Original Investigations

Mild Traumatic Brain Injury:

Tissue Texture Analysis Correlated to Neuropsychological and DTI Findings

Kirsi K. Holli, MSc, Minna Waljas, MSc, Lara Harrison, MD, MSc, Suvi Liimatainen, MD, PhD,Tiina Luukkaala, MSc, Pertti Ryymin, PhDlic, Hannu Eskola, PhD, Seppo Soimakallio, MD, PhD,

Juha Ohman, MD, PhD, Prasun Dastidar, MD, PhD

Ac

FrDeNAcFiTeTaDiUgrUFo(HAd

ªdo

10

Rationale and Objectives: The aim of this study was to evaluate whether texture analysis (TA) can detect subtle changes in cerebral tissue

caused by mild traumatic brain injury (MTBI) and to determine whether these changes correlate with neuropsychological and diffusion

tensor imaging (DTI) findings.

Materials and Methods: Forty-two patients with MTBIs were imaged using 1.5T magnetic resonance imaging within 3 weeks after head

injury. TA was performed for the regions corresponding to the mesencephalon, centrum semiovale, and corpus callosum. Using DTI, the

fractional anisotropic and apparent diffusion coefficient values for the same regions were evaluated. The same analyses were performed ona group of 10 healthy volunteers. Patients also underwent a battery of neurocognitive tests within 6 weeks after injury.

Results: TA revealed textural differences between the right and left hemispheres in patients with MTBIs, whereas differences were minimal

in healthy controls. A significant correlation was found between scores on memory tests and texture parameters (sum of squares, sumentropy, inverse difference moment, and sum average) in patients in the area of the mesencephalon and the genu of the corpus callosum.

Significant correlations were also found between texture parameters for the left mesencephalon and both fractional anisotropic and

apparent diffusion coefficient values.

Conclusions: The data suggest that heterogeneous texture and abnormal DTI patterns in the area of the mesencephalon may be linkedwith verbal memory deficits among patients with MTBIs. Therefore, TA combined with DTI in patients with MTBIs may increase the ability to

detect early and subtle neuropathologic changes.

Key Words: Mild traumatic brain injury; texture analysis; magnetic resonance imaging; diffusion tensor imaging.

ªAUR, 2010

Establishing a diagnosis of mild traumatic brain injury

(MTBI) has proven to be challenging. Many systems

have been developed to classify traumatic brain

injuries (TBIs) from mild to severe. Most of these systems,

such as the Glasgow Coma Scale, the duration of loss of

consciousness, the presence of posttraumatic amnesia, and

assorted neuroimaging techniques, rely on acute injury

characteristics. The prognosis of moderate to severe traumatic

brain injury is somewhat well correlated with these character-

istics, but the relationship is not as clear for milder forms of

traumatic brain injuries. Neuropsychological testing provides

ad Radiol 2010; 17:1096–1102

om the Medical Imaging Center (K.K.H., L.H., P.R., H.E., S.S., P.D.), thepartment of Neurology and Rehabilitation (S.L.), the Department of

eurosurgery (M.W., J.O.), and the Department of Emergency Medicineuta (S.L.), Post Box 2000, Tampere University Hospital, 33521 Tampere,

nland; the Department of Biomedical Engineering, Tampere University ofchnology, Tampere, Finland (K.K.H., L.H., H.E.); Tampere Medical School,mpere, Finland (L.H., S.S., P.D.); the Science Center, Pirkanmaa Hospitalstrict, Tampere, Finland (T.L.); and Tampere School of Public Health,niversity of Tampere, Tampere, Finland (T.L.). This study was supported byants for research funding from the Pirkanmaa Hospital District, Tampereniversity Hospital (Tampere, Finland); the Jenny and Antti Wihuriundation (Helsinki, Finland); and the Instrumentarium Science Foundationelsinki, Finland). Received February 25, 2010; accepted April 12, 2010.dress correspondence to: K.K.H. e-mail: kirsi.holli@tut.fi

AUR, 2010i:10.1016/j.acra.2010.04.009

96

an assessment and quantification of brain function by

examining brain-behavior relationships. These measurements

evaluate memory recall, attention and concentration,

problem-solving abilities, visual tracking, reaction time, and

the speed of information processing, as well as other measures

of cognitive function. Neuropsychological assessment is

sensitive to the effects of traumatic brain injury and can be

helpful in obtaining more information about subjective

cognitive complaints in MTBI (1,2). Because brain injury

does not produce a unique pattern of neuropsychological

deficits, similar results may be seen that arise from the

cognitive difficulties associated with other injury-related

factors, such as insomnia, stress, and mood disturbances

(1,2). These issues must be taken into account when

neuropsychological data are analyzed.

Some victims of MTBIs may suffer from severe life-altering

problems, including cognitive and emotional difficulties.

Among patients with these chronic problems, the vast

majority have normal findings on computed tomographic

(CT) and magnetic resonance imaging (MRI) scans, the

current clinical modalities of choice for the diagnosis of acute

MTBI. With recent improvements, including diffusion-

weighted MRI, diffusion tensor imaging (DTI), and new

MRI sequences (3,4), certain MRI techniques provide even

more sensitivity for MTBI than standard MRI. However,

Academic Radiology, Vol 17, No 9, September 2010 TEXTURE ANALYSIS OF MTBI MR IMAGES

even though MRI has been found to be more sensitive for

traumatic lesions than CT imaging, it still has a tendency to

underestimate the condition; thus, an obvious need exists

for advanced complementary techniques for imaging MTBI.

DTI has emerged as a valuable additional technique for

investigating traumatic axonal injury in mild to severe trau-

matic brain injury (5–8). DTI is an advanced MRI

technique that can detect in vivo anisotropic diffusion

properties in white matter. Several diffusion indices can be

derived from the tensor matrix that characterizes the size,

shape, and orientation of the diffusion tensor. These include

the fractional anisotropic (FA) and apparent diffusion

coefficient (ADC) values. The ADC describes the average

diffusion of the media and has low values in white matter,

while FA values represent the magnitude of the diffusion

tensor and are high in white matter. In MTBI, FA

reduction has been demonstrated in the corpus callosum

(CC), internal and external capsules, and centrum

semiovale, in both the acute and chronic stages (5,9). ADC

and FA values differ between patients with MTBI and

normal controls and provide an early indicator and

prognostic measure of subsequent brain damage (5).

Texture is an image feature that corresponds to both the

brightness value and pixel locations. Basically, texture is the

visual cue resulting from the repetition of image patterns.

These patterns can be described, for example, as smooth or

rough, regular or irregular, or coarse or fine. On the basis of

the distribution of gray-level values, texture analysis (TA)

allows the calculation of mathematical texture features that

can be used to characterize the properties of tissues (10). TA

is a rather novel field of research, and its applications range

from the segmentation of anatomic structures and lesion detec-

tion to the differentiation of pathologic versus healthy tissue.

TA has already been successfully applied in many neurologic

disorders, such as Alzheimer’s disease, hippocampal sclerosis,

and multiple sclerosis (11–13) and in the identification of

pathologic versus normal tissues, such as in brain, liver,

breast, and other tumors (14–22). A number of techniques,

such as gray-level co-occurrence matrix (COM), run-length

matrix, an autoregressive model, fractal geometric methods,

and wavelet transforms, have been used for TA in medical

applications; among them, COM seems to be the most widely

used and most effective method. The COM measures the

probability that a pixel of a particular gray level occurs at a spec-

ified direction and distance from neighboring pixels.

The purpose of this study was to evaluate whether TA can

detect subtle changes in cerebral tissue caused by MTBI and to

determine whether these texture changes correlate with

neuropsychological and DTI findings.

MATERIALS AND METHODS

Study Subjects

Patients with MTBIs (Glasgow Coma Scale score, 13–15)

were recruited from the emergency room of the Tampere

University Hospital in 2006 and 2007. For the TA study, 42

consecutive patients (17 men, 25 women; mean age, 38.8 �13.6 years; age range, 18–60 years) were included. Clinical

examination on admission and CT imaging on the day of

the accident were performed on all patients. Neuropsycho-

logical assessment was performed within 6 weeks after injury,

and MRI was performed within approximately 3 weeks after

the date of admission. All patients met criteria for MTBI

according to the World Health Organization’s Collaborating

Centre for Neurotrauma Task Force on Mild Traumatic Brain

Injury (23). Exclusion criteria for this study were age < 18 or

> 65 years, severe traumatic brain injury, previous brain

trauma, preexisting major cognitive disorder, or history of

major alcohol or drug abuse. All 42 patients were evaluated

as having normal findings on CT and MRI scans by a special-

ized radiologist. In addition, 10 healthy age-matched and

sex-matched volunteers (four men, six women; mean age,

39.8 � 12.9 years; age range, 28–61 years) were recruited to

form a control group. The same exclusion criteria used for

the patients with MTBIs were applied to all controls. All

patients and healthy volunteers gave written consent, and

the study was approved by the ethics committee of Tampere

University Hospital.

Neurologic and Neuropsychological Examinations

To classify and assess the severity of trauma, neurologic and

neuropsychological examinations were performed on 34

patients. Each patient’s degree of consciousness was assessed

using the Glasgow Coma Scale to determine the severity of

brain injury (24). Estimated durations of loss of consciousness

and posttraumatic amnesia were recorded (minutes or hours).

A battery of neurocognitive tests was applied to assess

memory function within 6 weeks after injury. Two composite

scores were created: a verbal memory composite score and

a visual memory composite score. These scores are considered

to provide global estimates of a subject’s verbal and visual

memory in the acute postinjury phase. For the verbal memory

composite score, we measured total immediate recall, postin-

terference recall, and recognition with the Rey Auditory

Verbal Learning Test (25) and the Four Word Short-Term

Memory Test (26). For the visual memory composite score,

the total adjusted errors and the total adjusted errors for six

shapes from the Cambridge Neuropsychological Test

Automated Battery Paired Associates Learning Subtest (27)

and Rey-Osterrieth Complex Figure Test immediate recall

(28) were used. All raw scores from neurocognitive tests

were converted into z scores and adjusted for age and educa-

tion (29). The composite scores for verbal and visual memory

were calculated by averaging the standardized scores from the

tests that assessed each domain.

MRI Examinations

All 42 patients were studied using a 1.5-T MRI machine

(Magnetom Avanto; Siemens Medical Solutions, Erlangen,

1097

HOLLI ET AL Academic Radiology, Vol 17, No 9, September 2010

Germany) with a 12-channel head matrix coil. The MRI

machine is maintained on a quality control protocol that

includes daily, monthly, and quarterly evaluations; in addition,

the main magnetic field homogeneity and radiofrequency

amplifier properties are measured and calibrated four times

a year. A prescan normalization filter was used to correct

the intensity inhomogeneity in the images. The data used

for homogenization were acquired using a preliminary low-

resolution measurement. An elliptical filter was used within

the slice planes to improve the signal-to-noise ratio.

The MRI protocol included all routinely used sequences,

but only axial fluid-attenuated inversion recovery (FLAIR;

repetition time, 9000 ms; echo time, 109 ms; inversion

time, 2500 ms; slice thickness, 5.0 mm; slice gap, 1.5 mm;

matrix size, 256 � 256; field of view, 230 mm) and sagittal

T1-weighted three-dimensional magnetization prepared

gradient echo (repetition time, 1910 ms; echo time, 3.1 ms;

inversion time, 1100 ms; slice thickness, 1.0 mm; slice gap,

0 mm; matrix size, 256 � 256; field of view, 250 mm; flip

angle, 15�) sequences were used in this study for TA. Diffusion

tensor was estimated by an axial multidirectional diffusion-

weighted spin echo echoplanar imaging sequence (b = 0,

b = 1000; repetition time, 3600 ms; echo time, 96 ms; inver-

sion time, 0 ms; slice thickness, 5.0 mm; slice gap, 1.5 mm;

matrix size, 128 � 128; field of view, 230 mm) with 12

diffusion gradient orientations.

TA

TA was performed using the software package MaZda version

4.5 (Technical University of Lodz, Institute of Electronics,

Lodz, Poland), which was specially designed for TA by Hajek

et al (30) as part of the European COST B11 program. For

TA, two different imaging series were selected from among

all of the magnetic resonance images: an axial FLAIR

(T2-weighted FLAIR) and sagittal T1-weighted three-

dimensional magnetization prepared gradient echo (T1w

Figure 1. T2-weighted fluid-attenuated inversion recovery images from

centrum semiovale, (c) corpus callosum in the caudo-cranial direction. R

1098

3D magnetization prepared gradient echo T1-weighted

MPR]). Two image slices from the T2-weighted FLAIR

sequence at selected levels of interest and one image slice of

the sagittal view in the caudocranial direction were chosen

for each subject. The levels of interest in the T2-weighted

FLAIR images were at the levels of the mesencephalon and

centrum semiovale. Regions of interest (ROIs) were drawn

manually around the mesencephalon (ROI size, about 1200

pixels, depending on the size of the mesencephalon) and

included both the left and right sides to cover the entire

mesencephalon (Fig 1a). Circular, standard-size ROIs (177

pixels) were manually placed on the image, located symmetri-

cally on both sides at the level of the centrum semiovale (Fig

1b). Circular ROIs (68 pixels) were also placed on the genu,

body, and splenium of the CC (Fig 1c).

In this study, we considered only the COM-based parame-

ters. The calculated COM parameters were as follows: angular

second moment, contrast, correlation, sum of squares, inverse

difference moment, sum average, sum entropy, entropy, differ-

ence variance, and difference entropy (30). In MaZda, the

distances d = 1, 2, 3, 4, and 5 and directions q = 0�, 45�,90�, and 135� are considered. In this study, however, given

the small size of some ROIs, we evaluated the textural features

from the COM only when the distance was 1, 2, or 3. To

minimize the influence of contrast variation and brightness,

gray-level normalization for each ROI was performed using

a method that normalizes image intensities in the range

(m � 3s and m + 3s, where m is the gray-level mean, and s

is the gray-level standard deviation).

DTI Analysis

Diffusion parameter analyses were performed on 34 patients

using the experimental software DTI Task Card version

1.70, developed at Massachusetts General Hospital (Boston,

MA) in cooperation with Siemens. Analysis of the remaining

patients and healthy volunteers were performed using the

selected levels of interest. (a) level of mesencephalon, (b) level of

egions of interest are drawn on the images.

TABLE 1. Numbers and Percentages of COM Parameters with Statistically Significant Differences (P < .05) Between Hemispheresand Segments of the Corpus Callosum Analyzed with Wilcoxon’s Test

Differing Parameters

Patient Healthy Volunteers

Region of Interest COM Sequence (n = 42) (n = 10)

Mesencephalon, right vs left n = 123 T2-weighted FLAIR 55 (45%) 14 (12%)

Centrum semiovale, right vs left n = 123 T2-weighted FLAIR 27 (22%) 2 (2%)

Corpus callosum, all segments n = 123 T1-weighted MPR 30 (24%) 1 (1%)

COM, co-occurrence matrix; FLAIR, fluid-attenuated inversion recovery; MPR, magnetization prepared gradient echo.

In all, 123 texture parameters were evaluated.

Academic Radiology, Vol 17, No 9, September 2010 TEXTURE ANALYSIS OF MTBI MR IMAGES

newer software Neuro 3D, which runs on the Syngo MMWP

workstation version VE26A (Siemens AG, Berlin, Germany).

Anisotropic diffusion can be visualized as ellipsoids, FA maps,

and tractographic views. In the tensor mode, FA, ADC, diffu-

sion weighted imaging (DWI), eigenvalues, and low-b maps

can be visualized. In our study, ADC and FA values were

measured symmetrically at the mesencephalon, centrum

semiovale, and CC (genu, body, and splenium).

Statistical Analysis

Differences in textural features between the right and left

perimesencephalic areas and centrum semiovale were

analyzed using Wilcoxon’s signed-rank test, and the textural

parameters calculated from the segments of the CC (genu,

body, and splenium) were analyzed using Friedman’s test.

These tests were run for both patients and healthy volunteers.

Spearman’s r test was performed to evaluate the relation-

ship in patients between the composite scores for verbal and

visual memory functions and the TA parameters. For areas

in which significant interactions were found between the

composite scores and TA parameters, we also compared the

DTI and TA results and the DTI findings with the composite

scores, using a nonparametric Spearman’s r test. These anal-

yses were performed with SPSS for Windows version 14.0.2

(SPSS, Inc, Chicago, IL). A P value < .05 was considered

statistically significant.

RESULTS

TA Results

The numbers and percentages of the COM texture param-

eters (n = 123, 11 parameters calculated in four different

directions over three distances) that were statistically signif-

icantly different (P < .05) between the areas analyzed in

patients compared to healthy volunteers are given in

Table 1.

In the mesencephalic area, 45% of all parameters were

significantly different when images from patients were

analyzed. These parameters consisted mainly of the entropy-

based parameters and angular second moment. With healthy

volunteers, there were clearly fewer significantly different

parameters; however, the parameter angular second moment

in the mesencephalic area also differed in healthy volunteers.

Texture Parameter Correlation to Memory Test Scores

We compared the TA results to the memory composite scores.

The final number of subjects in this part of the analysis was 34,

because eight subjects had dropped out of the neuropsycholog-

ical examinations. Significant correlations were found in the

mesencephalic area and the genu of the CC (Tables 2 and 3).

In the mesencephalic area, many textural parameters

correlated significantly with results from the memory tests.

The sum of squares and sum entropy in particular were

negatively correlated with verbal memory on the right side,

while on the left side, the sum of squares was negatively corre-

lated with verbal memory, and the sum average was positively

correlated with visual memory. Although the correlations

were not very strong, they were consistent.

In addition, several parameters measured in the genu of the

CC correlated with results from the visual memory tests

(Table 3). Correlations were also observed in the CC

splenium, but these were not consistent for any specific

parameters.

Texture Correlation to DTI Results

For areas in which significant interactions were found

between composite scores and TA parameters, we also

compared (1) DTI results to TA results and (2) DTI to

composite memory scores. We only included the DTI results

of 34 patients produced by the older software, to avoid errors

that might have been caused by using different software.

Texture parameters and FA and ADC values were

compared to determine whether the defined texture parame-

ters and changes in DTI values were correlated. Statistically

significant correlations were observed in patients in the left

mesencephalon. The texture parameters inverse difference

moment and angular second moment were positively

correlated with FA values, while the entropy-based parame-

ters were negatively correlated with FA values. However,

these correlations were not very strong, with the correlation

coefficient varying in negative correlations from r = �0.379

(P = .030) to r = �0.346 (P = .049) and in positive

1099

TABLE 2. Correlations and Statistical Significance Between Texture Parameters From the Right and Left Sides of theMesencephalon and the Memory Composite Scores (n = 34)

Mesencephalon Right Mesencephalon Left

Texture Parameter Verbal Memory Texture Parameter Verbal Memory

Sum of squares

(12/12 parameters)

r �0.393 to �0.490 Sum of squares

(6/12 parameters)

r �0.341 to �0.372

P .022 to .003 P .049 to .030

Sum entropy

(8/12 parameters)

r �0.344 to �0.437 Inverse difference

moment (3/12 parameters)

r 0.356 to 0.385

P .046 to .010 P .039 to .025Visual memory Visual memory

No statistically

significant

correlation

Sum average

(5/12 parameters)

r 0.387 to 0.494

P .029 to .004

HOLLI ET AL Academic Radiology, Vol 17, No 9, September 2010

correlations from r = 0.352 (P = .044) to r = 0.492 (P = .018).

The correlations between texture parameters and ADC values

were in opposing directions but were of similar magnitude.

Correlation of DTI to Memory Test Scores

We compared DTI values to results from the memory tests to

determine whether the memory test results and the differ-

ences observed in diffusion parameters ADC and fractional

anisotropy were correlated.

A positive correlation was observed between FA values in

the left side of mesencephalon and verbal memory

(r = 0.387, P = .046), but not visual memory (r = 0.280,

P = .186). No significant correlation was observed between

any memory test and ADC or FA values in the centrum semi-

ovale. A statistically significant negative correlation was

observed between ADC values in the body of the CC and

visual memory (r = �0.588, P = .005).

DISCUSSION

Conventional MRI and CT imaging often underestimate the

extent of injury following MTBI. Therefore, additional tools

are being developed to improve the diagnosis of MTBI. DTI

enables the measurement of the directional diffusivities of

water and is more sensitive than conventional MRI sequences

for detecting white matter injury. TA for medical images has

also attracted many investigators, several of whom have devel-

oped computer-aided detection systems to increase the

precision and accuracy of lesion detection and tissue

TABLE 3. Correlations and Statistical Significance BetweenTexture Parameters From the Genu of the Corpus Callosumand the Visual Memory Composite Scores (n = 34)

Texture Parameter Visual Memory

Entropy (4/12 parameters) r �0.390 to �0.397

P .027 to .025

Angular second moment

(3/12 parameters)

r 0.356 to 0.385

P .039 to .025

1100

characterization; however, to the best of our knowledge,

TA has not been applied to studying MTBI. Textural proper-

ties are evaluated on a millimeter scale and may capture the

local coherence, direction, and density of fiber bundles; their

myelinization status; and the density and direction of vessels

supplying and draining the white matter. Most of the time,

patients with MTBIs have normal MRI results; thus, TA

accompanied by DTI analysis may enable detection of

damaged structures.

Textural changes between the hemispheres in the mesence-

phalic area, centrum semiovale, and several segments of the

CC in patients with MTBIs and healthy volunteers were

evaluated. On conventional MRI, no lesions were observed

in patients, but TA revealed textural differences between the

hemispheres and in segments of the CC. In the mesencephalic

area in particular, many texture parameters differed at statisti-

cally significant levels between the left and right sides. Forty-

five percent of the COM-based parameters evaluated differed

statistically; these included mainly the entropy-based parame-

ters and angular second moment. Entropy indicates the

complexity and randomness within the ROI, while the

angular second moment measures the monotonicity of

gray-level transition in the image texture. These findings

indicate that one side of the mesencephalon is more complex

and heterogeneous on texture, while the other side is more

homogeneous. A homogeneous texture can be thought of

as containing repeated ideal structures; this uniformity

produces idealized patterns. A heterogeneous texture mostly

refers to an image in which repetition and spatial self-

similarity are absent. The textural differences between hemi-

spheres in healthy volunteers, conversely, were minimal.

We compared the TA results with the composite scores for

verbal and visual memory. Significant correlations were

identified in the mesencephalic area and in segments of the

CC. The sum of squares and sum entropy were negatively

correlated with verbal memory test scores on the right side,

while on the left side, the sum of squares was negatively

correlated and the inverse difference moment positively

correlated with verbal memory test scores.

The inverse difference moment is an indicator of homoge-

neity within the image texture, similar to the angular second

Academic Radiology, Vol 17, No 9, September 2010 TEXTURE ANALYSIS OF MTBI MR IMAGES

moment. These parameters are greater in a smooth, homoge-

neous image. The sum of squares, on the other hand, describes

the extent to which a measurement differs from the mean.

A homogeneous scene contains only a few gray levels, so the

sum of squares will be high. The sum entropy, like entropy, is

a measure of the complexity of gray-level distribution, and

higher entropy values indicate greater randomness in the image.

These results indicate that if the texture of the measured region

is complex and heterogeneous, the patient will not perform as

well on verbal memory tests, compared to an area found to be

very homogeneous. We therefore assume that if the texture in

the mesencephalic area is random and heterogeneous, then

MTBI has damaged this region, and the patient may have diffi-

culty with the memory tests. On the right side of the mesen-

cephalon, there was no correlation between the textural

parameters and visual memory tests. On the left, a correlation

was found between the sum average and visual memory tests.

The sum average is an indication of homogeneous brightness

and measures the average normalized gray-tone image in the

spatial domain. It takes on higher values if the measured area

is homogeneously bright and low values if the area is homoge-

neously dark. These correlations were not very strong but were

consistent. Correlations between texture parameters and the

visual memory composite score were also observed in the

genu of the CC. In this case, entropy was negatively correlated,

whereas the angular second moment was positively correlated,

indicating that if the entropy value were to increase and the

angular second moment to decrease, the area measured was

likely to be heterogeneous, and the patient’s performance on

visual memory tests would be poor. Some correlations were

also seen in the splenium of the CC, but these were not

consistent for any particular parameters.

For areas in which significant interactions were found

between memory composite scores and TA parameters, we

also compared the DTI and TA results and the DTI findings

to the memory composite scores. Significantly correlating

parameters were observed in patients at the left side of mesen-

cephalon with FA and ADC values. The texture parameters

inverse difference moment and angular second moment were

positively correlated with FA values, whereas entropy-based

parameters were negatively correlated with FA values. The

correlations between texture and ADC values were in opposite

directions, but themagnitudeswere similar.Thus, in a heteroge-

neous area, the entropy values increased, the inverse difference

moment and angular second moment decreased, the FA value

decreased, and the ADC increased. Decreased FA and increased

ADC values have been previously demonstrated to follow

MTBIs (7,8,31,32). The DTI changes in white matter have

also been found to correlate with post-concussion syndrome

(PCS) severity and emotional distress (8,33). DTI values were

compared to the memory test results to determine whether

memory test results and the differences observed in the

diffusion parameters ADC and fractional anisotropy were

correlated, and a significant positive correlation was found

between FA values and verbal test results in the left side of

mesencephalon.

Newer MRI techniques such as DTI may provide more

information on MTBI than current clinical neuroimaging

methods, especially when complemented by other techniques

such as TA. Although our findings suggest that DTI combined

with TA may provide biologic insights into the damage caused

by MTBI, more studies are needed to identify more precise

correlations between DTI, TA changes, and clinical outcomes

before the role of these methods can be established in the

diagnosis of MTBI. Nonetheless, our results are consistent

with the outcomes from memory tests.

CONCLUSIONS

For detecting subtle texture changes on magnetic resonance

images of patients with MTBIs that cannot be visually

observed, TA in combination with DTI and other neurologic

and neuropsychological tests may provide valuable additional

insights into the pathophysiology of MTBI. The data suggest

that heterogeneous texture and abnormal DTI patterns in the

mesencephalic area and the genu of the CC may be associated

with memory deficits in patients with MTBI. Combining TA

with DTI in patients with MTBI may improve our chances of

detecting early and subtle neuropathologic changes, thereby

facilitating better management of MTBI.

REFERENCES

1. Alexander MP. Mild traumatic brain injury: a review of physiogenesis and

psychogenesis. Neurology 1995; 45:1253–1260.

2. Binder LM. A review of mild head trauma. Part II: clinical implications. J Clin

Exp Neuropsychology 1997; 19:432–457.

3. Posse S, Tedeschi R, Risinger R, et al. High speed 1H spectroscopic

imaging in human brain by echo planar spatial-spectral encoding. Magn

Reson Med 1995; 33:34–40.

4. Raucher A, Sedlacik J, Deistung A, et al. Susceptibility weighted imaging:

data acquisition, image reconstruction and clinical applications. Z Med

Phys 2006; 16:240–250.

5. Inglese M, Makani S, Johnson G, et al. Diffuse axonal injury in mild

traumatic brain injury: a diffusion tensor imaging study. J Neurosurg

2005; 103:298–303.

6. Huisman TA, Schwamm LH, Schaefer PW, et al. Diffusion tensor imaging

as potential biomarker of white matter injury in diffuse axonal injury. AJNR

Am J Neuroradiol 2004; 25:370–376.

7. Rutgers DR, Fillard P, Paradot G, et al. Diffusion tensor imaging character-

istics of the corpus callosum in mild, moderate, and severe traumatic brain

injury. AJNR Am J Neuroradiol 2008; 29:1730–1735.

8. Kumar R, Gupta RK, Husain M, et al. Comparative evaluation of corpus

callosum DTI metrics in acute mild and moderate traumatic brain injury:

its correlation with neuropsychometric tests. Brain Inj 2009; 23:675–685.

9. Arfanakis K, Haughton VM, Carew JD, et al. Diffusion tensor MR imaging in

diffuse axonal injury. AJNR Am J Neuroradiol 2002; 23:794–802.

10. Tuceryan M, Jain AK. Texture analysis. In: Chen CH, Pau LF, Wang PSP,

eds. The handbook of pattern recognition and computer vision. 2nd ed.

River Edge, NJ: World Scientific Publishing, 1998; 207–248.

11. Freeborough PA, Fox NC. MR image texture analysis applied to the diag-

nosis and tracking of Alzheimer’s disease. IEEE Trans Med Imaging 1998;

17:475–479.

12. Bonilha L, Kobayashi E, Castellano G, et al. Textural analysis of hippo-

campal sclerosis. Epilepisa 2003; 44:1546–1550.

13. Zhang J, Tong L, Wang L, et al. Texture analysis of multiple sclerosis:

a comparative study. Magn Reson Imaging 2008; 26:1160–1166.

14. Kjaer L, Ring P, Thomsen C, Henriksen O. Texture analysis in quantitative

MR imaging—tissue characterization of normal brain and intracranial

tumours at 1.5T. Acta Radiol 1995; 36:127–135.

1101

HOLLI ET AL Academic Radiology, Vol 17, No 9, September 2010

15. Mahmoud-Ghoneim D, Alkaabi MK, de Certaines JD, et al. The impact of

image dynamic range on texture classification of brain white matter. BMC

Med Imaging 2008; 23:8–18.

16. Schad LR, Bluml S, Zuna I. MR tissue characterization of intracranial

tumors by means of texture analysis. Magn Reson Imaging 1993; 11:

889–896.

17. Herlidou-Meme S, Constans JM, Carsin B, et al. MRI texture analysis on

texture test objects, normal brain and intracranial tumours. J Magn Reson

Imaging 2003; 21:989–993.

18. Jirak D, Dezortova M, Taimr P, Hajek M. Texture analysis of human liver.

J Magn Reson Imaging 2002; 15:68–74.

19. Holli KK, Laaperi A-L, Harrison L, et al. Characterization of breast cancer

types by texture analysis of magnetic resonance images. Acad Radiol

2010; 17:135–141.

20. Gibbs P, Turnbull LW. Textural analysis of contrast-enhanced MR images

of the breast. Magn Reson Med 2003; 50:92–98.

21. Harrison L, Dastidar P, Eskola H, et al. Texture analysis on MRI images of

non-Hodgkin lymphoma. Comput Biol Med 2008; 38:519–524.

22. Harrison L, Luukkaala T, Pertovaara H, et al. Non-Hodgkin lymphoma

response evaluation with MRI texture classification. J Exp Clin Cancer

Res 2009; 28:87–100.

23. Carrol LJ, Cassidy JD, Holm L, et al. Methodological issues and research

recommendations for mild traumatic brain injury: the WHO Collaborating

Centre Task Force on Mild Traumatic Brain Injury. J Rehabil Med Suppl

2004; 43:113–125.

1102

24. Teasdale G, Jennett B. Assessment of coma and impaired consciousness.

A practical scale. Lancet 1974; 11:81–84.

25. Lezak M, Hovieson D, Loring D. Neuropsychological assessment. 4th ed.

New York: Oxford University Press, 2004.

26. Morrow LA, Ryan C. Normative data for a working memory test: the Four

Word Short-Term Memory Test. Clin Neuropsychol 2002; 16:373–380.

27. CANTAB�, the Cambridge Neuropsychological Automated Testing

Battery. Cambridge, UK: Cambridge Cognition, 2004.

28. Strauss E, Sherman EMS, Spreen O. A compendium of neuropsychologi-

cal tests: administration, norms, and commentary. 3rd ed. New York:

Oxford University Press, 2006.

29. Mitrushina M, Boone KB, Razani J, D’Elia L. Normative data for

neuropsychological assessment. 2nd ed. New York: Oxford University

Press, 2005.

30. Hajek M, Dezortova M, Materka A, Lerski R, eds. Texture analysis for

magnetic resonance imaging. Prague, Czech Republic: Med4Publishing,

2006.

31. Lo C, Shifteh K, Gold T, et al. Diffusion tensor imaging abnormalities in

patients with mild traumatic brain injury and neurocognitive impairment.

J Comput Assist Tomogr 2009; 33:293–297.

32. Mayer AR, Ling J, Mannell MV, et al. A prospective diffusion tensor imaging

study in mild traumatic brain injury. Neurology 2010; 74:643–650.

33. Chu Z, Wilde EA, Hunter JV, et al. Voxel-based analysis of diffusion tensor

imaging in mild traumatic brain injury in adolescents. AJNR Am J Neurora-

diol 2010; 3:340–346.