Evolution of T1 Relaxation, ADC, and Fractional Anisotropy during Early Brain Maturation: A Serial Imaging Study on Preterm Infants

BACKGROUND AND PURPOSE: The alteration of brain maturation in preterm infants contributes to neurodevelopmental disabilities during childhood. Serial imaging allows understanding of the mechanisms leading to dysmaturation in the preterm brain. The purpose of the present study was to provide reference quantitative MR imaging measures across time in preterm infants, by using ADC, fractional anisotropy, and T1 maps obtained by using the magnetization-prepared dual rapid acquisition of gradient echo technique. MATERIALS AND METHODS: We included preterm neonates born at <30 weeks of gestational age without major brain lesions on early cranial sonography and performed 3 MRIs (3T) from birth to term-equivalent age. Multiple measurements (ADC, fractional anisotropy, and T1 relaxation) were performed on each examination in 12 defined white and gray matter ROIs. RESULTS: We acquired 107 MRIs (35 early, 33 intermediary, and 39 at term-equivalent age) in 39 cerebral low-risk preterm infants. Measures of T1 relaxation time showed a gradual and significant decrease with time in a region- and hemispheric-specific manner. ADC values showed a similar decline with time, but with more variability than T1 relaxation. An increase of fractional anisotropy values was observed in WM regions and inversely a decrease in the cortex. CONCLUSIONS: The gradual change with time reflects the progressive maturation of the cerebral microstructure in white and gray matter. Our study provides reference trajectories from 25 to 40 weeks of gestation of T1 relaxation, ADC, and fractional anisotropy values in low-risk preterm infants. We speculate that deviation thereof might reflect disturbed cerebral maturation; the correlation of this disturbed maturation with neurodevelopmental outcome remains to be addressed.

MR imaging, specifically addressing the question of the maturation of the preterm brain.
Longitudinal imaging of the growing brain between 25 and 40 weeks of gestation allows assessing neuronal differentiation, gyral maturation, connecting fiber development, and early myelination. 6 To analyze these features of normal/abnormal maturation, efficient tools and reference values are still lacking. Several authors have described serial quantitative measures by using apparent diffusion coefficients and fractional anisotropy (FA) in various cohorts. [7][8][9][10][11] These sequences probe tissue microstructure and are used as markers of maturation, especially for axonal and dendritic organization and myelination. Recently, magnetizationprepared dual rapid acquisition of gradient echo (MP2RAGE) emerged as a new technique, which, by obtaining a purely T1weighted image, allows the extraction of whole-brain T1 tissue relaxation time maps to provide quantitative tissue characterization. 12 The descriptive properties of T1 relaxometry are of particular interest in the preterm population because they give structural information about tissue, such as water content and lipid and macromolecule composition, and draw a picture of the chronologic maturation of myelin. Moreover, there is a lack of quantitative T1 values for the assessment of brain development. 11,[13][14][15] In this serial imaging study in very preterm infants with cerebral low risk, we aimed to provide, for the first time, reference values for T1 relaxation time, and we hypothesized that their evolution is comparable with that of ADC and FA values, conferring greater and more precise information about tissue structure.

Patients
Neonates born before 30 weeks of gestation between February 2011 and May 2013 in our level III neonatology unit were considered for inclusion during the first days of life. Noninclusion criteria were the following: severe cardiorespiratory instability, intraventricular hemorrhage grade III and/or parenchymal hemorrhagic infarction on early sonography, severe congenital malformations, and genetic abnormalities. Patients who subsequently developed severe lesions on MR imaging, who died during the study, or who had abnormal neurologic assessment at term equivalent age (TEA) according to the Hammersmith Neonatal Neurologic Examination 16 were excluded from the final analysis. We thus defined the remaining patients as "cerebral low-risk." Neonatal variables were registered prospectively from the medical records.
Ethics approval was provided by the local committee, and written informed consent was obtained. Specific risks arising from imaging children younger than 2 years of age were assessed by the medical team and the institutional review board before the MR imaging examination.
Using T2, inversion recovery T1, and MP2RAGE, we calculated scores for severity at TEA according to Kidokoro et al, 18 including 6 items in the WM and 7 items in the GM and cerebellum. A global score (WM ϩ GM and cerebellum score) was calculated and classified as normal (0 -3), mild (4 -7), moderate (8 -11), and severe (Ն12). Two neonatologists experienced in reading MR imaging calculated the score. Intraventricular hemorrhages were graded according to Papile,19 and WM and cerebellar lesions were also described.
Twelve ROIs were identified with anatomic landmarks on 5 different sections for the WM (frontal, central, parietal, posterior limb of internal capsule [PLIC]; corpus callosum genu and splenium; and optic radiations) and the GM (frontal, perirolandic, and parietal cortices; thalamus; and lentiform nucleus). Freehand ROIs were drawn to maximize the size and avoid the risk of GM/WM contamination, as shown in On-line Fig 1. Each ROI was placed on the DTI sequence to measure ADC and FA and on the T1 map obtained from the MP2RAGE sequence to calculate the effective T1 relaxation time. 12

Neurodevelopmental Outcome
The patients were offered neurodevelopmental follow-up at 6 and 18 months of corrected age. A developmental pediatrician blinded to the neuroimaging findings performed a developmental assessment by using the Bayley Scales of Infant Development II, which entails a mental developmental index and a psychomotor devel-opmental index. The test mean is 100 Ϯ 15; a score Ͻ2 SDs means a severe delay.

Statistical Analyses
Statistical analyses were performed by using Matlab R2014b (MathWorks, Natick, Massachusetts) and STATA 13.0 (Stata-Corp, College Station, Texas). The different associations were analyzed with linear and quadratic regression. The correlation coefficient adjusted for the df (R adj 2 ) was used to identify the model with the best explanatory power. The Pearson linear correlation (R) was used to compare T1 relaxation time and ADC or FA values. The dispersion of ADC and T1 values was compared with a 1-sided t test applied on the normalized root mean square error of each marker. Demographic and neonatal variables were compared with the Student t test (continuous variables) and with the Fisher exact test (categoric variables). Statistical significance was defined as P Ͻ .05.

Description of the Population
Among 126 eligible patients, 51 preterm neonates were recruited. Reasons for not being included were parental refusal (n ϭ 26), early death (n ϭ 9), cardiorespiratory instability (n ϭ 13), early transfer to peripheral hospital (n ϭ 11), or absent recruiting person (n ϭ 16). Twelve patients were excluded from the final analysis because of severe lesions on brain MR imaging (MR imaging scores Ն8 or parenchymal hemorrhagic infarction), death, withdrawal of consent, or abnormal neurologic examination findings at TEA. We thus show the characteristics of the population based on 39 cerebral low-risk preterm neonates (Table).

Conventional MR Imaging and Scoring System
One hundred seven MR imaging examinations were performed; 35 early, 33 intermediary, and 39 at TEA. Thirty patients underwent 3 serial MRIs, 8 patients had 2, and 1 patient had only 1. The assessment of the image quality allowed considering 86% of the scans as good or with minimal motion artifacts.
Several mild cerebral lesions were diagnosed on the conventional sequences, including intraventricular hemorrhages grade I (n ϭ 5) and grade II (n ϭ 2), punctuate WM lesions (n ϭ 3), and punctuate cerebellar hemorrhages (n ϭ 4). Two patients had 2 types of lesions (intraventricular hemorrhage grade I and punctuate WM lesions).
For each ROI, we produced reference values stratified by gestational weeks, expressed as mean Ϯ SD for T1, ADC, and FA (On-line Table 2).
There was a significant and strong correlation between the T1 relaxation time and ADC values for all the ROIs and all MR im-ages at different gestational ages (Pearson correlation R 2 ϭ 0.616, P Ͻ .001). Furthermore, T1 values exhibited a significantly lower dispersion than ADC values (P ϭ 1.06 ϫ 10 Ϫ4 ). The correlation between T1 relaxation and FA (Pearson correlation R 2 ϭ Ϫ0.128) was negative and less significant.

DISCUSSION
The present study provides quantitative reference values for cerebral development, based on 107 MRIs acquired between 25 and 40 weeks in 39 very preterm infants. We used a newly developed sequence, MP2RAGE, which gives the T1 relaxation time, and compared it with MR imaging markers, ADC and FA. The selected cohort can be considered cerebral-low-risk, according to the exclusion criteria. Our findings were comparable with existing data (On-line Tables 3 and 4) issued from fetuses and preterm infants, detailed below.
Given fetal diffusion values and maturation curves obtained between 22 and 36 weeks, our findings of ADC and FA values were comparable with the ones presented by different groups, [20][21][22]  though subtle differences between fetuses at 37 weeks of gestation and preterm infants at TEA were reported. 23 No fetal data of T1 values are available.
While a multitude of data exist for preterm infants at TEA, 24,25 only a few studies have described the longitudinal evolution of quantitative brain MR imaging markers. In the late 1990s, Hüppi et al 9 reported changes of ADC and FA in the WM of preterm infants between early life and TEA. Their group showed differences in WM fiber organization and delay of development at TEA compared with term. Miller et al, 7 by using DTI, also showed serial differences in maturation in 23 infants with and without WM injury. Later, Nossin-Manor et al 11 assessed tissue organization longitudinally and were able to show a difference in maturation according to the different ROIs and the different techniques used, such as magnetization transfer, DTI, and T1 imaging. Recently Kersbergen et al 8 provided reference diffusivity values from scans obtained between 30 weeks and TEA. Compared with these studies, our ADC and FA values were similar to those in Nossin-Manor 11 and Partridge et al, 26 and FA values were slightly higher than those reported by other groups. 8,24 The relatively large heterogeneity of FA values in the literature is difficult to explain with certainty. However, it may involve several potential confounders: 1) b-value ranges from 600 to 1000 s/mm 2 , 27 2) slightly different tensor reconstruction strategies, 3) drawing and selection of the ROIs (this may actually be the main causative agent), and 4) some unsuspected systematic differences between the cohorts.
Concerning T1 relaxometry, only a few studies relate T1 values in infancy, 13 neonates 15 and premature infants, 11,14 albeit it provides reliable quantitative measures and high contrast images. We were able not only to measure T1 relaxometry serially in premature brains but also to show a strong correlation between ADC and T1 values, enhancing its validity toward clinical use. Moreover, we described a closer distribution of T1 values compared with ADC, in particular at TEA. Compared with existing data, 11,14 our findings were similar.
When performing serial imaging of preterm brain by using specific MR imaging markers, it is important to understand the different processes involved in brain maturation during the last trimester of gestation, such as neuronal differentiation, premyelination with water-content reduction, increase of lipid concentration, maturation of preoligodendrocytes, and finally the beginning of axonal myelination and development of connecting fibers. 5,28 Diffusion and T1 relaxation time are sensitive to changes in tissue water content and compartmentalization. Mean diffusivity reflects intra-and extracellular water mobility and provides information about cellular and axonal density and myelination. Moreover, T1 relaxation time also provides information about lipid concentration associated with myelin production, cholesterol, and macromolecules (galactocerebrosides) 11,13 and can, therefore, be considered as an optimal marker of brain maturation. FA represents a measure of tissue directionality sensitive to the degree of axonal alignment, fiber diameter, and consecutive early processes of premyelination. 9 To draw brain maturational trajectories in very preterm infants, we used the above-mentioned 3 imaging biomarkers. In the WM fiber tracts (PLIC, optic radiation, and corona radiata), the linear decline of ADC and T1 reflects reduction in water content, fiber packaging, and early processes of myelination, especially for the PLIC from 36 weeks onward. In these structures, the steep slope of FA represents the progressive development of unidirectional (PLIC) or multidirectional (corona radiata) fibers. The splenium and genu of the corpus callosum consist of tightly packed fibers with a high degree of coherent parallel organization, which myelinate only at 3 and 5 months after term, respectively. 11,29 This feature accounts for little change with time for ADC and T1 values and high absolute FA values. In the frontal and parietal WM, we observed a shouldered curve on ADC and T1 maps that could be explained by the inclusion of the subplate zone that peaks between 29 to 32 gestational weeks and then gradually disappears. The subplate has a high water content, 5,30,31 is particularly voluminous in the frontal WM, 21,32 and accounts for elevated ADC and T1 values.
In the basal ganglia and thalamus, the ADC and T1 values showed a gradual decrease due to fast neuronal densification with ongoing myelination, as described starting around 26 weeks. 29 In FA, these subcortical GM structures exhibited little change with time because of the low directionality of neuronal and glial content. In the frontal and parietal cortex, the evolution of ADC and T1 values showed a shouldered curve with maximum values around 35 weeks, possibly related to programmed cell death and additional neuropil before 35 weeks 33,34 and higher neuronal attenuation afterward. The perirolandic cortex seemed to mature faster than other cortical regions, and this accelerated maturation has been described in areas with primary function, such as the sensorimotor cortex. 34,35 The observed decline of the FA is attributed to the preferential reduction in the radial component of water diffusivity, reflecting the loss of the radial glial cells and the extension of dendrites of pyramidal cells. [32][33][34][35] The present study has a number of limitations. We assumed that our cohort was at cerebral low-risk, given their clinical evolution and the absence of major cerebral lesions. Neurodevelopmental outcome at 6 and 18 months showed that no patient had cerebral palsy, blindness, or hearing loss, and the distribution of developmental scores was typical for this population of preterm infants. Furthermore, because patients with moderate or severe brain lesions were scarce, we could not compare their values with those obtained from the selected low-risk patients. Finally, comparison with healthy control fetuses and term neonates was not available.
In this study, we propose reference values of T1 relaxometry, which could represent a precise and complementary tool to investigate brain development with time. We speculate that deviation of the described trajectories might reflect disturbed maturation, and this could add valuable information for the diagnosis of encephalopathy of prematurity. 4,5 Kinney and Volpe 28 described "altered developmental trajectories, combined with acquired insults and reparative phenomena" to characterize this entity, in which all the structures detailed above are affected. Oligodendrocyte differentiation, axonal growth, subplate organization, and maturation of the subcortical structures represent features that are likely to be affected by prematurity.

CONCLUSIONS
Our study evaluated, longitudinally and serially, the cerebral developmental trajectories of a cohort of cerebral low-risk preterm infants born at fewer than 30 weeks of gestation. On the successive MP2RAGE and DTI sequences, we observed a gradual decline with time of ADC and T1 relaxation time and changes of FA in the described 12 ROIs, reflecting the specific and sequential maturational changes occurring during development in the WM and GM microstructures. T1 maps confer high contrast, are easy to analyze, and thus appear as a promising complementary biomarker of cerebral maturation. We provide reference values for T1 relaxation, ADC, and FA, and we speculate that deviation thereof might reflect disturbed cerebral maturation; the correlation of this disturbed maturation with neurodevelopmental outcome remains to be addressed.