic groups. Given that the global brain abnormality scoring system was originally developed using data from a North American cohort, the application of its cutoff values to Korean preterm infants may lead to misclassification of brain growth abnormalities and inaccurate prediction of neurodevelopmental outcomes.

Regional brain volumes at TEA are influenced by multiple perinatal factors including birth weight, sex, postnatal growth, and nutritional management, with the magnitude of these associations varying across different brain regions [9,10]. Among these factors, early nutritional support may be particularly relevant, as van Beek et al. [10] demonstrated that an optimized nutrition protocol with a higher protein and caloric intake in the first 28 days of life was associated with significantly larger brain volumes in extremely preterm infants, especially in the cortical gray matter and subcortical structures including the cerebellum and thalamus.

The observed enlargement of the cDGMA in our study compared to that in the original global brain abnormality score cohort may be attributed to several factors, including advances in overall NICU care and differences in nutritional practices between the two cohorts. Our center follows the European Society for Pediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN) and American Society for Parenteral and Enteral Nutrition (ASPEN) guidelines, prioritizing the initiation of enteral feeding on the day of birth combined with an intensified protocol of early lipid emulsion and high-dose amino acid supplementation. Previous studies have shown that optimized nutritional support, particularly enhanced protein and energy intake during early postnatal life, is associated with increased brain volume in preterm neonates, particularly in the cortical gray matter and subcortical structures, including the basal ganglia and thalami [11,12]. However, as direct nutritional data were not compared between the two cohorts, the contribution of nutritional practices to the observed differences in cDGMA remains speculative and warrants further investigation. Furthermore, as this study was conducted at a single center, the observed findings may partly reflect our specific clinical practices and institutional protocols rather than purely ethnic or physiological variations.

In contrast, the smaller cTCD observed in our study was likely multifactorial. The cerebellum undergoes rapid growth during the third trimester, making it particularly vulnerable to disruption by preterm birth and its associated complications [13]. Known risk factors for impaired cerebellar growth include intraventricular hemorrhage, postnatal corticosteroid exposure, prolonged mechanical ventilation, and systemic infection, all of which are frequently encountered in extremely preterm infants [9,13-17]. In addition, population-specific growth characteristics may contribute to the observed differences, as suggested by the smaller cTCD values in our cohort even after correction for PMA. Notably, there is a distinct lack of comparative literature on ethnic differences in cTCD among preterm infants. In this context, our findings of a comparatively smaller cTCD in the Korean population do not always indicate a developmental delay but rather provide pivotal baseline data that may redefine ‘normal’ cerebellar growth standards for Korean infants. This underscores the necessity of using population-specific tools such as the KAS to avoid clinical overdiagnosis and ensure more nuanced neurodevelopmental assessments. By applying Korea-specific cutoffs, the KAS showed improved diagnostic performance for predicting NDI compared to the conventional global brain abnormality score, with a slightly higher AUC (0.793 vs. 0.784) and greater explanatory power for Bayley-III composite scores. Although the difference was small, these preliminary findings are meaningful considering that the KAS was derived from a relatively small single-center cohort of 99 infants. Even a small improvement in predictive accuracy may have clinical significance when applied to a high-risk population of extremely preterm infants, where the early identification of NDI can guide timely interventions.

This study has several limitations. First, its retrospective design, single-center setting, and relatively small sample size of 99 infants limit the generalizability of the findings. Second, the inclusion of only infants who completed both TEA brain MRI and Bayley-III assessment at a corrected age of 18 to 24 months may have introduced a selection bias. Third, the timing of MRI differed significantly between the NDI and control groups, and although corrected values were used, this difference may have influenced the brain biometric measurements. Fourth, a direct comparison of nutritional strategies, neonatal morbidities, anthropometric indices, and other NICU management details between our cohort and the original global brain abnormality score cohort was not feasible, limiting the interpretation of the observed differences in brain biometric values. Finally, some results were limited by the lack of robust statistical significance for some variables, necessitating further validation through large-scale, multicenter Korean cohorts with long-term followup to establish the external validity and clinical applicability of the KAS.

Article information Continued