Main

Brain development in early life is thought to be more susceptible to injury than in the latter stages. Any insult at this stage can cause permanent injury resulting in disability, which is different from a mature cerebellum's injury that might be restored partially by rehabilitation. The previous literature of studies of the cerebellum clarified relationships between structural brain volume alterations and symptoms in various regions (1–4). Furthermore, based on a report by Krägeloh-Mann et al. (5), marked cerebellar atrophy in preterm babies was considered to be related to severe developmental delays. Thus, cerebellar atrophy was shown to be a bad prognostic factor of motor and cognitive outcomes in children.

In our daily pediatric neurology practice, we need a imaging tool to provide clinicians with the ability to trace developmental changes in the cerebellum of children and provide a reliable data reference base for healthy children. It would be helpful for interpreting information and communicating with the patients and their family.

Herein, we present a new method for 3D image reconstruction using a conventional MRI scanner connected to a workstation. After carefully setting up the parameters for MRI scanning, the reconstruction process was managed by a software program that uses data from the MRI scanner. With this technique, we were able to study cerebellar growth in children with volume measurements from reconstructed 3D images of the brain. Few studies outside the United States and Europe have systematically examined cerebellar development in childhood (6), and there are no detailed cerebellum development data for ethnic Chinese children. The purpose of this study was to establish such a database for further study.

METHODS

Patients.

We collected data on subjects for a 3-y period at the Department of Pediatrics, Wan-Fang Hospital, Taipei, Taiwan. Data were obtained from subjects undergoing ordinary health examinations or other reasons for which neurologist had suggested an MRI brain imaging study, except those who had any form of cerebellar dysfunction. Those suggested for further MRI study included paroxysmal disorders or functional disorders, such as headache, Tourette syndrome, autism, seizure, and so on. Brain MRI studies were applied to each of them followed by a 3D neuroimaging reconstruction processing study. If the subjects were later diagnosed as having cerebellar disease, then they were excluded from the study. In total, 50 cases were included during this period. All work performed in this study was approved by the ethics committee of Wan-Fang Hospital. Informed consent was obtained from all patients or a parent before study participation.

Technique for 3D reconstruction.

We performed this process of 3D reconstruction according to Schierlitz et al. (7) using 2D medical images and Avizo software (Version 5.1; Mercury Computer Systems, Inc., Germany). The method for 3D image reconstruction began by setting up the indicator for MRI scanning to a mode suitable for large datasets that can work well on personal computers (PCs) using the Avizo software. A GE 1.5 excite machine is used in this hospital. A subject's brain was scanned using transverse planes of ≤1.5 mm to obtain at least 80–120 slices. The conditions for MRI scanning were as follows: 1) patient position: supine, 2) coil: head, 3) T1W: 3D SPGR, 4) TR: 33, 5) TE: 3.0, 6) flip angle: 35°, 7) bandwidth: 15.63, 8) matrix: 256 × 192 zip 512, 9) field of vision (Fov): 22 cm, and 10) cover range: whole brain. Scans were collected for algorithmic reconstruction and transmitted from the MR unit database to an established workstation. The 3D Avizo software system was then used for image processing. The skull component of the brain was visually removed from the regions of interest using an arithmetic module to isolate the cerebellar hemisphere. We limited gray-scale values to 110–155 of 1024 (210) scales to approximate the boundary of the unmyelinated cerebellum. The software's “threshold” and “edge detection” tools allowed the unmyelinated cerebellum to be precisely delineated. Certain areas with fewer than 50 pixels (area <0.1 cm2) were removed to eliminate erroneously identified unmyelinated cerebellum. We then repeated the procedure for the myelinated cerebellum with gray-scale values in the range of 75–95 of 1024. Subsequently, we computed volumetric measurements of the unmyelinated cerebellum with the following formula: volume of cortex = (number of voxels within the cortex) × (volume per voxel), derived from the formula of Courchesne et al. (8), which can be modified to compute the myelinated cerebellum and entire cerebellar volumes. These procedures were done by a technician and a physician concurrently for three times to minimize the human operational errors and reconfirm the areas classified.

Reliability of regional volumetric measurements.

Interrater reliability was examined in a sample of 20 randomly selected brain scans analyzed by two raters (K.-H.W. and C.-Y.C.). The intraclass correlation for total cerebellum volumes ranged from 0.93 to 0.95 and those of myelinated and unmyelinated volumes from 0.90 to 0.92. The intrarater reliability was also examined in the same 20 scans analyzed by one of these raters (K.-H.W.), and the correlation for total cerebellum, myelinated, and unmyelinated volumes ranged from 0.94 to 0.97. The rater (K.-H.W.) then completed the analysis on the remaining scans.

Statistical analysis.

We conducted statistical analyses of the data with Excel 2003 and SPSS version 11.5. The total cerebellar volume, which was further differentiated into unmyelinated and myelinated cerebellum, was recorded by age and gender. The regression curves were individually graphed, and equations were determined. Our statistical probabilities and regression analyses used logarithmic functions.

RESULTS

During a 3-y period, we collected 50 cases that were enrolled in this study. The gender, age distribution, and unmyelinated, myelinated, and total cerebellar volume are given in Table 1. There are two preterm births in this study (in Taiwan, the definition of full-term babies is around 38–40 wk of postconceptual age): one was born at 36th wk with the birth weight of 3100 g, MRI was done at 9 mo and 27 d postnatal, hence the adjusted postnatal age for this infant was 9 mo; another one was born at 35th wk and had the MRI done at 2 mo and 25 d, hence the age was adjusted to 2 mo. In the research processes, all ages were adjusted, and all the parameters examined were within normal ranges, i.e. birth weights were between 2500 and 4000 g, birth heights were between 45 and 55 cm, head circumferences were between 34 and 38 cm, and all were within 3–97 percentile of normal range. The evaluations of whole cerebellar volume by radiologist and cerebellar function by neurologist also showed that there were no abnormal cases. Reconstructed 3D images of cerebellar growth of selected stages are shown in Figure 1. As Figure 2 shows, the regression curve of normal cerebellar growth at the ages of 0.2–12.7 y was y = 12.743Ln(x) + 91.562, r2 = 0.313, and the regression curve for the normal growth of unmyelinated cerebellum was y = 7.3662Ln(x) + 79.583, r2 = 0.1695, and normal growth of myelinated cerebellum was y = 5.3797Ln(x) + 11.974, r2 = 0.4166 (Fig. 2). In the formula, y is the volume and x the mean age. It is evident that the cerebellum volume substantially increased in the first 2 y of life. The majority of the cerebellar volume was accounted for by unmyelinated cerebellum, which increased by almost 130% in the first 2 y; myelinated cerebellar volume also obviously increased in the first 2 y. The ratio of the volumes of unmyelinated to myelinated cerebellum after 2 y old was ∼4:1. When we considered the gender factor in the analysis, the volume of cerebellar growth differed between females and males, with girls always lagging behind; it is not until ∼12 y of age that they had caught up to the volume of the male cerebellum (Fig. 3A). We also determined the male cerebellum growth equation to be y = 6.6346Ln(x) + 104.66, r2 = 0.0866 and the equation for females to be y = 16.625Ln(x) + 79.581, r2 = 0.6881 (Fig. 3A). Furthermore, as Figure 3B shows, it is evident that the volume of myelinated cerebellum increased with age, increasing more in females than in males, although the male myelinated cerebellar volume was greater than that of females in the infant stage. The male myelinated cerebellum growth equation was y = 4.4506Ln(x) + 13.613, r2 = 02485 and that of females was y = 6.0392Ln(x) + 10.881, r2 = 0.6485. However, the male unmyelinated cerebellar volume began to level off after 2 y old and was greater than that of females until around 11 y old, despite the rapidly increasing volume of females (Fig. 3C).

Table 1 The unmyelinated, myelinated, and total cerebellar volume by 50 aged patients
Full size table
Figure 1
figure 1

3D reconstructed neuroimages of the cerebellum at different ages are as follows: (A) anterior view, (B) vertical view of a 1-y-old child (volume: 79.12 cm3); (C) anterior view, (D) vertical view of a 2-y-old child (volume: 91.44 cm3); (E) anterior view, (F) vertical view of a 4-y-old child (volume: 92.27 cm3); and (G) anterior view, (H) vertical view of an 8-y-old child (volume: 102.52 cm3).

Full size image
Figure 2
figure 2

Growth curves of the total cerebellum, unmyelinated, and myelinated cerebellar volume (cm3) from 3 mo to 12.7 y of per age group; means of unmyelinated cerebellum, means of myelinated cerebellum, and means of cerebellum. The regression curve of normal cerebellar growth at the ages of 0.2–12.7 y was y = 12.743Ln(x) + 91.562, r2 = 0.313, and the regression curve for the normal growth of unmyelinated cerebellum was y = 7.3662Ln(x) + 79.583, r2 = 0.1695 and normal growth of myelinated cerebellum was y = 5.3797Ln(x) + 11.974, r2 = 0.4166. In the formula, y represents volume and x represents age.

Full size image
Figure 3
figure 3

Growth curves of the cerebellum of male and female children from 3 mo to 12.7 y of age. means of female and means of male. (A) Growth curves of the cerebellum of male was y = 6.6346Ln(x) + 104.66, r2 = 0.0866 and female was y = 16.625Ln(x) + 79.581, r2 = 0.6881. (B) Growth curves of the myelinated cerebellum of male was y = 4.4506Ln(x) + 13.613, r2 = 0.2485 and female was y = 6.0392Ln(x) + 10.881, r2 = 0.6485. (C) Growth curves of the unmyelinated cerebellum of male was y = 2.1892Ln(x) + 91.032, r2 = 0.0136 and female was y = 10.587Ln(x) + 68.696, r2 = 0.5731. In the formula, y represents volume and x represents age.

Full size image

DISCUSSION

The advancement of digital image processing with algorithms makes it possible to practically, effectively, and efficiently demonstrate 3D images. Our study used a high-resolution gray-scale (1024) method to define gray-scale levels of different parts of the brain on a delicately sliced MRI scan (slice thickness of 1.5 mm, with ∼80–120 slices in a normal brain). Then, we setup a level threshold and processed the images digitally. With this procedure, we can display each component of the brain using 3D images, including the cortex, myelinated brain tissue, ventricles, vessels, and pathological lesions. With delicately sliced MRI scans, this technology can become a new tool for more-precisely demonstrating both the normal anatomy and pathological lesions of the brain. This program can also be modified to perform volume arithmetic, which enables the evaluation of brain volume. Computation of volumetric measurements of the unmyelinated cerebellum used the following formula: volume of the cortex = (number of voxels within the cortex) × (volume per voxel). Modifications to this formula can allow computation of the volume of the myelinated cerebellum, ventricles, whole brain, and even brain lesions (8).

Brain development in early childhood is extremely dynamic and is thought to be critical in neurodevelopmental disorders. Quantitative MRI has significantly advanced our understanding of brain development during growth. Several large-scale longitudinal studies have been published (6,9). However, none of those large-scale studies focused on the cerebellum or covered the age from infancy to adolescent. Filling in this serious gap in our current knowledge was one of the motivations to perform this study.

Cerebellar growth in our study possessed a similar trend to that of the cerebrum, which was reported previously (10–16). We observed a dramatic increase in the size of the cerebellum in the first 2 y of life, in which overall unmyelinated cerebellar volumes rapidly increased to a lifetime maximum at around age 2 y (Fig. 3C), and myelinated cerebellum continually increased beyond childhood (Fig. 3B).

Giedd et al. (9) had reported that volume changes of cerebrum cortical gray matter were nonlinear and regionally specific, but grew with the similar pattern that they increased during preadolescence with a maximum size, followed by a decline during postadolescence. This phenomenon was not seen in our result of unmyelinated cerebellum growth. The volume of unmyelinated cerebellum increased as a sharply parabolic curve with age, increasing less in females than in males, and became leveling beyond ∼12 y old.

Another recent report using quantitative diffusion tensor imaging to evaluate cerebellar white matter development also demonstrated that myelination of the cerebellar white matter continued to sharply increase for up to 36 mo after birth, which was then followed by a gradual increase into childhood and adolescence (17). As myelination and synapse formation account for brain and white matter growth during the postnatal stage, these also support our finding of continual increases in the myelinated cerebellar volume.

Interestingly, when we analyzed the data differentiated by gender, female children seemed to lag behind males in development. We do not know whether this result is because of the limited study number or not. We believe that in the future, after more data become available, this condition will be elucidated.

Accumulating evidence reveals that cerebellar growth may underpin motor and cognitive development in infancy (18–20). Many cerebellar disorders are frequently reported to be related to structural brain volume alterations and/or morphological changes (1–4). Our quantitative information on cerebellar development may play a pivotal role in clarifying brain neurodevelopmental abnormalities. Also, this 3D neuroimaging method could serve clinicians as a way to visualize and to study brain development in patients, which might be helpful for physicians to communicate information to the patients and their family.