**3. Diffusion magnetic resonance imaging**

The diffusion phenomenon describes quantitatively the random (Brownian) molecular motion translations of water molecules originating ultimately from thermal energy. In a tissue, like the brain, the mechanism of diffusion involves mainly the motion of water molecules on a scale of 5 to 15 micrometers over the approximate 40 ms of measurement time. It is described by an Apparent Diffusion Coefficient (ADC), representing the average translational displacement of a water molecule during a time unit. The diffusion process has a vectorial nature, involving magnitude and direction. Its magnitude and direction in vivo depend on a variety of factors including permeability barriers and their spacing, microviscosity of the intracellular milieu, molecular obstructions to water displacements, duration of the diffusion measurement and, eventually the bulk flow within capillaries and water active transporters in tissue. These circumstances make the interpretation of diffusion measurements complex, but at the same time highly informative on all the aspects influencing it.

In vivo Diffusion Weighted Imaging (DWI) provides images with an inherent contrast different from that obtained by conventional structural MR techniques. In this sense, DWI gives unique information about the microstructure and viability of brain tissue, with the signal intensity of the image depending on the average translational motion of water molecules, an aspect known to be appreciably modified in cerebral tumors (Schaefer, Grant et al. 2000).

#### **3.1 Apparent Diffusion Coefficient**

The first description of a diffusion study was provided by Stejskal and Tanner, who used a spin-echo acquisition protocol in which two gradient pulses equal in magnitude but

In vivo measurements of ADC in the human brain by using MRI are well documented both in physiological and physiopathological conditions (Le Bihan, Breton et al. 1986; Thomsen, Henriksen et al. 1987; Chien, Buxton et al. 1990; Hajnal, Doran et al. 1991; Schaefer, Grant et al. 2000). Regions of increased or decreased diffusion due to pathology can be clearly identified in a diffusion weighted MRI study by the corresponding changes in intensity,

Although DWI studies were initially focused on stroke (Warach, Chien et al. 1992), multiple sclerosis (Larsson, Thomsen et al. 1992) and hydrocephalus (Gideon, Thomsen et al. 1994),

In 1994 Tien et al. acquired diffusion images in patients with high-grade gliomas using Echo Planar Imaging (EPI) techniques to measure ADC along cephalocaudal axis, comparing the obtained results with those found by using conventional (SE) or fast spin-echo (FSE) MRI (Tien, Felsberg et al. 1994). The ADC images enabled these authors to differentiate various regions within the tumor, concluding that echo-plannar DWI appeared a more powerful

In 1995 Brunberg et al. reported their study performed with a motion-insensitive spin-echo sequence measuring ADC along three orthogonal axes (Brunberg, Chenevert et al. 1995). They aimed to determine if there was a relationship between water diffusion coefficients or its anisotropy in MR-defined regions of normal and abnormal parenchyma in patients with cerebral gliomas. They were able to distinguish between normal white matter, cystic or necrotic areas, regions of edema and solid enhancing tumor by comparing anisotropic ADC

In 1997 Krabbe et al. measured ADC in different regions of patients with brain tumors (Krabbe, Gideon et al. 1997). They reported diffusion values in contrast-enhancing tumor, non-enhancing tumor, cyst or necrosis, edema and cerebro-spinal fluid (CSF) for malignant gliomas, low–grade gliomas, metastases, meningiomas and cerebral abscess. They found that mean ADC in contrast-enhancing tumor was significantly higher in metastases than in high-grade gliomas, while no significant differences were found between ADC in contrastenhancing tissue in meningiomas and metastases or high-grade gliomas. The highest ADC was always found in CSF. ADC in edema was higher than in contrast-enhancing tumor, and in edema around cerebral metastases was significantly higher than those around high-grade gliomas. These findings could help to distinguish on this basis high-grade gliomas and

After these initial reports, a large number of results have been published demonstrating the vast possibilities of diffusion MRI in brain oncology. These cover the applications of DWI from glioma characterization and evaluation to the identification of patients with poor treatment response or tumor recurrence (Provenzale, Mukundan et al. 2006; Hamstra, Rehemtulla et al. 2007; Hamstra, Galban et al. 2008). In addition, DWI imaging may allow the non invasive grading of tumor cellularity since cells constitute barriers that restrict microscopically the motion of the water molecules within the tissue. In this sense, the diffusion of water molecules across the tumor as compared to the diffusion in normal brain tissue is expected be different, depicting different diffusivity and ADC depending on the tumor cellularity and grade. As cells constitute a barrier to water motion, tumors with higher cellularity are expected to show lower ADC values than less cellularly crowded

either in the diffusion weighted images or in fitted ADC maps.

an increasing number of diffusion studies on brain is being performed.

tool in the characterization of brain neoplasm than conventional techniques.

**3.2 Diffusion weighted imaging of gliomas** 

values in those regions.

metastases prior to the surgery.

opposite in directions (figure 2) were intercalated before and after the 180 degrees refocusing pulse, thus allowing the measurement of molecular water motions in the direction of the applied gradient (Stejskal and Tanner 1965).

Fig. 2. Diffusion Weighted MR Imaging. Left: The Stejskal-Tanner pulse sequence. The basic spin-echo sequence is sensitized to diffusion using two phase gradients of duration , separated by the diffusional time , located symmetrically before and after the 180 pulse. Center: Influence of cellularity on the translational diffusion measurements. Note that in the same time, a water molecule will move translationally a shorter path in an increased cellular density and molecularly more crowded medium. A representative image of the ADC measured in the normal rat brain (Pacheco-Torres, Calle et al.). Reproduced with permission

Because of the same magnitude and the opposite directions of these two gradients, the MR signal of water molecules without net motion will be identical to its intensity as obtained without diffusion gradients in the same cerebral location. However, for locations showing net translational motion of water molecules in the direction of the applied gradient, the opposing effects of the two gradients will not be cancelled, and the resulting image will depict lower intensity in these regions. Signal intensity will be smaller, the faster the average movement of water molecules or the larger ADC. This leads to a DW image depicting hypointensity in the tissue voxels containing faster moving water molecules and hiperintensity in those with slower water displacements.

Under these conditions, the MR signal loss is related to: (i) the diffusion coefficient of the water and (ii) the duration and strength of the magnetic field gradients used in the measurement. The following equation applies:

$$\text{SI} = \text{SI}\_0 \times \exp\left(\cdot b \ge \text{ADC}\right) \tag{1}$$

where; SI0 represents the signal intensity for each pixel in the absence of gradient pulses, *b* is an experimentally modifiable parameter including the duration , intensity and delay separation of the applied magnetic field gradients and ADC is the apparent diffusion coefficient of the water in the investigated tissue voxel. Using images acquired with increasing *b* values and thus, increasingly sensitized to diffusion, a parametric map can be calculated by plotting for every pixel the linear relationship between the natural logarithm of the relative signal loss (SI/SI0) versus *b*. This affords the calculation of the corresponding ADC on a pixel-by-pixel base, representing then every pixel value in the resulting ADC map.

opposite in directions (figure 2) were intercalated before and after the 180 degrees refocusing pulse, thus allowing the measurement of molecular water motions in the

Fig. 2. Diffusion Weighted MR Imaging. Left: The Stejskal-Tanner pulse sequence. The basic spin-echo sequence is sensitized to diffusion using two phase gradients of duration , separated by the diffusional time , located symmetrically before and after the 180 pulse. Center: Influence of cellularity on the translational diffusion measurements. Note that in the same time, a water molecule will move translationally a shorter path in an increased cellular density and molecularly more crowded medium. A representative image of the ADC measured in the normal rat brain (Pacheco-Torres, Calle et al.).

Because of the same magnitude and the opposite directions of these two gradients, the MR signal of water molecules without net motion will be identical to its intensity as obtained without diffusion gradients in the same cerebral location. However, for locations showing net translational motion of water molecules in the direction of the applied gradient, the opposing effects of the two gradients will not be cancelled, and the resulting image will depict lower intensity in these regions. Signal intensity will be smaller, the faster the average movement of water molecules or the larger ADC. This leads to a DW image depicting hypointensity in the tissue voxels containing faster moving water molecules and

Under these conditions, the MR signal loss is related to: (i) the diffusion coefficient of the water and (ii) the duration and strength of the magnetic field gradients used in the

where; SI0 represents the signal intensity for each pixel in the absence of gradient pulses, *b* is an experimentally modifiable parameter including the duration , intensity and delay separation of the applied magnetic field gradients and ADC is the apparent diffusion coefficient of the water in the investigated tissue voxel. Using images acquired with increasing *b* values and thus, increasingly sensitized to diffusion, a parametric map can be calculated by plotting for every pixel the linear relationship between the natural logarithm of the relative signal loss (SI/SI0) versus *b*. This affords the calculation of the corresponding ADC on a pixel-by-pixel base, representing then every pixel value in the

SI = SI0 x exp (-*b* x ADC) (1)

direction of the applied gradient (Stejskal and Tanner 1965).

hiperintensity in those with slower water displacements.

measurement. The following equation applies:

Reproduced with permission

resulting ADC map.

In vivo measurements of ADC in the human brain by using MRI are well documented both in physiological and physiopathological conditions (Le Bihan, Breton et al. 1986; Thomsen, Henriksen et al. 1987; Chien, Buxton et al. 1990; Hajnal, Doran et al. 1991; Schaefer, Grant et al. 2000). Regions of increased or decreased diffusion due to pathology can be clearly identified in a diffusion weighted MRI study by the corresponding changes in intensity, either in the diffusion weighted images or in fitted ADC maps.

#### **3.2 Diffusion weighted imaging of gliomas**

Although DWI studies were initially focused on stroke (Warach, Chien et al. 1992), multiple sclerosis (Larsson, Thomsen et al. 1992) and hydrocephalus (Gideon, Thomsen et al. 1994), an increasing number of diffusion studies on brain is being performed.

In 1994 Tien et al. acquired diffusion images in patients with high-grade gliomas using Echo Planar Imaging (EPI) techniques to measure ADC along cephalocaudal axis, comparing the obtained results with those found by using conventional (SE) or fast spin-echo (FSE) MRI (Tien, Felsberg et al. 1994). The ADC images enabled these authors to differentiate various regions within the tumor, concluding that echo-plannar DWI appeared a more powerful tool in the characterization of brain neoplasm than conventional techniques.

In 1995 Brunberg et al. reported their study performed with a motion-insensitive spin-echo sequence measuring ADC along three orthogonal axes (Brunberg, Chenevert et al. 1995). They aimed to determine if there was a relationship between water diffusion coefficients or its anisotropy in MR-defined regions of normal and abnormal parenchyma in patients with cerebral gliomas. They were able to distinguish between normal white matter, cystic or necrotic areas, regions of edema and solid enhancing tumor by comparing anisotropic ADC values in those regions.

In 1997 Krabbe et al. measured ADC in different regions of patients with brain tumors (Krabbe, Gideon et al. 1997). They reported diffusion values in contrast-enhancing tumor, non-enhancing tumor, cyst or necrosis, edema and cerebro-spinal fluid (CSF) for malignant gliomas, low–grade gliomas, metastases, meningiomas and cerebral abscess. They found that mean ADC in contrast-enhancing tumor was significantly higher in metastases than in high-grade gliomas, while no significant differences were found between ADC in contrastenhancing tissue in meningiomas and metastases or high-grade gliomas. The highest ADC was always found in CSF. ADC in edema was higher than in contrast-enhancing tumor, and in edema around cerebral metastases was significantly higher than those around high-grade gliomas. These findings could help to distinguish on this basis high-grade gliomas and metastases prior to the surgery.

After these initial reports, a large number of results have been published demonstrating the vast possibilities of diffusion MRI in brain oncology. These cover the applications of DWI from glioma characterization and evaluation to the identification of patients with poor treatment response or tumor recurrence (Provenzale, Mukundan et al. 2006; Hamstra, Rehemtulla et al. 2007; Hamstra, Galban et al. 2008). In addition, DWI imaging may allow the non invasive grading of tumor cellularity since cells constitute barriers that restrict microscopically the motion of the water molecules within the tissue. In this sense, the diffusion of water molecules across the tumor as compared to the diffusion in normal brain tissue is expected be different, depicting different diffusivity and ADC depending on the tumor cellularity and grade. As cells constitute a barrier to water motion, tumors with higher cellularity are expected to show lower ADC values than less cellularly crowded

This is probably related with the fact that there is an increase in intracellular water in edematous white matter as well as within the tumor. Although this is true for almost all gliomas, its important to consider here, that interpretation of diffusion images and ADC values for diagnostic and prognostic purposes may be tumor type and individual-

Water molecules are not able to diffuse with equal freedom in all directions through the brain, as surrounding cellular membranes and tissue structures limit to different extents their translational mobility in different directions. As a result, the ADC values measured in different directions are not equal and cerebral diffusion becomes anisotropic. This occurs mainly in white matter since molecular diffusion is relatively unrestricted along the trajectory of the neuronal axons and nerve fibres, but significantly more restricted in the perpendicular plane. The ADC in the grey matter, containing mainly neuronal bodies is more isotropic, with similar restrictions in all directions (Hajnal, Doran et al. 1991). This anisotropic behaviour of the molecular motions of water molecules in the brain can be observed by MRI obtaining images with the diffusion gradients applied in different spatial directions. The dominant direction and magnitude of water diffusion in every pixel is then obtained by calculating the resulting tensor from the ADC vectors measured in all investigated directions. This methodology is known as Diffusion Tensor Imaging (DTI), a method requiring the sampling of the ADC in six or more directions (Nucifora, Verma et al. 2007). This relatively complex mathematical modelling approach allows the preparation of three-dimensional ADC anisotropy maps, a very potent tool in the examination of the brain microstructure, inaccessible to other neuroimaging modalities. Using DTI sequences, it is possible to explore the integrity of white matter structures and detect abnormalities not visible in the conventional MRI acquisitions. DTI also allows to calculate tractography maps, revealing white matter fiber tracts, very useful to determine the infiltration or degeneration of the original fiber tracts by invading tumors (Mori and van Zijl 2002; Witwer, Moftakhar et

Diffusion properties assessed by DTI, such as the fractional anisotropy (FA) and average of diffusion coefficient or mean diffusivity (MD), have become nowadays the most frequently used parametric images in neuro-oncology. Such images are correlated with cellular physiology and tissue microstructure, being extensively used to delineate glioma grades. MD measurements allow differentiation of necrotic areas within a high-grade gliomas and peritumoral edema from adjacent white matter; but they are not able to assess tumor infiltration adequately (Tropine, Vucurevic et al. 2004). On the other hand, FA values can be associated to functional tumoral characteristics, including WHO grade. So, combination of DTI related parameters in a multiparametric approach may facilitate glioma characterization and classification. In this respect, several reports aimed, at the precise elucidation of individual anisotropy patterns of brain water diffusivity which can reflect histopathological features used to graded gliomas (Jakab, Molnar et al.). Jakab et al. reported that these parameters can be used as indicators of glioma classification types and tumor physiopathology features like proliferation, metabolism or blood flow. These authors

dependent.

al. 2002).

**3.3.1 Fractional Anisotropy and Mean Diffusivity** 

**3.3 Diffusion Tensor Imaging** 

tumors (Sugahara, Korogi et al. 1999; Gauvain, McKinstry et al. 2001). On the other hand, an increase in ADCs usually is correlated with decreased cellularity, revealing successful therapy and/or induced necrosis (Lyng, Haraldseth et al. 2000). Although some authors have reported lower diffusivity of water in high-grade than in low grade gliomas (Bulakbasi, Kocaoglu et al. 2003), it should be mentioned that there is a considerable overlap between ADCs in both types of neoplasm (Kono, Inoue et al. 2001). Even so, there is clearly an inverse correlation between tumor cellularity and ADC values measured in brain tumors allowing to distinguish these from other intracranial lesions. Figure 3 illustrates the important differences found in ADC of intratumoral and extratumoral regions, in two different neoplasm types. Images (c) depict ADC maps in two patients with different brain tumor types. Upper panel shows a high average ADC in the patient with partially resected GBM, while in lower pannel reveals a low average ADC in a patient with a primary CNS lymphoma. Regions of white matter edema distal to enhancing tumor margins usually show diffusion coefficient values significantly different from normal white and grey matter, and from areas of tumor cyst or necrosis, respectively. Notably, there are not appreciable differences between these ADCs from edema regions and those from enhancing tumor.

Fig. 3. *Upper panels*: A patient with GBM and high ADC values. (a) Axial contrast-enhanced T1W image shows a mass; (b) T2W image shows lesion and edema; (c) ADC map shows that mean ADC in tumor is much higher than in normal brain. *Lower panels*: A patient with primary CNS lymphoma and low ADC values. (a) Axial contrast-enhanced T1W image shows non enhancing lesion; (b) T2W image shows signal intensity in tumor darker than gray matter; (c) ADC map shows mean ADC in tumor is lower than in normal tissue. (http://radiopaedia.org/encyclopaedia/cases) Reproduced with permission.

tumors (Sugahara, Korogi et al. 1999; Gauvain, McKinstry et al. 2001). On the other hand, an increase in ADCs usually is correlated with decreased cellularity, revealing successful therapy and/or induced necrosis (Lyng, Haraldseth et al. 2000). Although some authors have reported lower diffusivity of water in high-grade than in low grade gliomas (Bulakbasi, Kocaoglu et al. 2003), it should be mentioned that there is a considerable overlap between ADCs in both types of neoplasm (Kono, Inoue et al. 2001). Even so, there is clearly an inverse correlation between tumor cellularity and ADC values measured in brain tumors allowing to distinguish these from other intracranial lesions. Figure 3 illustrates the important differences found in ADC of intratumoral and extratumoral regions, in two different neoplasm types. Images (c) depict ADC maps in two patients with different brain tumor types. Upper panel shows a high average ADC in the patient with partially resected GBM, while in lower pannel reveals a low average ADC in a patient with a primary CNS lymphoma. Regions of white matter edema distal to enhancing tumor margins usually show diffusion coefficient values significantly different from normal white and grey matter, and from areas of tumor cyst or necrosis, respectively. Notably, there are not appreciable differences between these ADCs from edema regions and those from enhancing tumor.

(a) (b) (c)

Fig. 3. *Upper panels*: A patient with GBM and high ADC values. (a) Axial contrast-enhanced T1W image shows a mass; (b) T2W image shows lesion and edema; (c) ADC map shows that mean ADC in tumor is much higher than in normal brain. *Lower panels*: A patient with primary CNS lymphoma and low ADC values. (a) Axial contrast-enhanced T1W image shows non enhancing lesion; (b) T2W image shows signal intensity in tumor darker than gray matter; (c) ADC map shows mean ADC in tumor is lower than in normal tissue. (http://radiopaedia.org/encyclopaedia/cases) Reproduced with permission.

This is probably related with the fact that there is an increase in intracellular water in edematous white matter as well as within the tumor. Although this is true for almost all gliomas, its important to consider here, that interpretation of diffusion images and ADC values for diagnostic and prognostic purposes may be tumor type and individualdependent.

#### **3.3 Diffusion Tensor Imaging**

Water molecules are not able to diffuse with equal freedom in all directions through the brain, as surrounding cellular membranes and tissue structures limit to different extents their translational mobility in different directions. As a result, the ADC values measured in different directions are not equal and cerebral diffusion becomes anisotropic. This occurs mainly in white matter since molecular diffusion is relatively unrestricted along the trajectory of the neuronal axons and nerve fibres, but significantly more restricted in the perpendicular plane. The ADC in the grey matter, containing mainly neuronal bodies is more isotropic, with similar restrictions in all directions (Hajnal, Doran et al. 1991). This anisotropic behaviour of the molecular motions of water molecules in the brain can be observed by MRI obtaining images with the diffusion gradients applied in different spatial directions. The dominant direction and magnitude of water diffusion in every pixel is then obtained by calculating the resulting tensor from the ADC vectors measured in all investigated directions. This methodology is known as Diffusion Tensor Imaging (DTI), a method requiring the sampling of the ADC in six or more directions (Nucifora, Verma et al. 2007). This relatively complex mathematical modelling approach allows the preparation of three-dimensional ADC anisotropy maps, a very potent tool in the examination of the brain microstructure, inaccessible to other neuroimaging modalities. Using DTI sequences, it is possible to explore the integrity of white matter structures and detect abnormalities not visible in the conventional MRI acquisitions. DTI also allows to calculate tractography maps, revealing white matter fiber tracts, very useful to determine the infiltration or degeneration of the original fiber tracts by invading tumors (Mori and van Zijl 2002; Witwer, Moftakhar et al. 2002).

#### **3.3.1 Fractional Anisotropy and Mean Diffusivity**

Diffusion properties assessed by DTI, such as the fractional anisotropy (FA) and average of diffusion coefficient or mean diffusivity (MD), have become nowadays the most frequently used parametric images in neuro-oncology. Such images are correlated with cellular physiology and tissue microstructure, being extensively used to delineate glioma grades. MD measurements allow differentiation of necrotic areas within a high-grade gliomas and peritumoral edema from adjacent white matter; but they are not able to assess tumor infiltration adequately (Tropine, Vucurevic et al. 2004). On the other hand, FA values can be associated to functional tumoral characteristics, including WHO grade. So, combination of DTI related parameters in a multiparametric approach may facilitate glioma characterization and classification. In this respect, several reports aimed, at the precise elucidation of individual anisotropy patterns of brain water diffusivity which can reflect histopathological features used to graded gliomas (Jakab, Molnar et al.). Jakab et al. reported that these parameters can be used as indicators of glioma classification types and tumor physiopathology features like proliferation, metabolism or blood flow. These authors

Fig. 5. T1W (upper) and relative anisotropy maps (lower). Each color indicates the principal eigenvector. Color intensity correlates with anisotropy grade. A: right frontal brain tumor; B:

integrity in both cases. (Lu, Ahn et al. 2003; Kinoshita, Yamada et al. 2005). Reproduced with

Gliomas are infiltrating tumors in which the tumor cells may invade cerebral structures following the path of normal fiber tracts, making it difficult to differentiate tumoral from healthy tissue and detect the extension of white matter infiltration. Although several approaches from DTI have been reported to make the detection and reliable quantification of this diffuse tumor infiltration (Price, Burnet et al. 2003; Tropine, Vucurevic et al. 2004), this aspect still represents a very challenging question. One of the best developments to improve this situation is based in the visualization of diffusion tensor field by ellipsoids and colour coding of the main diffusion orientations (MDO) that allows the assessment of fibre bundles infiltration more reliably (Schluter, Stieltjes et al. 2005). The color maps and diffusion ellipsoids permit the identification of the particular fibre tracts bundles leading to the periphery of glioma by a reduced FA. Obviously, this approach is helping to improve therapeutic decisions and to minimize interventional risk. Additionally, it has been found that tumor recurrence has the tendency to occur along directions where the region of abnormal isotropic diffusion spreads beyond the region of abnormal anisotropic diffusion. On these grounds, it is eventually possible to predict regions of potential tumor infiltration and/or recidive by detecting abnormalities with high spatial resolution by using DTI. An aspect of increasing interest is the development of new applications that allow obtaining not only presurgical but also postsurgical information related to the tumor affected tissue for a

anaplastic ependymoma. In lower panels is clearly detected the loss of white matter

permission

**3.3.2 Tractography** 

more accurate postoperative prognosis.

proved a consistent relationship between DTI-related information and histopathological data obtained from biopsies.

FA and MD are very sensitive indicators of white matter integrity also. Some clear examples of color-based images obtained from DTI acquisition are shown in figure 4 and 5, where not only edema region can be clearly distinguished from solid tumor by MD and FA color-based maps (figure 4), but also anisotropic directional maps depict those regions of fiber tract degeneration.

The diffusion MRI study of a brain tumor patient may show a lack of a significant difference in mean diffusion anisotropy between cerebral cortex and solid tumor reflecting the disruption of myelinated fiber orientation by the pathology. In cystic or necrotic regions there is not anisotropy in diffusion. However, the mass effect of a tumor can induce a displacement or distortion of fiber tracts, by mechanical compression, yielding a decrease in mean ADC values and an increase in diffusion anisotropy relative to white matter in health hemisphere as it is shown in figure 5. Moreover, a significant difference in diffusion anisotropy appears between regions of white matter edema and tumor, enhancing or not. This difference is related to the existence of intact myelin membranes within the edema region and the loss of that integrity in areas of active tumor cell proliferation. This promises useful applications in diagnosis and preoperative planning since ADC values alone do not allow the distinction between enhancing central tumor and surrounding tissue.

Fig. 4. Metastatic brain tumor. A: T1W image; B: T2W image; C and D show MD and FA overlay maps respectively, and mean values from peritumoral edema. (Lu, Ahn et al. 2003; Kinoshita, Yamada et al. 2005). Reproduced with permission

232 Advances in the Biology, Imaging and Therapies for Glioblastoma

proved a consistent relationship between DTI-related information and histopathological

FA and MD are very sensitive indicators of white matter integrity also. Some clear examples of color-based images obtained from DTI acquisition are shown in figure 4 and 5, where not only edema region can be clearly distinguished from solid tumor by MD and FA color-based maps (figure 4), but also anisotropic directional maps depict those regions of fiber tract

The diffusion MRI study of a brain tumor patient may show a lack of a significant difference in mean diffusion anisotropy between cerebral cortex and solid tumor reflecting the disruption of myelinated fiber orientation by the pathology. In cystic or necrotic regions there is not anisotropy in diffusion. However, the mass effect of a tumor can induce a displacement or distortion of fiber tracts, by mechanical compression, yielding a decrease in mean ADC values and an increase in diffusion anisotropy relative to white matter in health hemisphere as it is shown in figure 5. Moreover, a significant difference in diffusion anisotropy appears between regions of white matter edema and tumor, enhancing or not. This difference is related to the existence of intact myelin membranes within the edema region and the loss of that integrity in areas of active tumor cell proliferation. This promises useful applications in diagnosis and preoperative planning since ADC values alone do not

allow the distinction between enhancing central tumor and surrounding tissue.

Fig. 4. Metastatic brain tumor. A: T1W image; B: T2W image; C and D show MD and FA overlay maps respectively, and mean values from peritumoral edema. (Lu, Ahn et al. 2003;

Kinoshita, Yamada et al. 2005). Reproduced with permission

data obtained from biopsies.

degeneration.

Fig. 5. T1W (upper) and relative anisotropy maps (lower). Each color indicates the principal eigenvector. Color intensity correlates with anisotropy grade. A: right frontal brain tumor; B: anaplastic ependymoma. In lower panels is clearly detected the loss of white matter integrity in both cases. (Lu, Ahn et al. 2003; Kinoshita, Yamada et al. 2005). Reproduced with permission

#### **3.3.2 Tractography**

Gliomas are infiltrating tumors in which the tumor cells may invade cerebral structures following the path of normal fiber tracts, making it difficult to differentiate tumoral from healthy tissue and detect the extension of white matter infiltration. Although several approaches from DTI have been reported to make the detection and reliable quantification of this diffuse tumor infiltration (Price, Burnet et al. 2003; Tropine, Vucurevic et al. 2004), this aspect still represents a very challenging question. One of the best developments to improve this situation is based in the visualization of diffusion tensor field by ellipsoids and colour coding of the main diffusion orientations (MDO) that allows the assessment of fibre bundles infiltration more reliably (Schluter, Stieltjes et al. 2005). The color maps and diffusion ellipsoids permit the identification of the particular fibre tracts bundles leading to the periphery of glioma by a reduced FA. Obviously, this approach is helping to improve therapeutic decisions and to minimize interventional risk. Additionally, it has been found that tumor recurrence has the tendency to occur along directions where the region of abnormal isotropic diffusion spreads beyond the region of abnormal anisotropic diffusion. On these grounds, it is eventually possible to predict regions of potential tumor infiltration and/or recidive by detecting abnormalities with high spatial resolution by using DTI. An aspect of increasing interest is the development of new applications that allow obtaining not only presurgical but also postsurgical information related to the tumor affected tissue for a more accurate postoperative prognosis.

have been reported to be dose dependent either on chemotherapy or radiotherapy in human brain tumors (Chenevert, Stegman et al. 2000). Concluding, by carrying out DWI studies and determining diffusion coefficients in glioma patients it is possible to distinguish different regions within the tumor, not accessible to conventional MRI techniques as well as to obtain

More recently Functional Diffusion Maps (fDMs) were developed as a new application of DWI that can be used as a surrogate biomarker for brain tumor cellularity (Ellingson, Malkin et al. ; Moffat, Chenevert et al. 2005). fDMs obtained computationally allow establishing the correlation between water diffusivity and cellularity mainly to predict the effectiveness of tumor treatment on a voxel-by-voxel pixel analysis. These color based maps are obtained by comparing diffusion MR images at two time points, before and after therapy, to evaluate quantitatively the therapeutic-induced changes in ADC values of the tumor. These maps are depicted in three colors by segmenting the tumor in different regions depending on the magnitude and directions of the ADC changes, allowing the use of this computational analysis as a surrogate marker of the early tumor response to the treatment

Fig. 7. fMD map involving coregistration of images before and after the treatment. A threecolor overlay of postprocessed and analyzed maps show ADC values in the tumor that are unchanged (green), significantly increased (red), or significantly decreased (blue). These regions may be associated to unresponsive (green) or responsive (red, blue) regions. Data can be also represented in a scatter plot allowing quantitative assessment of ADCs changes

Some authors have concluded that fDMs as obtained from the analysis of longitudinal diffusion images appear as a potent tool to detect the therapeutic response of tumors

(Moffat, Chenevert et al. 2006). Reproduced with permission.

an early prediction of the therapeutic efficacy.

**3.4.1 Functional Diffusion Maps** 

as depicted in figure 7.

Fig. 6. MR images of a grade III glioma in the superior medial region of the left frontal lobe of a patient before (upper panels) and after (lower panels) surgery. A, D: T2W images, B, E: Gdenhanced T1W images, C, F: anisotropy weighted color maps allow differentiation of WM fiber tracts (green arrows) located in the tumor (white arrow) vicinity, PRE, POST: tractograms of the cortico-spinal tract. Yellow arrow signals deviation and deformation of the tract in the tumor proximity. (Lazar, Alexander et al. 2006). Reproduced with permission.

As shown in figure 6, DTI is probably one of the most useful clinical applications in this direction, but one of the major obstacles for accurate reconstruction of white matter trajectories is the use of an adequate and suitable tracking algorithm. Simple tracking algorithms are not appropriate in anatomically complex fascicular pathways often leading to incorrect directional information. To solve this, sophisticated algorithms have been developed to improve reconstruction of fiber tracts in these complex areas. Even more, one of the main challenges in DTI fiber tractography studies is to have the skill to reconstruct white matter tracts related to a specific network. This becomes even more complicated in the presence of a tumor where functionally relevant regions are displaced or infiltrated in the surrounding tissue. That goal has encouraged some authors to develop strategies by combining functional MRI and DTI (fMRI-guided DTI) to define related structures based on functional anatomy for subsequent fiber tracking (Kleiser, Staempfli et al.). Obviously, the ability to extract the most useful information from DTI acquisition in diagnosis, prognosis and therapy validation of gliomas is an active and challenging area of interest in clinic and research.

#### **3.4 Assessment of treatment response with diffusion MRI**

Since ADC measurements are able to monitor cancer progression, they may be also sensitive to assess the effectiveness to therapy. An increase in the translational motion of water molecules is expected when there are physiological responses like a loss of membrane integrity, a decrease in cellularity, changes in cellular size, extracellular volume or membrane permeability (Hamstra, Rehemtulla et al. 2007). Even more, changes in ADC

Fig. 6. MR images of a grade III glioma in the superior medial region of the left frontal lobe of a patient before (upper panels) and after (lower panels) surgery. A, D: T2W images, B, E: Gdenhanced T1W images, C, F: anisotropy weighted color maps allow differentiation of WM fiber tracts (green arrows) located in the tumor (white arrow) vicinity, PRE, POST: tractograms of the cortico-spinal tract. Yellow arrow signals deviation and deformation of the tract in the

As shown in figure 6, DTI is probably one of the most useful clinical applications in this direction, but one of the major obstacles for accurate reconstruction of white matter trajectories is the use of an adequate and suitable tracking algorithm. Simple tracking algorithms are not appropriate in anatomically complex fascicular pathways often leading to incorrect directional information. To solve this, sophisticated algorithms have been developed to improve reconstruction of fiber tracts in these complex areas. Even more, one of the main challenges in DTI fiber tractography studies is to have the skill to reconstruct white matter tracts related to a specific network. This becomes even more complicated in the presence of a tumor where functionally relevant regions are displaced or infiltrated in the surrounding tissue. That goal has encouraged some authors to develop strategies by combining functional MRI and DTI (fMRI-guided DTI) to define related structures based on functional anatomy for subsequent fiber tracking (Kleiser, Staempfli et al.). Obviously, the ability to extract the most useful information from DTI acquisition in diagnosis, prognosis and therapy validation of gliomas is an active and challenging area of interest in clinic and

Since ADC measurements are able to monitor cancer progression, they may be also sensitive to assess the effectiveness to therapy. An increase in the translational motion of water molecules is expected when there are physiological responses like a loss of membrane integrity, a decrease in cellularity, changes in cellular size, extracellular volume or membrane permeability (Hamstra, Rehemtulla et al. 2007). Even more, changes in ADC

tumor proximity. (Lazar, Alexander et al. 2006). Reproduced with permission.

**3.4 Assessment of treatment response with diffusion MRI** 

research.

have been reported to be dose dependent either on chemotherapy or radiotherapy in human brain tumors (Chenevert, Stegman et al. 2000). Concluding, by carrying out DWI studies and determining diffusion coefficients in glioma patients it is possible to distinguish different regions within the tumor, not accessible to conventional MRI techniques as well as to obtain an early prediction of the therapeutic efficacy.
