| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Division of Emergency Radiology, Massachusetts General Hospital, Harvard
Medical School, 55 Fruit St., FND-210, Boston, MA 02114.
2 Division of Trauma and Burn Surgery, Massachusetts General Hospital, Harvard
Medical School, Boston, MA 02114.
3 Harvard Medical School, Boston, MA 02114.
4 Department of Pediatrics, Massachusetts General Hospital, Harvard Medical
School, Boston, MA 02114.
5 Department of Emergency Medicine, Massachusetts General Hospital, Harvard
Medical School, Boston, MA 02114.
Received March 21, 2003;
accepted after revision May 22, 2003.
Address correspondence to T. Ptak
(tptak{at}partners.org).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Fifteen patients were randomly selected from the trauma surgery service. Thirty control patients were randomly selected from the emergency department. All patients were subjected to MRI evaluation, including a diffusion-weighted sequence. Data extracted from the record of each subject included Glasgow Coma Scale, revised trauma score, Abbreviated Injury Scale, initial head CT results, patient disposition, length of hospital stay, and length of stay in intensive care unit. Region of interest measurements were made in fractional anisotropy maps in each of 12 white matter regions. Univariate statistics and a two-tailed t test were performed on the raw fractional anisotropy data. Data were then dichotomized using thresholds from univariate statistics. A C-FAST score was devised from the dichotomized data. Logistic regression analyses were performed among the C-FAST, outcome, and predictor data.
RESULTS. Good correlation was noted between the C-FAST and death, hospital stay greater than 10 days, and intensive care unit stay greater than 5 days. Correlation with discharge to rehabilitation facility was good when adjusted for age and sex. Glasgow Coma Scale, revised trauma score, and Abbreviated Injury Scale show good correlation as predictors of a critical C-FAST.
CONCLUSION. The C-FAST is a promising index derived from MRI diffusion fractional anisotropy measurements that shows successful correlation with outcome and predictor variables. A larger investigation is needed to verify the validity and stability of the correlations.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In addition to the diffusion tensor image and the associated apparent diffusion coefficient map, measurements acquired in the diffusion-weighted acquisition have been used to glean additional detail about the directional bulk fluid flow of water in the brain. Several quantitative indexes have been derived from these data to represent bulk water diffusion direction. In this investigation, we use the fractional anisotropy index, a unit derived from the diffusion tensor data that was thought to be the most representative of three-dimensional directional diffusivity [5].
This unique information has been used previously to infer characteristics that may be attributable to disorders, particularly diseases of white matter [6]. Fractional anisotropy has also been applied to evaluation of posttraumatic diffuse axonal injury [7], but its clinical usefulness has yet to be described.
We began from an assumption that focal injury to an axon fiber will be represented by malfunction along the entire length of the axon. In this way, we further assumed that the fractional anisotropy measurement will reflect the aggregate response of "sick" axon bundles, and that the exact site of injury need not necessarily be included in the region of interest (ROI) measurement. Using this physiologic imaging model as a basis, we propose a score devised from summation of dichotomized fractional anisotropy measurements from six representative white matter fiber regions of the brain. We hypothesize that this score will reflect aggregate abnormal axon membrane function in each region and subsequent malfunctioning nerve fibers. We believe that this score will reflect unapparent white matter injury in each major brain region and will correlate with clinical indicators such as the Glasgow Coma Scale (GCS), patient outcome, and hospital stay.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Subject Population
A retrospective evaluation of patients admitted to the trauma surgery
service through our emergency department during the period June 1,
2001–October 31, 2001, yielded 15 patients (eight men, seven women; mean
age, 54 years; range, 21–83 years) who underwent both unenhanced head CT
and cerebral MRI, including the tensor diffusion sequence, within 7 days of
the incident. Thirty atraumatic control patients (13 men, 17 women; mean age,
49 years; range, 22–78 years) were randomly selected from the records of
the emergency department during the same period. These control patients were
evaluated with the same imaging sequences as the case subject group.
Presenting complaints in the control group included headache (n =
17), dizziness (n = 8), change in mental status (n = 4), and
changes in vision (n = 1). All final reports and images in the
control group were evaluated and confirmed as radiographically normal. Control
patients reported to be radiographically positive for any diagnosis were
excluded from the study to avoid confounding as a result of pathologic white
matter changes that were already radiographically apparent. Excluded cases
included white matter microangiopathic change, lacunar stroke, lobar stroke,
and nonspecific white matter T2 and fluid-attenuated inversion recovery
(FLAIR) hyperintensities.
Posttraumatic MRI
All imaging was performed on a 1.5-T Signa Horizon LX System version 8.3
MRI platform (General Electric Medical Systems, Milwaukee, WI). All subjects
had at least one scan within 7 days of the trauma (on the day of trauma
[n = 32], 1 day after [n = 4], 2 days after [n =
2], 3 days after [n = 2], 4 days after [n = 2], 5 days after
[n = 2], and 6 days after [n = 1]). Diffusion tensor imaging
is performed as a matter of routine on all patients presenting to our MRI
suite for cerebral imaging. The standard MRI sequence after trauma included a
minimum of sagittal T1-weighted, axial T2-weighted, FLAIR, and gradient-echo
susceptibility sequences in addition to an axial echoplanar diffusion tensor
acquisition (TR/TE, 10,000/47.8; b value, 1,000 sec x
mm–1). Routine diffusion images include a diffusion-weighted
image, apparent diffusion coefficient map, exponential diffusion-weighted
image (represents the pixel-by-pixel ratio of diffusion tensor imaging to low
b-value image), low b-value image, and fractional anisotropy map.
Fractional Anisotropy ROI Measurement
Anisotropy measurements were taken from each of six major white matter
volumes. Regions were measured in each of the centra semiovale. Anatomically,
this axial slice is a consistent representation of commissural (corpus
callosal) and association (fascicular) fibers. The centrum semiovale is also a
white matter volume prone to tensile, compressive, and rotational stresses.
The corpus callosum ROI was measured at the genu anteriorly and in three
samplings (left, right, and central) from the splenium. These dense fiber
bundles are also prone to severe central stresses. A representative sampling
of the deep white matter was taken from the internal capsule. The anterior and
posterior limbs and the genu were sampled separately on each side, for a total
of 12 measurements. The extreme capsule was not selected because of its thin
configuration and close approximation to the local gray matter structures.
There was also a high likelihood of incorporating the gray matter claustrum in
the measurement. Because of the uncertainty of an imprecise measurement as a
result of histologic heterogeneity and small size, the external and extreme
capsule measurements were omitted. Injury to this white matter region will
probably be reflected in the surrounding white matter tracts that have been
already included. Measurements were made on a PACS (picture archiving and
communication system) workstation (IMPAX D3000, version 4.1, Agfa HealthCare,
Greenville, SC) using the ROI utility. Elliptic regions were placed on
fractional anisotropy map areas representing each of the white matter tracts
(Fig. 1A,
1B,
1C). Care was taken not to
include visible gray matter or cerebrospinal fluid regions. Regions were
centered in an area that could be consistently recognized in the white matter
structure for optimal precision in ROI placement from one patient to the next.
For instance, the slice immediately cephalad to the lateral ventricle roof was
used to place the centrum semiovale ROI. The ROI was sized as large as safety
allowed without including gray matter or cerebrospinal fluid. To foster a
reasonable degree of precision between measurements, the area and
configuration of the ROI were recorded for each measured area and were kept as
constant as possible from one patient to the next. Placement in white matter
was confirmed by comparing the architecture of the white matter on the
fractional anisotropy map with the corresponding low b-value map to help avoid
inclusion of gray matter and cerebrospinal fluid. Mean pixel values were
recorded from the ROI measurement in each of the 12 regions.
|
|
|
Clinical Outcome and Historical Data
Clinical assessment and outcome data for each of the patients were
extracted from a database maintained by the trauma surgery service. All case
patients were followed for a period of 6 months to ensure a complete medical
record through discharge. Patient medical records were evaluated when the
records from the database were incomplete or unclear. Data fields selected for
outcome evaluation included age, sex, Glasgow Coma Scale, final patient
disposition (died, discharged to rehabilitation facility, or discharged to
home), vital signs, revised trauma score, Abbreviated Injury Scale, head CT
result, discharge date and length of hospital stay, and days in the intensive
care unit.
Statistical Analysis
All calculations were performed using statistics software (version 7.0,
Stata, College Station, TX).
Raw Anisotropy ROI Data, Two-Tailed t Test, and Sample Power
Calculations
Patient data were stratified as trauma or nontrauma, and univariate
statistics were performed for each of the white matter regions. Mean and
variance data were used in routine sample size estimate calculations based on
the observed differences in mean for each of the sites. Statistical power (1
– ß) was held at 0.80 and the alpha value, at 0.05, as customary.
Two-tailed t tests were carried out as the statistical test of the
mean, and p values for each white matter region were tabulated along
with the univariate statistics and sample size estimates.
C-FAST Calculation
ROIs were dichotomized using the mean and SD data for each ROI as described
previously. The cutoff for dichotomizing into normal and abnormal was chosen
as 1 SD below the mean for each of the measured regions in the control
population. This value represents abnormality at the 17th percentile in the
measurement distribution for each region.
Complex white matter structures such as the internal capsule and splenium of corpus callosum were measured as three ROIs, typically center, left or right, and anterior or posterior, depending on the long-axis orientation of the structure. The entire white matter region was considered abnormal if any one of the three regions was abnormal. This reflects forces imparted to, and resultant injury of, white matter as a single volumetric unit as opposed to individual white matter fiber tracts (i.e., internal capsule as opposed to anterior limb vs genu vs posterior limb).
Processing the raw data as described results in a scheme representing six physical white matter regions: corpus callosum genu, corpus callosum splenium, left internal capsule, right internal capsule, left centrum semiovale, and right centrum semiovale. Each region had been previously dichotomized into normal equals 0 and abnormal equals 1 as stated previously. The final patient C-FAST was calculated as the numeric sum of all six dichotomized regions, resulting in a whole integer score between 0 and 6.
For the purposes of predicting a "bad" C-FAST from initial clinical predictors (e.g., positive head CT, Glasgow Coma Scale, revised trauma score, or Abbreviated Injury Scale), the C-FAST was categorized about a critical value. Analysis of the score data showed a statistically significant correlation between score and most outcomes at a C-FAST of 3 or higher. Below a score of 3, the correlation and statistical significance declined dramatically. The C-FAST was then dichotomized, with an abnormal score being considered 3 or higher. This dichotomization makes it possible to analyze the data so that one can predict from clinical indexes the likelihood that the C-FAST will be critical before the diffusion MRI is obtained.
Categorization of Demographic and Clinical Outcome Data
Scores were categorized as critical or noncritical using common clinical
cutoff values for undesirable clinical outcome. A Glasgow Coma Scale of less
than 8, an Abbreviated Injury Scale greater than 20, and a revised trauma
score of less than 6 were considered critical. Age (sample range, 21–83
years) was categorized into 10-year increments (because of small sample size)
from age 20 (e.g., 20–29, 30–39, 40–49, 50–59,
60–69, 70–79, and 80–89). Death and discharge to
rehabilitation facility were dichotomized to 1 if true and to 0 if false. A
hospital stay of more than 10 days and an intensive care unit stay of more
than 5 days were considered critical.
Uni- and Multivariate Logistic Correlation
Univariate logistic regression models were formed with each of the clinical
outcomes (death, rehabilitation facility discharge, hospital stay > 10
days, and intensive care unit stay > 5 days) with C-FAST as the predictor
variable. Odds ratios calculated in these models reflect the incremental
likelihood of observing the clinical outcome for each increment in score.
Goodness-of-fit was assessed using Pearson's correlation coefficient
(r) of the model.
Multivariate logistic regression models were calculated in this manner but with categorized age and sex introduced as covariates. This stratification reflects the age- and sex-adjusted odds ratios as described.
Univariate regression was also performed using as the outcome variable a
critical C-FAST and using as the predictors positive findings on head CT,
Glasgow Coma Scale, revised trauma score, and Abbreviated Injury Scale. In
this model, the odds ratio reflects the likelihood of observing a critical
C-FAST (
3) when given a critical predictor value.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Score Distribution
Using the threshold criteria as described, the distribution of scores in
our population was 26 of 30 control patients with a score of less than 3 and
12 of 15 case patients with a score of 3 or greater. Each score level was
tested for criticality by dichotomizing the score into "good" and
"bad" at a threshold score. Outcome correlation was performed
using each outcome variable and the critical value. Six iterations of this
activity revealed that a C-FAST of 3 showed the earliest statistically
significant result with high sensitivity and specificity. Hence, the
separation between good and bad outcomes occurred most consistently at a score
of 3. In this way, a C-FAST of 3 was taken to indicate the clinical threshold
for a poor outcome in our subject population.
Regression Models
Correlation of score with all but one of the outcome variables and with all
predictor variables showed good results at a p value of 0.05 or less.
Pearson's correlation coefficients in all models were low, with the highest
value being seen in the univariate regression with death outcome. The
discharge to rehabilitation facility variable was not significant in the
univariate model, but when adjusted for age and sex, became statistically
significant (Table 2). Aside
from the univariate death outcome model, regressions showed overall good
sensitivity but excellent specificity. When age and sex were adjusted in the
multivariate model, sensitivity increased and specificity remained largely
unchanged (Table 2).
|
Predictor Effectiveness
Effectiveness of the C-FAST was compared with the effectiveness of other
predictors in the model by constructing the receiver operating characteristic
(ROC) curve for predicting death using C-FAST and each of the clinical
predictors. The family of ROCs produced is shown in
Figure 2. The effectiveness of
C-FAST in combination with each of the other clinical predictors was also
tested by first creating the interaction term for each combination of C-FAST
with Glasgow Coma Scale, revised trauma score, and Abbreviated Injury Scale. A
grand all-inclusive combination term was also calculated using C-FAST with all
predictors. The family of ROCs for the interaction terms is shown in
Figure 3. The comparative areas
under the ROC (i.e., effectiveness) for predicting death in each of these
iterations are tested for difference from the C-FAST alone using a chi-square
analysis. The ROC area, SE, and p value for comparison of each of the
ROC curves with reference to the C-FAST curve are shown in
Table 3.
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Previous works using diffusion tensor imaging have concentrated on the site of focal injury in the axon [4]. Finding the injury site with precision has proven difficult. More recently, work has generalized to a wider area measurement of white matter in the hopes of capturing the focally injured area [7]. In our work, we generalize further, assuming that a focal injury affects function throughout the whole of the axon, and that inclusion of the injury site is not required. This assumption is not without basis. Changes in axon structure and function after injury are well described in the wake of trauma. Graham and Gennarelli [9] describe spillage of protoplasm, altered organelle morphology, and changes in axon membrane and myelin sheath configuration throughout the axon length both proximal and distal to the focal site of injury as soon as 24 hr after the inciting event. Further, structural injury can continue up to 24 hr after the initial injury, with axotomy occurring where previously intact or stretched axonal membranes were noted. Long-term changes in anisotropy are not defined. To keep posttraumatic changes relatively consistent, we approximated a scanning window of a maximum of 7 days by extrapolating from information obtained in the pathology literature.
An intact axonal membrane controls the movement of water in the interstitial spaces of white matter fiber tracts to a single direction of flow that is parallel to the fiber tract orientation, running along the axon but not across it (i.e., anisotropic). Theoretically, trauma to the white matter alters individual axon membrane integrity or permeability along the whole of the injured axon. These physiologic changes can occur to one, several, or many fibers in a fascicle. The resulting change in membrane permeability makes water movement across the long-axis plane of the axon shaft possible, thus increasing the number of potential water diffusion directions (i.e., movement toward isotropic diffusion).
Measurement of anisotropy in large white matter regions reflects the bulk water flow characteristics of the constituent white matter fiber tracts. Because focal injuries affect membrane function of the entire axon, altered anisotropy should be apparent throughout the entire axon. The white matter region measured need not necessarily include the injured site. Diffuse axonal injury should then be reflected in the aggregate anisotropy characteristics of the fiber tracts.
All patients in this pilot study of the C-FAST were older than 21 years. Water diffusion characteristics in the developing brain are considerably more complex than in the adult, especially in the early years of myelination. A study involving pediatric head injury and anisotropy indexes would be much more complex, requiring attention to changing water diffusion characteristics over time. Therefore, evaluation of pediatric blunt head injury patients was not within the scope of this initial evaluation.
In our investigation, the C-FAST predicted death with an incremental univariate odds ratio of 5.13, which decreased slightly to 4.49 when adjusted for age and sex. This means that for each increment in the C-FAST, the likelihood of a patient with head trauma dying is 4.49 times that of someone who did not experience head trauma. So a patient with a C-FAST of 1 is 4.49 times more likely to die than an atraumatic individual, and a patient with a C-FAST of 4 is 17.96 times more likely to die.
Brain injury is the major contributor to death in the multitrauma patient. The C-FAST is a score hypothetically related to cerebral white matter pathophysiology and does not take into account injury to other systems. This isolation of brain injury predictors from the remainder of body predictors is a complex issue and not within the scope of a pilot study. To satisfy the question of relationships between head predictors and whole body predictors, we used a stepwise regression model-building technique, incorporating multiple clinical assessment variables. All clinical predictors (Glasgow Coma Scale, revised trauma score, and Abbreviated Injury Scale) were rejected in both the forward and backward constructions using a significance level of 0.05 for inclusion (forward) or exclusion (backward). Forced multivariate correlations performed with adjustment for these clinical predictors did not contribute to a coherent result, showing p values greater than 0.4. This finding may be due to patchy and incomplete clinical predictor data in our small subject data set. When we analyzed raw contingency tables for predictor and outcome against trauma, many cells showed low numbers. The score most significantly affected was the revised trauma score, when a cell with a score of 0 was encountered in the critical revised trauma score variable in nontrauma patients. This finding skews the result toward uniform (100%) predictability because the score fails to produce a false-positive result. A prospective analysis with a larger subject population may provide additional stability in these correlations.
In an attempt to see how good the clinical scales were at predicting the likelihood of a bad C-FAST, we constructed regression models of Glasgow Coma Scale, revised trauma score, and Abbreviated Injury Scale against a dichotomized "critical" form of the C-FAST. We selected a C-FAST of 3 as the critical value as described in Materials and Methods and dichotomized C-FAST into two values representing good and bad scores. Similar dichotomization of the predictors had already been performed as previously described. Regression in these cases showed the best predictor of a bad C-FAST to be findings of head CT. Modeling with a positive head CT result, adjusted for age and sex, showed that the likelihood of observing a bad C-FAST was 14.28 times that of an atraumatic subject, which is not surprising given the relative insensitivity of CT for head injury. That is, if injuries to the brain are severe enough to show an abnormality on CT, and if our supposition that fractional anisotropy is a more sensitive physiologic indicator is true, then we would expect the C-FAST to be critical. The other predictors also showed significance in the model, but all showed low sensitivity (Table 4).
|
Relative to the other clinical predictors used in this investigation, the C-FAST showed good correlation with outcome. Using the death outcome as a standard, we performed an ROC analysis to compare the effectiveness of all predictors in this model. As was shown by the family of ROC curves and indicated in Table 3, the area under the ROC curve for C-FAST was not significantly different from any of the other predictors used in this investigation except the revised trauma score. The revised trauma score alone showed the best result overall, with an area under the curve of 0.94. Using interaction term analysis, we formed combinations of the predictors to see if using the C-FAST in any combination with the clinical predictors resulted in a synergistic effect. The combined revised trauma score and C-FAST showed the best outcome with the greatest area under the ROC curve and was the only set of terms significantly different from the rest. It should be noted from the univariate analysis as previously discussed, that this outstanding performance of the revised trauma score variable is suspect because of our small study population. Although our results are on the surface promising, current data are unreliable and require further analysis with a larger population. All combinations showed improved predictor performance over the individual terms alone. Except for the combination of C-FAST and revised trauma score, combining predictors did not improve areas under the ROC curve significantly beyond that observed with the C-FAST alone. This finding suggests that using the C-FAST alone as a predictor of death is as sensitive as any combination of C-FAST with Glasgow Coma Scale, with Abbreviated Injury Scale, or with revised trauma score, and as sensitive as all scores taken together.
In our investigation, we assembled a score representative of major white matter volumes symmetrically throughout the brain. The statistics in this relatively small case control study confirm a measurable and significant difference between fractional anisotropy values in patients with posttraumatic head injuries as compared with nontrauma patients. Pearson's correlation coefficients in almost all cases were less than 0.30. Some of the instability in the statistics lies in the small sample size and lack of statistical power of the measurement. Another possible source of error arises in our control population. For convenience, we used as our control subjects a population of patients who were scheduled for head MRI for nontraumatic complaints. These patients may have white matter dysfunction not apparent on MRI that is being accounted for in the fractional anisotropy value. Such a situation might help explain the few control subjects with middle-range scores. Using an asymptomatic volunteer population may show a more accurate representation of the differences between normal and posttraumatic fractional anisotropy values.
As with any proposed clinical score, the initial data should be reviewed and a prospective investigation planned to refine and validate the predictability, and to establish the sensitivity and specificity, of the C-FAST. We suggest from our experience a larger population as suggested by population estimate calculations and recruitment of a healthy volunteer control population.
In conclusion, in this pilot study we proposed and tested the first clinically relevant index derived from cerebral diffusion MRI data; C-FAST shows promise in predicting outcome in brain-injured trauma patients. Good correlation was also noted between the C-FAST and clinically proven predictors.
Analyses performed on the raw anisotropy data and on contingency tables formed from the dichotomized clinical values suggest instability in some of the data because of the small study population. Although initial observations in using the C-FAST as a predictor of severe brain injury are promising, additional evaluation with a larger population is needed to confirm the validity of this new clinical index.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Ozturk, A.D. Sasson, J.A.D. Farrell, B.A. Landman, A.C.B.S. da Motta, A. Aralasmak, and D.M. Yousem Regional Differences in Diffusion Tensor Imaging Measurements: Assessment of Intrarater and Interrater Variability AJNR Am. J. Neuroradiol., June 1, 2008; 29(6): 1124 - 1127. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D.R. Rutgers, F. Toulgoat, J. Cazejust, P. Fillard, P. Lasjaunias, and D. Ducreux White Matter Abnormalities in Mild Traumatic Brain Injury: A Diffusion Tensor Imaging Study AJNR Am. J. Neuroradiol., March 1, 2008; 29(3): 514 - 519. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Sidaros, A. W. Engberg, K. Sidaros, M. G. Liptrot, M. Herning, P. Petersen, O. B. Paulson, T. L. Jernigan, and E. Rostrup Diffusion tensor imaging during recovery from severe traumatic brain injury and relation to clinical outcome: a longitudinal study Brain, February 1, 2008; 131(2): 559 - 572. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. L. Mac Donald, K. Dikranian, P. Bayly, D. Holtzman, and D. Brody Diffusion Tensor Imaging Reliably Detects Experimental Traumatic Axonal Injury and Indicates Approximate Time of Injury J. Neurosci., October 31, 2007; 27(44): 11869 - 11876. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N Nakayama, A Okumura, J Shinoda, Y-T Yasokawa, K Miwa, S-I Yoshimura, and T Iwama Evidence for white matter disruption in traumatic brain injury without macroscopic lesions J. Neurol. Neurosurg. Psychiatry, July 1, 2006; 77(7): 850 - 855. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) (Az, 0.8109),
bad revised trauma score (x) (Az, 0.9359), and bad
Abbreviated Injury Scale (
) (Az, 0.7853). Positive
indicator in these bad indexes refers to critical value in abnormal range for
C-FAST, Glasgow Coma Scale, revised trauma score, and Abbreviated Injury
Scale, respectively, as defined in Materials and Methods.