Altered structural brain networks at term-equivalent age in preterm infants with grade 1 intraventricular hemorrhage

Advances in neonatal intensive care have markedly enhanced the survival rate and decreased the rates of intraventricular hemorrhage (IVH) in preterm infants [1, 2]. Although previous efforts have helped in the declining incidence of intraventricular hemorrhage and its neurological sequelae, 40% of very low birth-weight infants still show language, cognitive, and behavioral impairments [3‐5]. Research on the capacity of structural brain networks elucidated with diffusion magnetic resonance imaging (MRI) to predict functional impairments in very preterm infants has provided insight into the relationship between brain network topology and neurodevelopment. The diagnostic utility of brain imaging has prompted clinicians to explore the connectomes in infants and has considerably advanced the understanding of their neural circuits [6].

Preterm infants are at risk for structural disruption of brain connectivity due to the perinatal complications encountered during the fetal and neonatal period. Research has been conducted on the value of structural brain networks ascertained with diffusion MRI to predict the functional impairments and neurodevelopmental outcomes in very preterm children [7, 8]. Relative to full-term neonates, preterm neonates exhibit a decreased clustering coefficient, increased characteristic shortest path length, [6, 9] and decreased local and global efficiencies [10‐12]. Our previous study on preterm infants has revealed higher small-worldness in preterm babies when compared to full-term controls at a term-equivalent age [13].

IVH is an important cause of neurological and cognitive complications and can significantly impair brain development [14]. Several recent studies suggest that even germinal matrix hemorrhage of a low-grade IVH can cause neurodevelopmental impairment in the future [15, 16]. However, the mechanism of the organization of the images of abnormally integrated and segregated brain networks in preterm infants with grade 1 IVH during early development remains poorly understood. Furthermore, whether the reorganization of structural brain networks is independently affected by grade 1 IVH remains unclear. This study used advanced MRI analysis, including diffusion tensor imaging (DTI)-based connectomes, to investigate grade 1 IVH-induced alterations to the preterm structural brain network at near-term age.

The total population consisted of 55 preterm infants (26 males) with a mean GA at birth of 29.3 ± 4.1 (range 23–35) weeks and a mean age at scan of 37.2 ± 1.4 (range 36–41) weeks, and 24 term-born infants (18 males) with a mean GA at birth of 38.1 ± 1.0 (range 37–40) weeks and a mean age at scan of 39.8 ± 0.9 (range 37–41) weeks. Table 2 presents the demographic and clinical characteristics of both groups in the study. Thirteen of the 55 preterm infants (23.6%) were diagnosed with grade 1 IVH. Notably, 4 infants developed both grade 1 IVH, while 9 infants developed unilateral grade 1 IVH (Right 6 and Left 3). Compared to the full-term group, the preterm infants without grade 1 IVH underwent MRI significantly earlier (37.8 ± 1.7 vs 39.8 ± 0.9, p < 0.001) and featured a lower proportion of males (45.2% vs 75.1%, p = 0.028).

Relative to the preterm infants without grade 1 IVH, preterm infants with grade 1 IVH exhibited a low mean GA (27.4 ± 3.1 vs 30.1 ± 3.8, p = 0.002) and birth weight (909.2 ± 373.1 vs 1373.5 ± 471.6, p = 0.002). However, the age at the MRI scan and head circumference at age of MRI scan were not significantly different between preterm infants with and without grade 1 IVH. During the neonatal period, preterm infants with grade 1 IVH differed in that they exhibited a high rate of BPD (76.9% vs 50%, p = 0.116) and a longer duration of total parenteral nutrition days (44.42 ± 28.16 vs 30.61 ± 29.6, p = 0.156). This is associated with chronic intubation and poor nutritional support for proper brain growth; however, these differences were insignificant.

Table 3 shows the comparison of brain network parameters, including binary and weighted FA connectivity matrices in the following preterm sub-groups: extremely preterm (< 28 weeks GA, n = 21), very preterm (28–32 weeks GA, n = 18), and moderate to late preterm (33–36 weeks GA, n = 16). These exhibited no significant differences among preterm sub-groups in the statistical analyses, controlling for sex, BPD, and age at MRI scan as covariates in preterm infants. Table 4 shows the results of the comparison of brain network parameters, including binary and weighted FA connectivity matrices, among preterm infants with and without grade 1 IVH and full-term infants. Compared to the full-term group, preterm infants without grade 1 IVH underwent MRI significantly earlier (37.8 ± 1.7 vs 39.8 ± 0.9, p < 0.001) and featured a lower proportion of male infants (45.2% vs 75.1%, p = 0.028). The preterm neonates without grade 1 IVH exhibited slightly lower wFA clustering coefficients (0.088 ± 0.007 vs. 0.094 ± 0.007, p = 0.375), and wFA local efficiency (0.141 ± 0.010 vs 0.152 ± 0.014, p = 0.486) than full-term infants. However, the clustering, local efficiency, path length, global efficiency, and small-worldness in preterm infants without grade 1 IVH were comparable to those in full-term infants after adjusting for GA, sex, BPD, and age at scan. Remarkably, the preterm infants with grade 1 IVH exhibited higher clustering (0.093 ± 0.015 vs. 0.088 ± 0.007, p = 0.027) and local efficiency (0.151 ± 0.022 vs. 0.141 ± 0.010, p = 0.025), implying the potential for segregation. However, they exhibited longer path length (0.291 ± 0.035 vs. 0.275 ± 0.019, p = 0.020) and lower global efficiency (3.998 ± 0.473 vs. 4.212 ± 0.281, p = 0.048), indicating a decreased integration in the wFA connectivity matrix among preterm infants with grade 1 IVH than those without grade 1 IVH, after correcting for GA, sex, BPD, and age at scan. In contrast, we observed no significant differences in the binary network parameters with and without grade 1 IVH.

Developmental connectomics obtained from advanced neuroimaging techniques have emerged as an important tool for the characterization of the structural and functional brain connectivity during the early period of life in neonates. We explored potential alterations in structural network properties related to grade 1 IVH in preterm infants. The measures of structural segregation, such as modularity and local efficiency, were observed to increase, while global efficiency decreased in preterm infants with grade 1 IVH compared with those without grade 1 IVH, indicating a specialized reinforcement of network segregation at near-term age.

Previous studies have reported the effects of grade 1 IVH on the developmental white matter microstructure in preterm infants. Morita et al. [30] observed an association between grade 1 IVH in preterm infants and impaired cerebellar white matter from DTI images. Additionally, a study reported that the FA values in the corpus callosum, limbic pathway, and cerebellar tract are decreased in preterm infants with grade 1 IVH [31]. These infants with grade 1 IVH also had a poor development score at 24 months of corrected age. However, Payne et al. [32] exhibited that low grade intraventricular hemorrhage was not associated with adverse neurodevelopmental outcomes at 18 to 22 months in 1472 infants who were born before 27 weeks of GA. A previous study observed that except for grade 4 IVH, low grade IVH does not make a significant impact on the developmental outcome at 8 years of age in 270 infants who were born under 28 weeks of GA [33]. Despite many previous beliefs that it might not have any significant consequence, our study revealed that grade 1 IVH can alter the overall brain connectivity in grade 1 IVH.

The beginning of the development of the microstructure of white matter connections and the emergence and establishment of small worldness reportedly occurs throughout 30 PMA [6, 9, 10, 26]. When compared with preterm infants, the neural structural networks in full-term infants generally becomes more integrated with decreases in path length and increases in the global efficiency during the first phases of connectome development. Song et al. [34] suggested that increases in the global and local efficiencies indicate the development of both long distance and local connections in a weighted network matrix, which enable the brain to exchange information efficiently. This would further suggest that both integration and segregation processes increase with the age at which the scan is obtained, reflecting subtle changes in the neonatal brain. After early phases of the neonatal period, decreases in the clustering coefficient, modularity, and, hence, decreases in local efficiency as measured with diffusion MRI have been observed from the age of 6 months to 2 years in full-term infants [12, 35]. Evaluating the growth processes concomitant with the brain network shift from segregation to integration through the first 2 years of life, a few studies have reported that networks become more efficient in transferring information, indicating the improved integration in the full-term brain [11, 35]. We focused increasingly on normative preterm infants who were discharged successfully with no evidence of greater than grade 1 IVH. These results might reflect compensation-related network changes in preterm infants with grade 1 IVH, suggesting altered global efficiency and the relative preservation of segregated local efficiency in the structural brain networks during the rapid developmental period. In this study, the incidence (23.6%) of grade 1 IVH was relatively high in preterm infants less than 35 weeks, compared to that of a previous study by Bolisetty et al., in which 21.3% of preterm infants less than 28 weeks of gestational age were diagnosed with low grade IVH. These findings reflect that the preterm infants with grade 1 IVH in this study exhibited a low GA (27.4 ± 3.1 weeks) and birth weight (909.2 ± 373.1 g), involving higher incidence of IVH. Furthermore, we assume that the incidence of low grade IVH in this study was overestimated by exclusion of the infants with additional white matter injuries, cerebellar hemorrhage, and high grade IVH.

We explored the significance of the weighted networks and the differences between binary and FA-weighted networks. Discrepancies between unweighted or weighted connectome pipelines might be attributed to the differences in the speed at which networks change during dynamic developmental phases along with the abundant formation of new connections. To a lesser extent, differences in edge weights may account for the discrepancies. Compared to the binary approach, the FA-weighted approach could be more sensitive to the network organization of atypical early neural development in preterm infants than to the early development of the full-term brain [36]. Our observations by means of the FA-weighted matrices obtained from the diffusion MRI-based brain network indicated that the measures of structural segregation, such as clustering and local efficiency, augmented in preterm infants with grade 1 IVH; however, binary measures remained insignificant.

Ball et al. [27] demonstrated that, relative to term-born infants, preterm infants exhibit disrupted network capacities in both cortical-subcortical and cortico-cortical connections, possibly effected by the environmental stress induced by premature extra-uterine life on early network development during the neonatal period. After birth, the positive environmental motivation improves the integration capacity via the improvement of long-range connections. However, most preterm infants in the neonatal intensive care unit undergo an inevitable stressful environment or prematurity-related morbidity. Studies have shown the value of an optimal balance between integration and segregation in developmental connectome models after preterm birth with perinatal morbidity [8, 10]. With development, the balance between the integration and segregation of networks tends to be optimized with remarkable reorganization, resulting in maximized resilience against prematurity-associated pathology. Our previous study identified that the BPD is a critical factor in brain growth, which influences impaired white matter and cerebellar development in preterm infants at term-equivalent age [37]. We excluded the infants with intrauterine growth retardation and corrected for GA and BPD, which can negatively impact the neonatal brain growth and neurodevelopmental outcome, although not statistically different in the two groups. Despite the importance of the influence of prematurity-related morbidity on neurodevelopment, its correlation with grade 1 IVH has not been widely discussed. The prediction of the effect of grade 1 IVH on neurodevelopmental outcomes among preterm infants is currently among the clinical challenges of standard MRI studies, which are hampered by the limited capacity to evaluate neural reorganization processes and estimate differences in the architecture of structural brain networks. In the present study, after correcting for GA, sex, and BPD, the early development of grade 1 IVH led to a decrease of global efficiency and increase of local efficiency in structural brain networks compared to those of infants without grade 1 IVH. We identified aspects of network organization that mediates the effect of grade 1 IVH, which have revealed that such an organization reflects the increase of altered clustering at the expense of a global efficiency. Our findings provide insight to improve the understanding of how grade 1 IVH can subsequently alter later neurodevelopment using DTI-based connectomes. Accordingly, we aim to provide Bayley III scale data in the neurodevelopmental follow-up study to differentiate the preterm infants with abnormal brain development from relatively healthy preterm infants with normal brain development. This analysis may add novel findings to our results of grade 1 IVH-induced alterations to preterm brain network, implicating the involvement of cognitive and language development. The study is subject to a few limitations. First, our study contains a relatively small sample size when compared with studies performed with older age groups as the acquisition and analysis of neonatal MRI data are specialized and difficult tasks. Second, our diagnosis of grade 1 IVH was somewhat limited with the poor reliability, as all the cranial ultrasound scans were evaluated by a single pediatric radiologist. Furthermore, we could not include the repetitive longitudinal neuroimaging of the same cohort of preterm neonates and the implications for cognitive or language neurodevelopment.

Despite these limitations, knowledge of the neonatal human connectome may provide insight into preterm development and the influence of grade 1 IVH on brain development at near-term age. Further follow-up studies based on structural brain networks should examine whether long-term neurodevelopment in preterm infants with grade 1 IVH is mitigated by delayed or altered integration and segregation compared to that in full-term infants.