点击显示 收起
1 From the Department of Radiology, Health Outcomes Policy and Economics (H.O.P.E.) Center, Brain Institute, Miami Children’s Hospital, 3100 SW 62nd Ave, Miami, FL 33155. Received November 1, 2001; revision requested November 9; revision received and accepted November 13.
Index terms: Brain, diffusion, 10.139 • Brain, MR, 10.12144
... the most important issues in biostatistics are not expressed with statistical procedures. The issues are inherently scientific, rather than purely statistical, and relate to the architectural design of the research, not to the numbers with which the data are cited and interpreted.
A. R. Feinstein (1)
New imaging technology is initially developed at the laboratory level. Subsequently, it is translated into the clinical arena as a new imaging application. New imaging applications, however, should not be used empirically but rather scientifically by proving their diagnostic value. Hence, normal, or normative, data are fundamental in understanding any new diagnostic test. The Gaussian (normal distribution curve) definition of normal is based on measurements taken from a disease-free population (2). The normal range is usually defined as the range of measurements extending two SDs above and below the mean (average); that is, the range that includes the central 95% of all measurements (2). Most diagnostic tests define normal in this way (2). Providing the range of values in healthy subjects is fundamental in understanding any diagnostic test.
Basic laboratory research has shown that diffusion, or Brownian motion, results from the thermal translational motion of molecules (3). It is a random process that involves small displacement distances: For example, unrestricted water molecules usually diffuse a distance of 20 microns in any given direction in 100 msec, or 60 microns in 1 second (3). These distances are comparable to cellular dimension, which raises the possibility that measurements of water diffusion might provide a means of probing cellular integrity and pathology (3). Extensive magnetic resonance (MR) research was done to make diffusion-weighted imaging available as a new imaging application (4). However, before diffusion-weighted imaging could be a valuable clinical diagnostic test, extensive normative data for all age groups needed to be determined.
The study by Forbes et al (5) reported in this issue of Radiology focuses on the normal changes in brain water diffusion during the 1st year of life by using diffusion-weighted imaging. Forbes et al (5) studied 40 subjects to determine two important parameters: (a) if water diffusion was age dependent in infants and (b) if water diffusion was different between gray and white matter in different regions of the cerebral hemispheres.
In this issue, Forbes et al (5) report two important results that are statistically significant. First, the apparent diffusion coefficient (ADC) in all regions is lower as age increases during the 1st year of life (P < .01). Second, ADC is significantly higher in white than in gray matter (P < .001). Other authors have reported brain water diffusion data in multiple age groups. A decade ago, Sakuma et al (6) reported in Radiology the results in six adult volunteers, eight premature neonates, and three infants aged 5–10 months. Three years later, the same group of scientists from Mie University School of Medicine in Japan, but with Nomura as first author, reported a larger series (7). The Mie University group (7) evaluated 48 subjects grouped as 10 newborns (aged <1 month), 13 infants (aged 1–10 months), nine children (aged 1–11 years), and 16 adults (aged 20–79 years). Nomura et al (7) concluded that changes of diffusional anisotropy in white matter are completed within 6 months after birth. Recently, Chun et al (8) reported a series with 11 healthy adult volunteers between the ages of 26 and 86 years. They found that the average water diffusion of the human brain is nearly constant throughout most of adulthood (8). Although some of these series have few subjects, we are finally putting together a road map of what is normal brain water diffusion according to age and brain tissue.
Diffusion-weighted imaging is a noninvasive imaging technique. Recruiting "normal" subjects, however, in this age group is a major challenge. First, children less than 12 months of age almost universally require sedation or anesthesia for acquisition of motionless diffusion-weighted images. Although the morbidity and mortality of sedation or anesthesia are low, they are not negligible (9). Second, obtaining institutional review board approval is difficult because of ethical issues. Forbes et al (5) have defined strict inclusion and exclusion criteria to try to obtain a representative "normal" population for this age group. In great part, the authors were successful. However, some of the subjects had final neurologic diagnoses of kernicterus (n = 2) and viremia (n = 2), which could potentially have affected water diffusion in white and gray matter.
A skewed distribution (non–normally distributed) can sometimes be mathematically transformed into an approximately normal distribution (10). Common transformations in medical science are the logarithmic and square root transformations (10). In this issue, Forbes et al (5) have elegantly displayed the reduction in ADC in different regions of the brain by using logarithmic transformation. The multiple anatomically related logarithmic transformations of the ADC for age are displayed as a middle continuous line representing the best fit logarithmic regression line for all data points. This best fit line is surrounded by 95% CIs, such that 95% of data points fall within these bounds (11,12). Providing the readers with the 95% CI is key in determining the approximately normal distribution of brain water diffusion in newborns and infants.
The study by Forbes et al (5), like most articles, has some items that should be taken into consideration with regard to results. Although the authors studied 40 subjects, the age groups are unevenly represented. Thirty-three subjects were between the ages of 0 and less than 6 months, whereas only seven subjects were between the ages of 6 and 12 months. Therefore, the results are more robust for the former and limited for the latter age group.
Transforming the complex physics and molecular biology laboratory concept of diffusion (Brownian) motion into a noninvasive imaging technique has been a monumental task. More than 35 years of extensive work was required (6). Measurements of anisotropic diffusion with a pulsed gradient were done in the early days of nuclear MR diffusion measurements (13,14). In 1976, Cleveland et al (15) studied water diffusion in a rat muscle model. In the middle and late 1980s, several research groups (16–18) started to perform spatial mapping of in vivo water diffusion with MR imaging techniques. However, diffusion-weighted images were deteriorated by eddy currents, especially when the diffusion-sensitive gradient was applied on phase-encoding and section-selection axes (6). It was not until the early 1990s that clinical use of diffusion-weighted imaging came to be with the advent of excellent gradient performance by means of self-shielded gradient coils (6,19,20).
Understanding normal brain water diffusion has served as a template for the evaluation of multiple neurologic diseases. Substantial research has been performed in adult acute stroke by using diffusion-weighted imaging (21–24). These studies have demonstrated that diffusion-weighted imaging is a better diagnostic method than conventional MR imaging in detecting early cerebral infarction (21–24). However, brain water diffusion changes in neonatal hypoxic-ischemic encephalopathy have not been as clear as in adult acute stroke (25,26). After global perinatal hypoperfusion, diffusion-weighted imaging has shown deep gray matter and perirolandic white matter lesions before conventional MR imaging (26). However, diffusion-weighted imaging may result in underestimation of the extent of injury due to variations in the compartmentalization of edema, selective vulnerability, and delayed cell death (26). Diffusion-weighted imaging differences between symmetric versus diffuse and focal versus multifocal lesions may reflect differences in pathophysiology or timing of the injury (26). Although we are starting to understand some of the pathophysiologic mechanisms of water diffusion involved in pediatric and adult neurologic disorders, a substantial amount of research still lies ahead.
Before scientists start to decipher the hieroglyphics of water diffusion changes in multiple neurologic diseases, they must master the "Rosetta stone" of normative data. Forbes et al (5) have raised the bar to a higher level. Future studies with larger numbers of subjects are clearly needed. Ideally, these studies should have normative data not only for diffusion-weighted imaging but also for other newer techniques such as diffusion tensor imaging and multivoxel MR spectroscopy. These studies should encompass not only the postnatal period but also the prenatal one. With the advent of prenatal MR imaging, having reliable fetal normative data will be fundamental in our understanding of disease processes that may present early in the in utero period, such as inborn errors of the metabolism. All of these studies should be supported by statistical power analysis to determine the robustness of the data presented. Once we have a solid understanding of the brain’s normal water diffusion properties, we will then be able to decipher abnormal water diffusion patterns in important pediatric neurologic disorders such as hypoxic-ischemic encephalopathy, childhood stroke, and inborn errors of metabolism.
An alternative to larger series is meta-analysis. Meta-analysis is a method used for integrating and combining the results of independent studies (27). Combining data from a variety of scientific studies can increase the power to detect effects, more precisely estimate the impact of these effects, or address a question not posed by the original investigator (28). Traditionally, meta-analysis has been used to analyze multiple randomized controlled trials in which results may be inconsistent or inconclusive. However, meta-analysis also can be applied to data from observational studies if they are of sufficient quality (27).
In conclusion, more articles such as that by Forbes et al (5) will provide additional information about the normal diffusion of water in the human brain. Although these studies are difficult to perform in children, they can and should be performed ethically by using well-defined inclusion and exclusion criteria. The work by Forbes et al (5), although it does not uncover all the answers about normal brain water diffusion, is clearly a step in the right direction.
REFERENCES