Differentiation between low-grade and high-grade glioma using combined diffusion tensor imaging metrics.

G Model CLINEU-3499; No. of Pages 7

ARTICLE IN PRESS Clinical Neurology and Neurosurgery xxx (2013) xxx–xxx

Contents lists available at ScienceDirect

Clinical Neurology and Neurosurgery journal homepage: www.elsevier.com/locate/clineuro

Differentiation between low-grade and high-grade glioma using combined diffusion tensor imaging metrics Lin Ma a,b , Zhi Jian Song a,b,∗ a b

Digital Medical Research Center, Fudan University, Shanghai, China Shanghai Key Lab of Medical Image Computing and Computer Assisted Intervention, Shanghai, China

a r t i c l e

i n f o

Article history: Received 17 September 2012 Received in revised form 27 September 2013 Accepted 2 October 2013 Available online xxx Keywords: Diffusion tensor imaging High-grade glioma Low-grade glioma Magnetic resonance imaging

a b s t r a c t Objective: To ascertain whether diffusion tensor imaging (DTI) metrics including tensor shape measures such as planar and spherical isotropy coefficients (CP and CS) can be used to distinguish high-grade from low-grade gliomas. Methods: Twenty-five patients with histologically proved brain gliomas (10 low-grade and 15 highgrade) were included in this study. Contrast-enhanced T1-weighted images, non-diffusion weighted b = 0 (b0) images, fractional anisotropy (FA), apparent diffusion coefficient (ADC), CS and CP maps were coregistered and each lesion was divided into two regions of interest (ROI): enhancing and immediate peritumoral edema (edema adjacent to tumor). Univariate and multivariate logistic regression analyses were applied to determine the best classification model. Results: There was a statistically significant difference in the multivariate logistic regression analysis. The best logistic regression model for classification combined three parameters (CS, FA and CP) from the immediate peritumoral part (p = 0.02), resulting in 86% sensitivity, 80% specificity and area under the curve of 0.81. Conclusion: Our study revealed that combined DTI metrics can function in effect as a non-invasive measure to distinguish between low-grade and high-grade gliomas. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Gliomas are the most common primary malignant brain tumors. Although the multimodal treatment of gliomas in recent years has made great progress, relative to the low-grade gliomas, the prognosis of patients with high-grade gliomas remains poor [1,2] and their mean survival time is merely 14.6 months [3]. Treatment strategies including surgical resection, radiotherapy, and chemotherapy for these two grades are very different. Therefore, it is crucial to discriminate accurately between high-grade and low-grade gliomas preoperatively. With the gold standard of glioma grading is still based on the final postoperative histologic examination of appropriately sampled tissue, preoperative grading of glioma diagnosis relies usually on clinical and neuro-imaging features such as enhancing and peritumoral boundaries, mass effect, necrosis, tumor hemorrhage, and degree of enhancement signs. However, differential diagnosis is

∗ Corresponding author at: Digital Medical Research Center, Fudan University, P.O. Box 251, 138 Yixueyuan Road, Shanghai 200032, China. Tel.: +86 21 54237054; fax: +86 21 54237797. E-mail addresses: [email protected] (L. Ma), [email protected] (Z.J. Song).

still challenging when both low-grade and high-grade gliomas may have similar enhancement patterns and exhibit extensive edema [4] on magnetic resonance imaging (MRI), especially when the clinical history cannot provide valuable information. Diffusion tensor imaging (DTI) as an advanced magnetic resonance technique provides visibility into the movement of water molecules. Thus far, a few researches were involved in analysis on DTI metrics of enhancing and peritumoral edema areas such as fractional anisotropy (FA) and apparent diffusion coefficient (ADC) for identification of gliomas and histopathological features of tumor, however, with mixed results. Here, we tried to assess the effect of Combined DTI metrics with similar enhancing imaging because single metrics have been argued in favor and against every option in the evaluation of glioma grading. Some reports [5–9] have suggested that FA, ADC and tumor cell density of the tumor core show good correlation that is helpful for the differentiation. Kinoshita et al., Beppu et al., etc. [5,6,8] found that FA has a good positive correlation and mean ADC (MD) has a good negative correlation with tumor cell density within the tumor core, while Stadlbauer et al., Lee et al., etc., have got the opposite conclusions that FA is negatively correlative and the ADC is positively correlative [6,9]. In 2005, Inoue et al. investigated 41 patients historically proved with WHO I–IV and reported the higher grade the higher FA value [10].

0303-8467/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.clineuro.2013.10.003

Please cite this article in press as: Ma L, Song ZJ. Differentiation between low-grade and high-grade glioma using combined diffusion tensor imaging metrics. Clin Neurol Neurosurg (2013), http://dx.doi.org/10.1016/j.clineuro.2013.10.003


ARTICLE IN PRESS L. Ma, Z.J. Song / Clinical Neurology and Neurosurgery xxx (2013) xxx–xxx

2

Whereas, other studies suggest that FA [11–13] or ADC [10,14–17] cannot be used for glioma grading. Regarding ADC researches, Lu et al. found no significant difference in ADC values between low-grade and high-grade astrocytoma [16]. Other reports show similar results. Inoue et al. concluded ADC values were affected by many factors that make ADC a poor indicator to accurate glioma grading [10]. Conflicting results are also found in employing FA to predict the malignancy of gliomas [10–13]. Contrary to previous views, Einar reported results in 2006 by testing WHO grade II and grade III that demonstrated FA values were lower with the higher glioma grade in peritumoral area [12]. Besides, a few other reports found no significant difference in FA values between glioma grades [11,13]. These discrepancies may be caused by the differences in acquisition and analytical methods applied. It can also be noted that FA and ADC may provide limited information and other tensor shape obtained from DTI measurements, such as linear and planar anisotropy coefficients (CL and CP), also explain additional information and more tissue characterization [18–21]. The majority of earlier studies have merely focused on FA or/and ADC two DTI parameters for differential diagnosis of gliomas. In addition, some researchers confirmed that DTI metrics can also be helpful in discrimination between primary glioma and other intracranial tumors [22–24]. However, till now there has been no report on applying the combined DTI metrics to distinguish between high-grade and low-grade gliomas, which might be a more effective way of characterizing gliomas. To date, the preoperative distinction between high- and lowgrade gliomas is making progress. For example, metabolic imaging such as spectroscopy and perfusion scanning are promising, but their accuracy and reliability are controversial [25–27]. This study, therefore, aimed to determine whether integrated DTI parameters could be correlated with classification by calculating FA, ADC, CP, and CS values in enhancing region and immediate peritumoral region. 2. Materials and methods 2.1. Patients The patients recruited to this study were admitted to the neurosurgery department of Huashan Hospital from November 2006 to May 2010. The tumor imaging of all the patients we collected is positive with Gd contrast. A total of 25 patients (16 males, 9 females; average age 45.08 years; age range 11–71 years) were diagnosed on postoperative histology to be World Health Organization grade I–IV; 10 had low-grade gliomas (1 grade I and 9 grade II), while the remaining 15 had high-grade gliomas (2 grade III and 13 grade IV). All patients underwent Contrast-enhanced (CE) T1-weighted and DTI preoperatively. No medical therapy was received for their tumor prior to imaging.

geometric distortion. A total of 5 averages were collected to ensure sufficient signal-to-noise ratio for high quality tensor or mapping, leading to a total DTI data acquisition time of approximately 5 min. Axial T1-weighed images were obtained with field of view 240 × 240, TR 440, TE 13 ms, one excitation, 320 × 192 matrix, 5-mm-thick sections with 1.5-mm intersection gap. 2.3. Image processing All the diffusion images were processed off-line. DTI series and b0 image series were spatially normalized and computed using the DTI Studio, Version 2.4 software (H. Jiang, S. Mori, John Hopkins University, http://cmrm.med.jhmi.edu) Pixelwise ADC, FA, CP and CS maps were computed using the following standard algorithms [28]: ADC =

1 + 2 + 3 3

3 2

FA =

2

2

(1 − ) + (2 − ) + (3 − )

2

21 + 22 + 23

2(2 − 3 )

CP =

CS =

21 + 22 + 23 33

21 + 22 + 23

where 1 , 2 and 3 are the three eigenvalues of the diffusion tensor, the representation of the water molecules movement in the X, Y and Z axis which reflects the ellipsoid shape and structure [29,30], and denotes the mean of the three eigenvalues. 2.3.1. Determination of region of interest The scalar DTI maps (FA, ADC, CS and CP) were co-registered to non-diffusion weighted b = 0 (b0) images overlayed on CE T1weighted images to synchronously demonstrate the tumor core and peritumoral area applying a 3D non-rigid transformation and mutual information using a DTI analysis software self-complied by our Lab which was certified by the China State Food and Drug Administration [31]. For each case, each lesion was divided into two regions of interest (ROI), later on, rectangle ROIs (20–40 pixels) were placed in the tumor strong signal enhancing area (ROI i) on the CE T1-weighted images and the hyperintense areas most adjacent (1–3 mm) to the enhancing portion of the tumor on b0 images which was defined as the immediate peritumoral edema area (ROI ii) (Fig. 1). Areas of necrosis and hemorrhage were removed from the analysis. The ROIs were automatically delivered onto the above-mentioned co-registered scalar DTI maps. For each patient, 2 or 3 ROIs were applied (depending on the size of edema) in each area to calculate the average FA, ADC, CP and CS values respectively. All ROIs were drawn on the axial slice with maximum tumor size, locating in the tumor core and immediate peritumoral edema area. Another operator validated the data through repeating the experiment with the same methodology.

2.2. MRI

2.4. Statistical analysis

MRI imaging was performed by using a 1.5-T MR system (Sonata; Siemens, Erlangen, Germany) with a diffusion-weighted echo-planar imaging sequence and following parameters: flip angle, 90◦ ; repetition time/echo time, 1000/112 ms; slice thickness, 5 mm; interval, 0 mm; matrix size, 128 × 128; field of view, 220 mm × 220 mm; number of excitations, 1. Diffusion weighting factor (b) of b = 1000 s/mm2 , as well as b = 0 s/mm2 (no diffusion gradient), 23 diffusion directions. The sequence design was based on balancing diffusion gradients to minimize eddy currents and

The difference of the registration error between low-grade and high-grade groups was assessed using a Mann–Whitney U test. A p-value of less than 0.05 was considered significant. Association between each DTI parameter and gliomas classification was evaluated using a univariate logistic regression and the parameters with a high predictive power (p < 0.30) were chosen. Areas under the receiver operating characteristic (ROC) curves (AUC) were calculated for each parameter to further assess their efficacy. Cutoff values were then estimated by maximizing the sum


ARTICLE IN PRESS


L. Ma, Z.J. Song / Clinical Neurology and Neurosurgery xxx (2013) xxx–xxx

3

Fig. 1. MRI of a patient with right triangle gliosarcoma. A, contrast-enhanced T1-weighted image showing the enhancing lesion. B, non-diffusion weighted b = 0 (b0) image showing extensive edema area. C, the co-registered map of b0 image (B) in blue overlayed on CE-T1 image (A) in red showing the tumor core and peritumoral area synchronously. D, the co-registered map of the scalar DTI maps (green) overlayed on CE-T1 W and b0 images (C) indicating regions of interests (ROI): ROI i (blue rectangle) placed in the enhancing portion of tumor (areas of necrosis and hemorrhage are exempted) on T1, ROI ii (yellow rectangle) placed the hyperintense areas adjacent to tumor enhancing area as immediate peritumoral edema on b0 image.

of sensitivity and specificity. The selected parameters were incorporated into a multivariate logistic regression analysis to decide the best independent predictors when controlling for other potential confounding variables. Thus, an optimal logistic regression model (LRM) was constructed to distinguish low-grade and highgrade gliomas (p < 0.05). Model fit was assessed by means of the Hosmer–Lemeshow goodness-of-fit test [32]. Areas under the ROC curves (AUC) were also calculated employing combination of DTI parameters to further assess their classified abilities. Cutoff values of combination were estimated as well. All statistical analyses were performed using the SPSS 15.0 (SPSS Inc., Chicago, IL, USA) software package.

3. Results By use of our self-compiled software, the DTI map and b0 image were registered to CE-T1 successfully as Fig. 1C and D shown. The registration errors of high-grade gliomas range from 0.60 mm to 2.10 mm, averaging (1.70 ± 0.48) mm (mean ± SD) and that of lowgrade gliomas from 0.9 mm to 2.1 mm, averaging (1.62 ± 0.45) mm. There is no significant statistical difference between registration errors of two groups (p > 0.05). Representative images of low-grade glioma(WHO II) and highgrade glioma (WHO III) are displayed in Figs. 2 and 3 which showed the enhancing region on CE T1-weighted (Figs. 2A and 3A),

Table 1 Comparison of DTI values in various ROIs between low grade and high grade glioma. ADC (10−3 mm2 /s)

FA

Low-grade High-grade

ROI i

ROI ii

ROI i

ROI ii

0.131 ± 0.056 0.127 ± 0.050

0.234 ± 0.097 0.264 ± 0.066

1.412 ± 0.664 1.250 ± 0.307

1.243 ± 0.228 1.349 ± 0.282

CP

Low-grade High-grade

CS

ROI i

ROI ii

ROI i

ROI ii

0.091 ± 0.003 0.106 ± 0.044

0.142 ± 0.043 0.164 ± 0.049

0.780 ± 0.082 0.754 ± 0.130

0.641 ± 0.113 0.595 ± 0.082

ROI i: the enhancing area. ROI ii: the peritumoral edema area.


G Model CLINEU-3499; No. of Pages 7 4


Fig. 2. A 57-year-old female with an anaplastic astrocytoma in the right frontal lobe, WHOIII. Contrast-enhanced T1-weighted (A) and b0 (B) images show enhancing lesion and extensive edema. ADC map (D) shows restricted diffusion of the enhancing part. C, E and F shows FA, CP and CS images, respectively.

extensive edema region on b0 images (Figs. 2B and 3B). The ADC map (Figs. 2D and 3D) demonstrated restricted diffusion of water molecules from the enhancing part and also exhibited low anisotropy relative to the normal white matter as proof of the three anisotropy maps in Figs. 2C, E and F and 3C, E and F. On visual inspection, it would be difficult to distinguish gliomas in case they have similar imaging characteristics on conventional CE T1-weighted, T2-weighted (b0) images and DTI maps. For each patient, the FA, ADC, CS and CP values in the immediate peritumoral edema and in enhancing region were calculated. The mean data for the two groups are summarized in Table 1.

3.1. Classification model generated from statistical analysis Univariate analyses were experimented for all parameters from each area to obtain the ones with high predictive capability (p < 0.30). The efficacy of each parameter was evaluated using ROC analysis as displayed in Table 2. FA from the immediate peritumoral area (AUC = 0.69) was the best single parameter for the discrimination, followed by CP (AUC = 0.63) and ADC (AUC = 0.62) in the same area. Later on, the selected four parameters (FA, ADC, CS and CP) from immediate peritumoral area were fully integrated into a multivariate logistic regression analysis by method of backward stepwise selection. There was a statistically significant difference in the multivariate logistic regression analysis (p = 0.02). The best LRM for the probability of high-grade gliomas was reached by three

parameters (CS, FA and ADC) from the immediate peritumoral area as follows: f (CS, FA, ADC) =

1 1 + exp(−(ˇ0 + ˇ1 CS + ˇ2 FA + ˇ3 ADC))

where ˇ0 = 91.55, ˇ1 = −107.95, ˇ2 = −118.37 and ˇ3 = 3.92. The Hosmer–Lemeshow test demonstrated rare departure from fit (p = 0.36). The ROC curves for individual parameter from the immediate peritumoral region and best LRM were displayed in Fig. 4. There seems to be little doubt that the LRM of the three parameters

Table 2 Sensitivity and specificity of imaging parameters in differentiation of low-grade from high-grade glioma using receiver operating characteristic curve analysis. Parameter

ROI

Cutoff value

Sensitivity

Specificity

AUC

FA

i ii

0.15 0.25

0.40 0.60

0.80 0.80

0.49 0.69

CS

i ii

0.89 0.43

0.13 1.00

1.00 0.10

0.46 0.30

CP

i ii

0.10 0.14

0.53 0.67

0.80 0.70

0.59 0.63

ADC

i ii

1.44 1.32

0.33 0.60

0.70 0.70

0.36 0.62

ROI i: the enhancing area. ROI ii: the peritumoral edema area.




5

Fig. 3. A 22-year-old male with an astrocytoma in the right frontal lobe, WHOII. Contrast-enhanced T1-weighted (A) and b0 (B) images show the similar enhancing lesion with extensive edema as shown in Fig. 2A and B. ADC map (D) also shows restricted diffusion of the enhancing part. C, E and F shows FA, CP and CS images, respectively.

(CS, FA and ADC) was more accurate than single one with a cutoff value = 0.58, sensitivity = 86.7%, specificity = 80% and AUC = 0.85. The AUC for combination of all four DTI parameters (FA, ADC, CS and CP) and commonly used parameters (CS and FA) from immediate peritumoral region were 0.81 and 0.76, respectively. Fig. 5 demonstrated a scatter plot of CS and FA of the immediate peritumoral region. It is shown that the diagonal solid line (cutoff line for the combined model of CS and FA) distinguished low grade from high grade glioma much better, while the vertical dotted line (cutoff line for CS) and the horizontal dashed line (cutoff line for FA) both have lower accuracy, which explained that FA alone was a poor predictor (AUC = 0.69), however, when combined with CS, the discrimination between high- and low-grade gliomas improved greatly (AUC = 0.76). 4. Discussion

Fig. 4. Receiver operative characteristic (ROC) curves for FA, CS, CP, ADC and CS + FA + ADC from the edema region of the tumor. The three parameters (CS, FA and ADC) combined were more accurate than single one with area under the curve (AUC) 0.81.

Conventional, breakdown of the blood–brain barrier (BBB) resulting in contrast-enhanced MR imaging is often associated with higher tumor grade while grade I and II may not have enhancement imaging with gadolinium. However, some researches described that contrast enhancement alone is not always accurate in predicting tumor grade [27,33,34]. In Ginsberg’s report [35], Shortage of enhancement in supratentorial gliomas does not indicate a lowgrade glioma. Knopp et al. [36] revealed that all low-grade tumors showed contrast enhancement as we found in our research. That is why we have to distinguish similar images since the gliomas WHO


G Model CLINEU-3499; No. of Pages 7 6


the diffusion ellipsoid is more planar-like due to its diffusion in a plane spanned. When CS increased, the shaped of the tensor is more sphere-like and no preferred diffusion direction exists. As proved by our research, the diffusion tensor shape parameters of DTI can provide additional spatial information to better classify the gliomas. Other researchers got the similar conclusion. Kumar in 2007 confirmed the DTI morphological parameters of CL, CP had the advantage for discriminating true from pseudo fiber inside abscess cavity [37]. Wang et al. [24] in 2009 studied 38 cases of glioblastoma, 25 patients with brain metastases and conclude that CP used individually or combined, has the potential as non-invasive measure to differentiate two tumors. Taken together, it revealed that diffusion tensor shape parameters of DTI are playing an important role in histopathological study of intracranial tumors [24,37]. As the associated indices of the predicted probability and observation value, the area under the ROC curve (ranged from 50 to 100%) is a valuable tool for assessment of accuracy of diagnostic test. Less than 50% demonstrated low efficacy. Based on our research, when FA, CS and ADC of immediate peritumoral area combined together, the largest curve area (AUC = 0.81) and the best classification ability (86.7% sensitivity and 80% specificity) were achieved. Further studies are still necessary on correlation between DTI metrics values and histological change such as the density of tumor cell and blood vessel, edema, the size of tumor cell. 5. Conclusion Fig. 5. Scatter plot of CS and FA from the peritumoral region of low-grade (triangle) and high-grade glioma (dot). The vertical dotted line is the cutoff line for CS (cutoff value = 0.43 as shown in Table 2), which is inadequate for discrimination of two gliomas. The horizontal dashed line is the cutoff line for FA (cutoff value = 0.25 as in Table 2), which is better than CS. The diagonal solid line is the cutoff line for the combined model of CS and FA, which distinguish low grade from high grade glioma much better than CS or FA alone.

Grade I–IV recruited in this study demonstrate different contrast enhancement. As a result of a careful selection of gliomas with similar enhancement patterns and limitations of retrospective nature, less Grade I and Grade III are included. Therefore, we only divide gliomas into two groups. Severity of infiltration is the most fundamental characteristics to distinguish low-grade and high grade. It is well known that microinvasive tumor cells are not detected on conventional MR imaging, which can only display the rough scope of tumor core and peritumoral edema rather than quantified infiltrative tumor cells resulting in poor prognoses for patients diagnosed with high-grade gliomas. Other non-invasive diagnoses methods are thus needed before operation. We analyzed the correlation of the DTI parameters including FA, ADC, CP and CS with glioma classification using univariate and multivariate logistic regression. The results showed that multiparameter logistic model combined the immediate peritumoral edema FA, CS and ADC value has better differential diagnosis accuracy, which is beneficial for the preoperative non-invasive diagnosis of gliomas. Furthermore, our study confirmed the feasibility of using multiple DTI parameters in the differential diagnosis. We found that each parameter alone could not contribute to differentiate low-grade from high-grade gliomas. Despite of their slight difference in each region, we did not regard FA and ADC the same statistical significance as a few researchers reported (p < 0.05) [5,6,8]. FA and ADC in our opinion may provide only partial information about the movement of water molecules and may also be affected by various factors such as the density and composition of tumor cells, edema, tumor necrosis, etc., resulting in inability to discriminate gliomas accurately. Some others found the same [10–17]. In our study, we also included two diffusion tensor shape parameters (CP and CS) as ellipsoid shape descriptors. When CP increased,

Combined DTI metrics accomplished quantitative analysis on glioma grading and severity of infiltration because it can provide metabolic and functional information on cellular and molecular level. This classification is a preoperatively diagnosis without biopsy, which could to some extent decrease uncertainty of biopsy diagnosis due to sampling location and depth. It could therefore be concluded that combined DTI metrics have the potential as a new and promising tool of non-invasive diagnosis to distinguish between high-grade and low-grade glioma. Funding This work was supported by the Science and Technology Commission of the Shanghai Municipality of China (Grant No. 10dz2211800) and Shanghai Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention (MICCAI). The authors have no personal financial or institutional interest in any of the drugs, materials, or devices described in this article. References [1] Hadziahmetovic M, Shirai K, Chakravarti A. Recent advancements in multimodality treatment of gliomas. Future Oncol 2011;7:1169–83. [2] Van Gool S, Maes W, Ardon H, Verschuere T, Van Cauter S, De Vleeschouwer S. Dendritic cell therapy of high-grade gliomas. Brain Pathol 2009;19:694–712. [3] Stupp R, Hegi ME, Mason WP, van den Bent MJ, Taphoorn MJ, Janzer RC, et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol 2009;10:459–66. [4] Sartor K. MR imaging of the brain: tumors. Eur Radiol 1999;9:1047–54. [5] Beppu T, Inoue T, Shibata Y, Yamada N, Kurose A, Ogasawara K, et al. Fractional anisotropy value by diffusion tensor magnetic resonance imaging as a predictor of cell density and proliferation activity of glioblastomas. Surg Neurol 2005;63:56–61. [6] Lee EJ, Lee SK, Agid R, Bae JM, Keller A, Terbrugge K. Preoperative grading of presumptive low-grade astrocytomas on MR imaging: diagnostic value of minimum apparent diffusion coefficient. AJNR Am J Neuroradiol 2008;29:1872–7. [7] Talos I, Mian A, Zou K, Hsu L, Goldberg-Zimring D, Haker S, et al. Magnetic resonance and the human brain: anatomy, function and metabolism. Cell Mol Life Sci 2006;63:1106. [8] Kinoshita M, Hashimoto N, Goto T, Kagawa N, Kishima H, Izumoto S, et al. Fractional anisotropy and tumor cell density of the tumor core show positive correlation in diffusion tensor magnetic resonance imaging of malignant brain tumors. NeuroImage 2008;43:29–35.




[9] Stadlbauer A, Ganslandt O, Buslei R, Hammen T, Gruber S, Moser E, et al. Gliomas: histopathologic evaluation of changes in directionality and magnitude of water diffusion at diffusion-tensor MR imaging. Radiology 2006;240:803–10. [10] Inoue T, Ogasawara K, Beppu T, Ogawa A, Kabasawa H. Diffusion tensor imaging for preoperative evaluation of tumor grade in gliomas. Clin Neurol Neurosurg 2005;107:174–80. [11] Beppu T, Inoue T, Shibata Y, Kurose A, Arai H, Ogasawara K, et al. Measurement of fractional anisotropy using diffusion tensor MRI in supratentorial astrocytic tumors. J Neurooncol 2003;63:109–16. [12] Goebell E, Paustenbach S, Vaeterlein O, Ding XQ, Heese O, Fiehler J, et al. Low-grade and anaplastic gliomas: differences in architecture evaluated with diffusion-tensor MR imaging. Radiology 2006;239:217–22. [13] Wieshmann UC, Clark CA, Symms MR, Franconi F, Barker GJ, Shorvon SD. Reduced anisotropy of water diffusion in structural cerebral abnormalities demonstrated with diffusion tensor imaging. Magn Reson Imaging 1999;17:1269–74. [14] Guo AC, Cummings TJ, Dash RC, Provenzale JM. Lymphomas, high-grade astrocytomas: comparison of water diffusibility and histologic characteristics. Radiology 2002;224:177–83. [15] Kono K, Inoue Y, Nakayama K, Shakudo M, Morino M, Ohata K, et al. The role of diffusion-weighted imaging in patients with brain tumors. AJNR Am J Neuroradiol 2001;22:1081–8. [16] Lu S, Ahn D, Johnson G, Law M, Zagzag D, Grossman RI. Diffusiontensor MR imaging of intracranial neoplasia and associated peritumoral edema: introduction of the tumor infiltration index. Radiology 2004;232: 221–8. [17] Zimmerman RD. Is there a role for diffusion-weighted imaging in patients with brain tumors or is the bloom off the rose. AJNR Am J Neuroradiol 2001;22:1013–4. [18] Alexander AL, Hasan K, Kindlmann G, Parker DL, Tsuruda JS. A geometric analysis of diffusion tensor measurements of the human brain. Magn Reson Med 2000;44:283–91. [19] Kim S, Pickup S, Hsu O, Poptani H. Diffusion tensor MRI in rat models of invasive and well-demarcated brain tumors. NMR Biomed 2008;21:208–16. [20] Westin CF, Peled S, Gubjartsson H, Kikinis R. Geometrical diffusion measures for MRI from tensor basis analysis. In: Proceedings of ISMRM. 1997. [21] Westin CF, Maier SE, Mamata H, Nabavi A, Jolesz FA, Kikinis R. Processing and visualization for diffusion tensor MRI. Med Image Anal 2002;6:93–108. [22] Provenzale JM, McGraw P, Mhatre P, Guo AC, Delong D. Peritumoral brain regions in gliomas and meningiomas: investigation with isotropic diffusion-weighted MR imaging and diffusion-tensor MR imaging. Radiology 2004;232:451–60.

7

[23] Santhosh K, Thomas B, Radhakrishnan VV, Saini J, Kesavadas C, Gupta AK, et al. Diffusion tensor and tensor metrics imaging in intracranial epidermoid cysts. J Magn Reson Imaging 2009;29:967–70. [24] Wang S, Kim S, Chawla S, Wolf RL, Zhang W-G, O’Rourke DM, et al. Differentiation between glioblastomas and solitary brain metastases using diffusion tensor imaging. NeuroImage 2009;44:653–60. [25] Kim JH, Chang KH, Na DG, Song IC, Kwon BJ, Han MH, et al. 3T 1H-MR spectroscopy in grading of cerebral gliomas: comparison of short and intermediate echo time sequences. AJNR Am J Neuroradiol 2006;27:1412–8. [26] Al-Okaili RN, Krejza J, Woo JH, Wolf RL, O’Rourke DM, Judy KD, et al. Intraaxial brain masses: MR imaging–based diagnostic strategy—initial experience. Radiology 2007;243:539–50. [27] Jenkinson M, Smith S. A global optimisation method for robust affine registration of brain images. Med Image Anal 2001;5:143–56. [28] Melhem ER, Mori S, Mukundan G, Kraut MA, Pomper MG, van Zijl PC. Diffusion tensor MR imaging of the brain and white matter tractography. AJR Am J Roentgenol 2002;178:3–16. [29] Ito R, Mori S, Melhem ER. Diffusion tensor brain imaging and tractography. Neuroimaging Clin N Am 2002;12:1–19. [30] Li L, Liu YX, Song ZJ. Three-dimensional reconstruction of registered and fused Chinese visible human and patient MRI images. Clin Anat 2006;19:225–31. [31] Lemeshow S, Hosmer DJ. A review of goodness of fit statistics for use in the development of logistic regression models. Am J Epidemiol 1982;115:92–106. [32] Mihara F, Numaguchi Y, Rothman M, Kristt D, Fiandaca M, Swallow L. Nonenhancing supratentorial malignant astrocytomas: MR features and possible mechanisms. Radiat Med 1995;13:11. [33] Scott J, Brasher P, Sevick R, Rewcastle N, Forsyth P. How often are nonenhancing supratentorial gliomas malignant. A population study. Neurology 2002;59:947–9. [34] Law M, Yang S, Babb JS, Knopp EA, Golfinos JG, Zagzag D, et al. Comparison of cerebral blood volume and vascular permeability from dynamic susceptibility contrast-enhanced perfusion MR imaging with glioma grade. AJNR Am J Neuroradiol 2004;25:746–55. [35] Ginsberg LE, Fuller GN, Hashmi M, Leeds NE, Schomer DF. The significance of lack of MR contrast enhancement of supratentorial brain tumors in adults: histopathological evaluation of a series. Surg Neurol 1998;49:436–40. [36] Knopp EA, Cha S, Johnson G, Mazumdar A, Golfinos JG, Zagzag D, et al. Glial neoplasms: dynamic contrast-enhanced T2*-weighted MR imaging. Radiology 1999;211:791–8. [37] Kumar M, Gupta RK, Nath K, Rathore RKS, Bayu G, Trivedi R, et al. Can we differentiate true white matter fibers from pseudofibers inside a brain abscess cavity using geometrical diffusion tensor imaging metrics. NMR Biomed 2008;21:581–8.


Diffusion tensor imaging of the anterior cruciate ligament graft after reconstruction: repeatability and diffusion tensor imaging metrics.

Differentiation between Subtypes of Primary Progressive Aphasia by Using Cortical Thickness and Diffusion-Tensor MR Imaging Measures.

Accelerated human cardiac diffusion tensor imaging using simultaneous multislice imaging.

Central Motor Conduction Time and diffusion tensor imaging metrics in children with complex motor disorders.

Linear associations between clinically assessed upper motor neuron disease and diffusion tensor imaging metrics in amyotrophic lateral sclerosis.

Combined intravoxel incoherent motion and diffusion tensor imaging of renal diffusion and flow anisotropy.

Bimanual motor deficits in older adults predicted by diffusion tensor imaging metrics of corpus callosum subregions.

Parallel imaging and compressed sensing combined framework for accelerating high-resolution diffusion tensor imaging using inter-image correlation.

Anisotropic fractional diffusion tensor imaging.

Discrimination between Lumbar Intraspinal Stenosis and Foraminal Stenosis using Diffusion Tensor Imaging Parameters: Preliminary Results.

REDUCING CSF PARTIAL VOLUME EFFECTS TO ENHANCE DIFFUSION TENSOR IMAGING METRICS OF BRAIN MICROSTRUCTURE.

Discrimination between glioblastoma multiforme and solitary metastasis using morphological features derived from the p:q tensor decomposition of diffusion tensor imaging.

Partial correlation analyses of global diffusion tensor imaging-derived metrics in glioblastoma multiforme: Pilot study.

Visualization of the medial forebrain bundle using diffusion tensor imaging.

Differentiation of the Infarct Core from Ischemic Penumbra within the First 4.5 Hours, Using Diffusion Tensor Imaging-Derived Metrics: A Rat Model.

Voxel-based clustered imaging by multiparameter diffusion tensor images for glioma grading.

Diffusion Tensor Imaging for Glioma Grading: Analysis of Fiber Density Index.

Cerebral Microstructural Alterations after Radiation Therapy in High-Grade Glioma: A Diffusion Tensor Imaging-Based Study.

Usefulness of diffusion tensor imaging for the differentiation between low-fat angiomyolipoma and clear cell carcinoma of the kidney.

Diffusion tensor image registration using hybrid connectivity and tensor features.

Combined Diffusion Tensor Imaging and Arterial Spin Labeling as Markers of Early Parkinson's disease.

Longitudinal diffusion tensor imaging in frontotemporal dementia.

Peripheral neuropathy: detection with diffusion-tensor imaging.

Diffusion tensor imaging in hemorrhagic stroke.