Advanced diffusion MRI and biomarkers in the central nervous system: A new approach

Radiología. 2017;59(4):273---285

www.elsevier.es/rx

UPDATE IN RADIOLOGY

Advanced diffusion MRI and biomarkers in the central nervous system: A new approach夽 T. Martín Noguerol ∗ , J.P. Martínez Barbero Sección de Neurorradiología, Clínica las Nieves, SERCOSA, Grupo HealthTime, Jaén, Spain Received 8 January 2017; accepted 16 April 2017

KEYWORDS Magnetic resonance imaging; Diffusion; Diffusion Tensor Imaging; Biomarkers

PALABRAS CLAVE Resonancia magnética; Difusión; Tensor de difusión; Biomarcadores

Abstract The introduction of diffusion-weighted sequences has revolutionized the detection and characterization of central nervous system (CNS) disease. Nevertheless, the assessment of diffusion studies of the CNS is often limited to qualitative estimation. Moreover, the pathophysiological complexity of the different entities that affect the CNS cannot always be correctly explained through classical models. The development of new models for the analysis of diffusion sequences provides numerous parameters that enable a quantitative approach to both diagnosis and prognosis as well as to monitoring the response to treatment; these parameters can be considered potential biomarkers of health and disease. In this update, we review the physical bases underlying diffusion studies and diffusion tensor imaging, advanced models for their analysis (intravoxel coherent motion and kurtosis), and the biological significance of the parameters derived. © 2017 SERAM. Published by Elsevier Espa˜ na, S.L.U. All rights reserved.

RM-Difusión avanzada y biomarcadores en el sistema nervioso central: un nuevo enfoque Resumen La introducción de las secuencias potenciadas en difusión ha supuesto una revolución para la detección y caracterización de la patología del sistema nervioso central. Sin embargo, en numerosas ocasiones, la valoración de dichos estudios se limita a una estimación cualitativa. Además, la complejidad fisiopatológica de las distintas entidades que afectan al

夽

Please cite this article as: Martín Noguerol T, Martínez Barbero JP. RM-Difusión avanzada y biomarcadores en el sistema nervioso central: un nuevo enfoque. Radiología. 2017;59:273---285. ∗ Corresponding author. E-mail address: [email protected] (T. Martín Noguerol). 2173-5107/© 2017 SERAM. Published by Elsevier Espa˜ na, S.L.U. All rights reserved.

274

T. Martín Noguerol, J.P. Martínez Barbero sistema nervioso central no siempre puede ser correctamente explicada por los modelos clásicos. El desarrollo de nuevo modelos para el análisis de las secuencias de difusión aporta numerosos parámetros que podrían permitir un abordaje cuantitativo tanto desde el punto de vista del diagnóstico como pronóstico, así como para llevar a cabo la monitorización terapéutica, y podrían ser considerados como potenciales biomarcadores de salud y enfermedad. Realizamos por este motivo una actualización que incluye las bases físicas de los estudios de difusión y tensor de difusión, sus modelos de análisis avanzado (IVIM y Kurtosis) y la significación biológica de los parámetros derivados. © 2017 SERAM. Publicado por Elsevier Espa˜ na, S.L.U. Todos los derechos reservados.

Introduction Diffusion weighted magnetic resonance imaging (DWI-MRI) are capable of providing us with non-invasive estimates of the movement of water molecules in a biological environment.1 DWI-MRI studies allow us to perform quantitative and qualitative assessments of such movement and provide us with anatomical and functional information on the tissues. The design of DWIs has been perfected thanks to multiple technical innovations. This is how we have been able to reduce the times of acquisition and the artifacts and improve the signal-noise correlation. Other technical optimizations of DWIs have allowed us to develop diffusion tensor imaging (DTI) --- a tool that has revolutionized the assessment of white matter in the central nervous system (CNS). DWIs have proven to be of great utility for the study of CNS diseases. However, the physical bases of DWIs have remained unchanged during the last 20 years. This is why, today, innovations focused on the phase of analysis and interpretation of the data obtained in the acquisition phase. Small modifications in the acquisition process but, above all, it is the use of different mathematical and biological models for the analysis of signal intensity drops that is allowing us to understand more accurately the physiopathological processes that occur in the CNS. Such analytical models will allow us to obtain multiple parameters that will be potentially used as biomarkers. In this update we will discuss in a more pedagogical way the physical bases of the conventional models of DWI and DTI, and delve into new analytical models based on the intravoxel incoherent motion (IVIM) and Kurtosis models. The origin and biological significance of the different parameters derived from these analytical models and the prospective clinical applications of such parameters will be fully explained in detail.

Physical and acquisition bases of classic DWIs and DTIs The classic DWI sequences are based on the implementation of two (2) diffusion gradients (identical in magnitude and duration) into one spin echo (SE) sequence.

Technical innovations have made it possible to develop sequences based on turbo spin echo (TSE) sequences and, above all, the most widely used today --- the echo-planar imaging (EPI) sequences.2 The first gradient implemented will dephase water molecules and after a while, the second gradient will rephase such molecules in the same proportion that they were dephased in the first place. The molecules that remain stationary after the second gradient will fully recover their initial energetic state, which will translate into high signal intensities in high b values of the diffusion sequence, the so-called ‘‘diffusion restriction’’. On the other hand, the molecules experimenting movement during the time elapsed between both gradients will lose their position and will not be able to recover it completely, meaning that their energetic state will be lower and that signal attenuation proportional to the degree of movement will occur, the so-called ‘‘free or facilitated diffusion’’.3 This is how DWIs make distinctions between different kinds of tissues based on the freedom of water movement inside them (Fig. 1). Diffusion may be considered isotropic or anisotropic based on the direction of water movement.4 In isotropic diffusion, water movement occurs with equal probability in all directions of space, whether such space is limited or not. In anisotropic diffusion, there is a dominant direction of water movement in a given tissue that is usually conditioned by the existence of anatomical and physiological barriers. This is the case of water movement inside axons and between myelin sheaths where the dominant direction of movement occurs across the long axis of the axon (Fig. 2). For the study of anisotropic diffusion, and as a technical optimization of the DWI sequences, the diffusor tensor imaging (DTI) has been developed.5 DTIs are based on the implementation of diffusion gradients into multiple orthogonal directions in space (at least 6). This is the what to assess the mobility of water molecules in every direction and detect if there is one dominant diffusion direction. From the mathematical standpoint one 3 × 3 matrix is generated that will diagonalize the dominant direction in each plane of space represented by one vector (eigenvector) of a determined magnitude (eigenvalue).3 One of the added values that DTIs provide us with is the performance of tractography studies.6 Tractographies are based on 3D representation of white matter bundles through

Advanced diffusion MRI and biomarkers in the central nervous system

90º

Dephase gradient

180º

275

Rephase gradient

Reading

Signal intensity

1

2

3

Acute ischaemia

Healthy tissue

Chronic ischaemia

Figure 1 Diffusion sequence scheme. (1) The molecules with limited movement or that do not experiment movement in the time elapsed between the two (2) diffusion gradients (dephase and rephase) will maintain their energy level at the end of the sequence since they will fully recover their phase and, therefore, will associate high signal intensity in high b values as it is the case with ischemia in the acute phase where the cytotoxic edema leads to a reduced extracellular space that makes it hard for the water molecules to move at such level. (2) In a normal tissue there is a certain physiological degree of freedom of movement for water molecules in the extracellular space that is consistent with an intermediate level of signal intensity. (3) When it comes to the ventricular system or chronic ischemic lesions, there is total movement of water molecules between the two (2) diffusion sequence gradients that leads to a total loss of phase and signal intensity in high b values.

the determination of the dominant direction of water movement in each voxel.

Physical bases and advanced analytical models of DWIs: the IVIM and Kurtosis models Yet despite the quantitative and qualitative jump that the development and implementation of DWI and DTI sequences has meant in the daily radiological and clinical practice, numerous studies have confirmed that the physiological and pathological processes occurring underneath the tissues are much more complex than we thought they were.7---9 The comparison of in vivo and in vitro studies has determined the existence of high complexity based on the presence of several compartments at tissue level (intracellular, extracellular and intravascular), which makes the classical models of acquisition and analysis of diffusion sometimes insufficient to explain the physiopathology of the CNS. In order to study such compartments and biological processes we can intervene both in the acquisition process and in the analysis of the data obtained. Starting from the same classical diffusion sequence we will modify the number of b values acquired and their magnitude, which will allow us, in

a second phase, to obtain a more precise model to analyze signal intensity drops. The selection and weighting of b values is the key point in the development of these new analytical models and this is why we believe that it is necessary to come up with a brief reminder of the physical and biological significance of such parameter. The b value is determined by the gyromagnetic coefficient (), the intensity of the magnetic field (G), the coding of the gradients-pulse widths (ı), and the time elapsed between both gradients () following this formula: b =  2 · G2 · ı2 ( − ı/3). The intensity of the magnetic field is usually unmodified in the diffusion studies. However, both the gradients-pulse widths, and the time elapsed between both gradients may be modified, which leads us to the concept of low b values and high b values10 (Fig. 3). Low b values are the key of the IVIM model, which may assess, thanks to the multiple use of b values < 150 s/mm2 , the random movement (incoherent) of water inside the capillary network of a certain voxel (intravascular compartment).1 High b values are the key of the Kurtosis model, which may determine, in a very limited way, cellular membrane-linked water movement (intracellular and extracellular compartments) (Fig. 4).

276

T. Martín Noguerol, J.P. Martínez Barbero

e1 λ 1 e2 λ 2

e2 λ 2

e1

λ1

e3 λ 3

Isotropic diffusion λ1 ≈ λ2 ≈ λ3

e3 λ 3

Anisotropic diffusion λ1 >> λ2 ≈ λ3

Figure 2 Main types of water movement in the tissues. We talk about isotropic diffusion whenever there is one dominant direction of water movement in the extracellular space, whether limited or not. We talk about anisotropic diffusion whenever there is one dominant direction of water movement in the tissues, as it is the case with white-matter fiber bundles. In the case of isotropic diffusion, there are no significant magnitude differences among the three (3) main eigenvectors (1 ≈ 2 ≈ 3 ). Whenever there is anisotropic water diffusion, we will be finding one axial dominant direction of water movement across the major axis of the axon (1  2 ≈ 3 ) with respect to the two (2) minor 2 eigenvectors (perpendicular). These vectors are the foundation of the main parameters derived from DTI studies (AD: axial diffusivity; RD: radial diffusivity; FA: fractional anisotropy; MD: mean diffusivity).

IVIM model The IVIM model is based on the consideration of two (2) main compartments within the same tissue voxel: extracellular and intravascular tissues.1 This approach allows us to know, through one single study, the degree of cellularity of a tissue, its real diffusion coefficient (D), and estimate the vascularization of a tissue through D* that is consistent with pseudodiffusion, that is, the measurement of diffusion in that vascular compartment (one random type of movement that matches the classical study of diffusion). Also, f, a parameter derived from D*, represents the fraction of perfusion or the percentage of molecules moving within the capillary vessels of a certain voxel.11 We have been able to see that the ADC values obtained through classic diffusion sequences are different from in vivo and in vitro samples.12 The ADC values of in vivo tissues may present higher numbers than in vitro data. Hence, the apparent nomenclature since it is not an exact measurement of the degree of molecule diffusion in the tissues. The difference between both measurements has been attributed to the existence of another type of additional movement that interferes with the measurement of the predicted ADC. Such movement occurs in the intravascular compartment. The movement of molecules inside the capillary makes it possible to overestimate ADC values when assessing in vivo biological tissues13 (Fig. 5). Through the implementation of b values below 150 s/mm2 we manage to dephase, that is, terminate the contribution to the signal intensity of such more intravascular movement. The real value of the diffusion coefficient will be

estimated using b values above 150 s/mm2 in order to avoid such contamination.14

Kurtosis model The conventional analysis of DWI signal intensity drops includes the presence of a random movement of water molecules in a given tissue. The odds that one proton diffuses in a given direction are determined by one Gaussiantype function and also there is standard deviation around the ADC values. However, it has been proven that the presence of physiological barriers like cellular membranes or organelles limits movement randomization. Water does not diffuse with the same freedom and degree of restriction in all directions. This membrane-linked water movement may be studied using the Kurtois model that considers that there is a non-Gaussian distribution in the diffusion process15 (Fig. 6). The Kurtosis model, imported from the field of statistics, refers to the degree of concentration shown by the values around the central zone of distribution (0 in the case of Gaussian distribution---normal---and >0 when the values obtained do not follow Gaussian distribution). In order to be able to detect this membrane-linked water movement, which is a very slow water movement by the way, it is necessary to implement extremely high b values (>2000 s/mm2 ).16 This analytical model allows us to estimate the heterogeneity or complexity of tissues, since it brings us close to the degree of water interaction with biological membranes,

Advanced diffusion MRI and biomarkers in the central nervous system

277

Molecule with very facilitated movement (free water, intravascular,…)

Molecule with very limited movement (protein-linked water, membrane-linked water,…)

Δ

b = γ 2.G 2 .δ 2 (Δ – δ /3)

G Unit of time

Low b value

δ

δ Δ

G

High b value

δ

δ

Figure 3 Biological significance of low and high b values. Low b values mean that the diffusion gradient is applied for a short span of time and, also, that there is a short time elapsing between both gradients. This means that if one water molecule moves very slowly then we will not be able to confirm such movement after such short span of time, with the corresponding loss of signal. On the contrary, if one molecule moves very fast, we will just need to apply low b values in order to confirm it changed position and, therefore, its loss of signal. However, in the case of high b values (>1000---2000 s/mm2 ), the molecules that barely have any movement are studied with a wide span of time between both gradients, so after this time has elapsed we will be able to detect some position variations, which will be confirmed by partial losses of signal intensity.

something that goes beyond the ‘‘simple’’ restriction of water movement in the extracellular space. Several studies have proven that when the ADC values of different tissues are the same, that is, tissues that apparently show the same degree of diffusion restriction, the estimates from the Kurtosis model allow us to make distinctions among them.17 The further away one tissue moves from normal distribution, the higher its complexity will be in the Kurtosis model.15

reproducibility, show the characteristics of a tissue from a quantitative standpoint, and indicate the existence of any clinically relevant underlying physiological or pathological processes.18 These parameters have been summarized in Table 1.

Parameters derived from diffusion sequences

Parameters derived from DWIs and DTIs The parameters derived from DWI and DTI studies allow us to assess the different physiopathological process from the quantitative viewpoint in a more precise way and with higher diagnostic accuracy. Many of these parameters are still in the phase of clinical validation since we still do not have a global standardization of the acquisition and analysis protocols of diffusion sequences, except for the technical limitations inherent to diffusion sequences like higher sensitivities to movement artifacts or magnetic susceptibility. This is why it is essential to know the biological significance of these parameters as a starting point to be able to conduct future studies and contribute to such validation. In order to be considered biomarkers, these parameters need to meet certain prerequisites such as reliability,

The ADC (apparent diffusion coefficient), the most widely used parameter derived from the acquisition of conventional diffusion sequences, is capable of estimating the degree of water restriction in the extracellular space. Such ADC values will be reduced in those situations where the extracellular space is reduced due to a larger number of cells or the presence of oversized cells (cytotoxic edema). In order to assess the ADC we need the participation of two (2) b values, usually 0 and the highest b value of all (1000---1500 s/mm2 in the CNS). The ADC allows us to correct the influence that the T2-weighted effect has on diffusion itself and confirm whether the hyperintensity of a given tissue for the highest b values is due to a real diffusion restriction (if so, the ADC values will be low), or whether we have a T2 ‘‘shine-through’’ effect (that will associated high ADC values).

278

T. Martín Noguerol, J.P. Martínez Barbero

IVIM

ADC

KURTOSIS

1 × 104 8 × 104 6 × 104 4 × 104 2 × 104 0 0

Intravascular compartment

500

1500 1000 b factor [s/mm2] Extracellular compartment

2000

2500

Cellular compartment

Figure 4 Models of signal intensity drop of the diffusion sequence. The acquisition of b values below 150 s/mm2 and above 1500 s/mm2 allows us to obtain one curve of signal intensity drop (red curve) that reflects one exponential multicompartmental model more adjusted to reality than the classic linear monocompartmental model (blue line). In the case of low b values there is a pronounced drop of signal intensity (first part of the red curve) due to fast dephase of water molecules inside the intravascular space. In the case of intermediate b values there is a maintained drop of signal intensity that reflects the behavior of water in the extracellular space, parallel to a hypothetical drop of signal intensity of the monocompartmental model (blue line). In the case of high b values there is a lower drop of signal intensity than expected based on the monocompartmental model that reflects the interaction of water molecules and membranes at cellular level. This analysis is conducted using post-processing tools that allow us to adjust the drop of signal intensity for every b value in order to know what the state of every compartment really is by using only one single acquisition.

Here we need to remember that the ADC is an apparent value because on top of measuring the diffusion of extracellular water in the tissues, it is also influenced by other movements like respiration, cardiac and vascular movements, cerebrospinal fluid, or secretions. The ADC main advantage on all the other parameters that we will be discussing is its wide availability.19

The main clinical application of conventional diffusion sequences and the ADC in the CNS is the detection and characterization of ischemic pathology. The degree of diffusion restriction allows us to establish the time of progression of ischemic strokes with great accuracy.20 Diffusion studies have also allowed us to improve the characterization of tumor lesions. Numerous publications

Figure 5 IVIM. In healthy tissues (1), water movement in the extracellular space is slightly influenced by the presence of water movement in the intravascular compartment, so it is possible to find ADC values that are of equal or slightly greater value than the pure diffusivity values (D). In the case of tumor pathology of the central nervous system (2) and yet despite the fact that there is a reduced extracellular space due to a higher number of tumor cells, the ADC values may be increased by the contribution made by water movement in the intravascular space (ADC > D). Such capillary component is usually enlarged in high-grade lesions due to a phenomenon known as tumor neoangiogenesis. The IVIM model allows us to perform correct catheterizations of tumor lesions while avoiding the influence of the vascular component that may also be estimated without using IV contrast.

Advanced diffusion MRI and biomarkers in the central nervous system

279

Extracellular space

K

Cellular space

Figure 6 Kurtosis. The interaction of water molecules with cellular membranes follows the non-Gaussian distribution (red arrows) and moves away from the normal distribution of water molecule movement in the extracellular spapce (blue arrows). Kurtosis shows the difference of such non-Gaussian distribution with respect to normal distribution that will be greater with more heterogeneous tissues.

have proven that tumors with lower ADC values have more cellularity, which is consistent with a higher degree of tumor aggressiveness.21 Similarly, the ADC values may be used to monitor these tumors response to treatment, also there is an increase of ADC values in tumors that respond to treatment due to phenomena of necrosis.22 Using the ADC values we may predict the response to treatment since lesions with the lowest ADC values prior to treatment show more positive responses than lesions with higher ADC values that normally associate poorly oxygenated necrotic areas, with less capillary supply that reduces the effectiveness of treatment both chemotherapy and radiotherapy23 (Fig. 7). The use of the ADC allows us to assess intracranial infectious diseases, as it is the case with abscesses where diffusion is of great help during differential diagnosis with other type of necrotic lesions with peripheral

Table 1

Physical bases

Biological significance

Units

ADC (apparent diffusion coefficient) MD (mean diffusivity)

Monoexponential drop of signal intensity Arithmetic average of 3 eigenvalues Index of tissue organization

Water movement in extracellular compartment Water movement in extracellular compartment Integrity of white matter

mm2 /s

Main eigenvalue Arithmetic average of 2 minor eigenvalues Drop of signal intensity above150 s/mm2 Drop of signal intensity below150 s/mm2 Active capillary density

Axonal integrity Integrity of the myelin sheath Real tissue diffusion

mm2 /s mm2 /s

Water pseudo-diffusion in the vascular compartment Percentage of water molecules in the vascular space in one voxel Tissue complexity, lesion heterogeneity

mm2 /s

D (diffusion coefficient) D* (pseudo-diffusion) f (perfusion fraction)

Kurtosis

DTI studies allow us to obtain a greater number of parameters, most of them of great utility in the assessment of white matter. The most widely known parameter is fractional anisotropy (FA) that shows the degree of tissue anisotropy, that is, how dominant the direction of water movement really is in a given voxel, ranging from 1

Parameters derived

FA (fractional anisotropy) AD (axial diffusivity) RD (radial diffusivity) IVIM

Parameters derived from DTI sequences

Parameters derived from basic and advanced analyses of DWI and DTI studies.

Monocompartmental DTI

enhancement like metastases. Inside the abscesses, the presence of debris, cellular remains, and microorganisms conditions a very important diffusion restriction, although this finding is not always specific of abscesses and there may be overlapping with certain histological subtypes of metastases in the CNS.24,25

K (Kurtosis)

Degree of deviation of signal intensity with respect to normal

mm2 /s 0 < FA > 1

mm2 /s

Per cent

---

280

T. Martín Noguerol, J.P. Martínez Barbero

Figure 7 Prediction of treatment response and follow-up. Fifty-six year old patient with right insular lesion and biopsy results of WHO grade III astrocytoma. In the study prior to treatment (January 2015) one infiltrating lesion may be identified without apparent enhancement in the contrast study (a) showing regions of noticeable diffusion restriction (b) with ADC values of 0.6 × 10−3 mm2 /s (c) that are consistent with hypercellular regions. The patient underwent radiotherapy and tumor mass reduction was confirmed (d), and a lower degree of diffusion restriction with increased ADC values of 1.5 × 10−3 mm2 /s (f) consistent with an adequate response to treatment due to cell death. In this case, the diffusion study allows us to predict the favorable response to treatment and confirm the correct progression of the disease.

(maximum anisotropy) to 0 (maximum isotropy). FA may be considered a biomarker of axonal integrity since it has high sensitivity in the assessment of white matter pathology and is altered (usually reduced) in almost in all the processes that affect it; however, and precisely for the same reason, its specificity is not the most desirable one.26 The dominant direction in every plane of space is determined by the eigenvectors that include one value, and one magnitude known as eigenvalues that is expressed in mm2 /s (same as the ADC). The average value of the three (3) main eigenvalues is known as mean diffusivity (MD), that may be regarded as the most exact measurement of the ADC since it takes the three (3) main directions of water movement into consideration, and behaves similar to the ADC in different clinical settings. Another two (2) parameters derived from DTIs are axial diffusivity (AD), and radial diffusivity (RD). The AD shows the movement of water molecules in the longitudinal direction---dominant of the axon, and is consistent with the main eigenvalue. Such measurement will usually be high since, as we already pointed out, diffusivity is facilitated in such axis. The AD will allow us to assess the integrity of axonal conduction, or the existence of a lesion at main neural level

(that would alter the axon flow due to a mechanism similar to Wallerian degeneration). The RD is the expression of the arithmetic average of the other (2) minor eigenvalues --- those located perpendicularly to the main eigenvalue. Diffusivity in the radial plane will be conditioned by the existence of myelin sheaths.7 It is these biological characteristics that make RD an adequate parameter for the assessment of myelin integrity, probably the most specific of all parameters derived from DTI. The loss of myelin will increase the levels of RD, since the molecules will find less problems moving in such transverse plane.19 (Fig. 8). There are numerous clinical applications of DTI-based studies.27 Among them, the most significant one due to the qualitative jump it means is the assessment of tumor pathology. Both AF and MD have been used to characterize the area of altered peritumor signal intensity in T2 and FLAIR-weighted sequences. In the case of metastases o meningiomas, such edema shows lower AF values and higher MD values than with high-grade gliomas.8,28 The explanation to this phenomenon seems to be based on tumor cell infiltration of the adjacent white matter into high-grade glial lesions, which allows its differential diagnosis with

Advanced diffusion MRI and biomarkers in the central nervous system

T2-weighted TSE images

FA map

281

0.94

0.69

0.44

0.19

–0.06

RD map

2.73

AD map

3.43

1.97

2.57

1.21

1.71

0.45

0.85

–0.31

–0.01

Figure 8 Twenty-six year old patient with a diagnosis of multiple sclerosis (MS) showing one lesion in the left cerebral peduncle with reduced values of fractional anisotropy (FA) (0.3) with respect to the contralateral peduncle (0.8). Such lesion does not show any significant variations in the values of axial diffusivity (AD), however, increased values of radial diffusivity (RD) (1.8 × 10−3 mm2 /s) may be identified with respect to the contralateral peduncle (0.5 × 10−3 mm2 /s), indicative of selective damage to the myelin sheath and compatible with the patient’s baseline condition.

CNS-associated metastases; however, these data do not appear to be generalizable to the assessment of intratumor component.29 The use of DTI-based 3D reconstructions through the tractography technique allows us to assess the relation and dependency of CNS lesions with the main white matter bundles, and estimate the resectability degree of such bundles based on the existence of movement, invasion, or infiltration of such tracts.5 Other clinical applications of DTIs include the assessment of demyelinating lesions that usually identifies the increase of MD values and decrease of FA values due to the existing inflammatory component, being the RD the most specific parameter for the assessment of myelin damage.30 Numerous studies have confirmed the higher sensitivity of DTIs for the detection of white matter disorders that look apparently normal in patients with demyelinating disease.31,32 Also DTIs have been used to study epilepsy,33 or even neurodegenerative conditions, and have proven their potential utility in the differential diagnosis of different parkinsonian syndromes.34,35 Some authors have proposed the use of AF as a valid parameter to monitor responses to treatment in patients with chronic hydrocephalia in adults after one shunt catheter has been placed.36

Parameters derived from the IVIM model As we mentioned before, the parameter derived from D allows us to obtain a more accurate and reproducible measurement of the true tissue diffusion coefficient. This is very important in the assessment of tumor pathology since it is not rare that high-grade tumors where low ADC values are expected due to their high cellularity paradoxically show high ADC values due to phenomena of associated necrosis, and due to the influence of the vascular component in the solid portion of the tumor. Such vascular component in highgrade tumors is usually enlarged due to the angiogenesis process that is inherent to this type of lesions.37 D* and f are parameters that make reference to the intravascular component of one determined tissue voxel. D* reveals water movement inside the capillary network, and f (derived from such D*) shows the active capillary density, that is, the percentage of capillaries in whose interior water movement is detected.9 For the time being, the clinical applications of the IVIM model are limited to research only.38 Intuitively we may deduce that, in the case of acute ischemic pathology, on top of finding reduced ADC levels (and hence reduced D levels), both D* and f values will be reduced too due to decreased flow in the distal capillary network.39 For the assessment

282

T. Martín Noguerol, J.P. Martínez Barbero

Gd-enhanced T1-weighted image

VSC map

1197 844 490 137 –216.8

D∗ map

1.88

f map 168.

1.63 126.

1.38 84.

1.13

0.88

42. 32. 21. 11. 1.

Figure 9 Sixty-three year old patient with a diagnosis of gliobastoma. One wide lesion may be identified at the level of both frontal lobes and knee of the corpus callosum showing intense enhancement after the administration of contrast with increased values of cerebral blood volume (CBV). The analysis using the IVIM model shows an increase of pseudo-diffusion (D*) and the perfusion fraction (f) of up to 40 per cent that is consistent with areas of neovascularization. The upper chart shows signal intensity in the T2*-weighted perfusion imaging showing a more pronounced drop (blue curve) in the tumor solid compartment compared to the apparently healthy parenchyma (orange curve), suggestive of neoangiogenesis. The lower charts show the drop of signal intensity of the diffusion sequence following the IVIM model that confirms a more profound drop of the first part of the red curve in the tumor solid component (blue line) compared to the apparently healthy parenchyma (orange line) suggestive of a predominant vascular/capillary component with fast water molecule dephase inside.

of tumor pathology the existence of an adequate correlation between D and ADC values has been confirmed, being of great utility for tumor degree estimation; also there is a correlation between f and parameters derived from perfusion studies37,40 (Fig. 9). Also the use of the IVIM model has been proposed for survival rate prediction in high-grade tumors,41 or even for functional studies through the identification of increased D* and f values in activated cortical areas with the different motor, visual, or cognitive paradigms.42

Parameters derived from the Kurtosis model The main parameter derived from the non-Gaussian analysis of the diffusion sequence signal intensity is the Kurtosis (K) diffusion sequence, or average Kurtosis (AK). K is indicative of how much the distribution we study moves away from normal distribution (Gaussian), bearing in mind that if K = 0, then we have a normal distribution. Mathematically, K may be both positive and negative, although in biological tissues, only the positive Kurtosis is taken into consideration.43 High K values are indicative that the studied tissue is more complex, more heterogeneous, and much more different from a normal distribution.44 Such heterogeneity is

mainly due to the existence of a higher number of cells, membranes, vessels, and necrosis. Like we said, K adds a higher grade of specificity to tissue characterization compared to conventional diffusion. In this way, experimental studies have proven that there is a higher grade of Kurtois in high-grade tumors compared to low-grade tumors, since the presence of more mitosis, cellular membranes, angiogenesis, and necrosis basically makes this tissue more complex than low-grade tumors where the interstitium is predominant.45 The studies conducted using the Kurtosis model have also proven, with higher degree of statistical significance than FA and MD, their utility in the assessment of altered peritumor signal intensities. The regions of edema with tumor infiltration due to the existence of neoplastic cells from the primary tumor will show higher K values than the regions of vasogenic edemas reactive to metastasic lesions, where K values will be lower,46 which may help us make more accurate distinctions between high-grade gliomas and single metastases of the CNS (Fig. 10). Other studies have confirmed the existence of reduced K values in entities presenting with neural losses, as it is the case of Alzheimer-like neurodegenerative disorders, which allows a much earlier detection of changes.47

Advanced diffusion MRI and biomarkers in the central nervous system

FLAIR

Gd-enhanced T1-weighted image

283 224

DWI b 1.000

ADC map

K map

176

132

88 66 44 36

Figure 10 Seventy-three year old-patient with WHO grade III astrocytoma treated with surgery, chemotherapy, and radiotherapy. At the level of the right parietal lobe one complex lesion with solid and cystic component may be identified associated with severe vasogenic edema. The Kurtois analysis shows increased K values in the surgical bed represented in increased areas of tumor heterogeneity (white arrows). Note the presence of areas with increased K values in the thickness of the vasogenic edema (black arrows); such alteration, suggestive of tumor infiltration, also seems identifiable in the DWI/ADC study, but in a less evident way. The asterisk shows one focus of a probable false positive in the K map due to black-out effect secondary to the presence of hemorrhage remains.

DTI studies may show poor definition of the dominant direction when in the presence of several directions in the white matter crossing bundles. In order to overcome this limitation, advanced techniques in high resolution tractography have been developed (HARDI, q-ball imaging, spectral diffusion imaging) based on the implementation of a high number of directions (>200) or very high b values (>5000 s/mm2 ) at the expense of long times of acquisition. DTI studies based on the Kurtosis model allow us to obtain higher spatial resolutions in the cut-off points (due to better characterizations of water interactions with the myelin sheaths) and more adjusted times of acquisition.48

Conclusions The classical approach to diffusion studies and the development of new models to analyze signal intensity drops such as the IVIM or the Kurtois models give us a better understanding of the different physiopathological processes that occur in the CNS. The different parameters derived from these analytical models may be considered biomarkers and help us move from a purely qualitative interpretation of diffusion to a more multiparametric quantitative approach, which would increase our diagnostic accuracy, prognosis, and treatment follow-up in various clinical settings.

Authors 1. 2. 3. 4. 5. 6. 7. 8.

Manager of the integrity of the study: TMN and JPMB. Study Idea: TMN. Study Design: TMN. Data Mining: TMN y JPMB. Data Analysis and Interpretation: TMN and JPMB. Statistical Analysis: N/A. Reference: TMN and JPMB. Writing: TMN and JPMB.

9. Critical review of the manuscript with intellectually relevant remarks: TMN and JPMB. 10. Approval of final version: TMN and JPMB.

Ethical responsibilities Protection of people and animals. The authors declare that no experiments with human beings or animals have been performed while conducting this investigation. Data confidentiality. The authors declare that they have followed their centres protocols on the disclosure of data from patients. Right to privacy and informed consent. The authors declare that in this article there are no data from patients.

Conflicts of interests The authors declare no conflict of interests associated with this article whatsoever.

References 1. Le Bihan D, Breton E, Lallemand D, Grenier P, Cabanis E, Laval-Jeantet M. MR imaging of intravoxel incoherent motions: application to diffusion and perfusion in neurologic disorders. Radiology. 1986;161:401---7. 2. Hagmann P, Jonasson L, Maeder P, Thiran JP, Wedeen VJ, Meuli R. Understanding diffusion MR imaging techniques: from scalar diffusion-weighted imaging to diffusion tensor imaging and beyond. Radiographics. 2006;26:S205---23. 3. de Figueiredo EH, Borgonovi AF, Doring TM. Basic concepts of MR imaging, diffusion MR imaging, and diffusion tensor imaging. Magn Reson Imaging Clin N Am. 2011;19:1---22. 4. Dong Q, Welsh RC, Chenevert TL, Carlos RC, Maly-Sundgren P, Gomez-Hassan DM, et al. Clinical applications of diffusion tensor imaging. J Magn Reson Imaging. 2004;19:6---18. 5. Nucifora PGP, Verma R, Lee S, Melhem ER. Diffusion-tensor MR imaging. Radiology. 2007;245:367---84.

284 6. Romano A, D’Andrea G, Minniti G, Mastronardi L, Ferrante L, Fantozzi LM, et al. Pre-surgical planning and MR-tractography utility in brain tumour resection. Eur Radiol. 2009;19:2798---808. 7. Fox RJ, Cronin T, Lin J, Wang X, Sakaie K, Ontaneda D, et al. Measuring myelin repair and axonal loss with diffusion tensor imaging. AJNR Am J Neuroradiol. 2011;32:85---91. 8. Jakab A, Molnár P, Emri M, Berényi E. Glioma grade assessment by using histogram analysis of diffusion tensor imaging-derived maps. Neuroradiology. 2011;53:483---91. 9. Bisdas S. Are we ready to image the incoherent molecular motion in our minds. Neuroradiology. 2013;55:537---40. 10. Nicolas R, Sibon I, Hiba B. Accuracies and contrasts of models of the diffusion-weighted-dependent attenuation of the MRI signal at intermediate b-values. Magn Reson Insights. 2015;1:1---21. 11. Le Bihan D. Diffusion MRI. What water tells us about the brain. EMBO Mol Med. 2014;6:569---73. 12. Jafar MM, Parsai A, Miquel ME. Diffusion-weighted magnetic resonance imaging in cancer: reported apparent diffusion coefficients, in-vitro and in-vivo reproducibility. World J Radiol. 2016;8:21---49. 13. Le Bihan D, Turner R. The capillary network: a link between IVIM and classical perfusion. Magn Reson Med. 1992;27:171---8. 14. Le Bihan D, Breton E, Lallemand D, Aubin M-L, Vignaud J, LavalJeantet M. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology. 1988;168:497---505. 15. Grinberg F, Farrher E, Kaffanke J, Oros-Peusquens AM, Shah NJ. Non-Gaussian diffusion in human brain tissue at high b-factors as examined by a combined diffusion kurtosis and biexponential diffusion tensor analysis. Neuroimage. 2011;57:1087---102. 16. Fieremans E, Jensen JH, Helpern JA. White matter characterization with diffusional kurtosis imaging. Neuroimage. 2011;58:177---88. 17. Steven AJ, Zhuo J, Melhem ER. Diffusion kurtosis imaging: an emerging technique for evaluating the microstructural environment of the brain. AJR Am J Roentgenol. 2014;202:26---33. 18. Martí Bonmatí L, Alberich-Bayarri A, García-Martí G, Sanz Requena R, Pérez Castillo C, Carot Sierra JM, et al. Biomarcadores de imagen, imagen cuantitativa y bioingeniería. Radiologia. 2012;54:269---78. 19. Urger E, Debellis MD, Hooper SR, Woolley DP, Chen S, Provenzale JM. Influence of analysis technique on measurement of diffusion tensor imaging parameters. AJR Am J Roentgenol. 2013;200:W510---7. 20. Cauley KA, Thangasamy S, Dundamadappa SK. Improved image quality and detection of small cerebral infarctions with diffusion-tensor trace imaging. AJR Am J Roentgenol. 2013;200:1327---33. 21. Kralik SF, Taha A, Kamer AP, Cardinal JS, Seltman TA, Ho CY. Diffusion imaging for tumor grading of supratentorial brain tumors in the first year of life. AJNR Am J Neuroradiol. 2014;35:815---23. 22. Huang WJ, Wu Y, Yin G, Li B, Geng JJ. Diffusion-weighted imaging for predicting and monitoring. Am. J. Neuroradiol. 2016:1---9. 23. Brandão LA, Shiroishi MS, Law M. Brain tumors: a multimodality approach with diffusion-weighted imaging, diffusion tensor imaging, magnetic resonance spectroscopy, dynamic susceptibility contrast and dynamic contrast-enhanced magnetic resonance imaging. Magn Reson Imaging Clin N Am. 2013;21:199---239. 24. Tung GA, Evangelista P, Rogg JM, Duncan JA. Diffusion-weighted MR imaging of rim-enhancing brain masses: is markedly decreased water diffusion specific for brain abscess. Am J Roentgenol. 2001;177:709---12. 25. Hayashida Y, Hirai T, Morishita S, Kitajima M, Murakami R, Korogi Y, et al. Diffusion-weighted imaging of metastatic brain tumors: comparison with histologic type and tumor cellularity. AJNR Am J Neuroradiol. 2006;27:1419---25.

T. Martín Noguerol, J.P. Martínez Barbero 26. Oida T, Nagahara S, Kobayashi T. Acquisition parameters for diffusion tensor imaging to emphasize fractional anisotropy: phantom study. Magn Reson Med Sci. 2011;10:121---8. 27. Sundgren PC, Dong Q, Gómez-Hassan D, Mukherji SK, Maly P, Welsh R. Diffusion tensor imaging of the brain: review of clinical applications. Neuroradiology. 2004;46:339---50. 28. Lee S-K. Diffusion tensor and perfusion imaging of brain tumors in high-field MR imaging. Neuroimaging Clin N Am. 2012;22:123---34. 29. Jiang R, Du F-Z, He C, Gu M, Ke Z-W, Li J-H. The value of diffusion tensor imaging in differentiating high-grade gliomas from brain metastases: a systematic review and meta-analysis. PLoS ONE. 2014;9:e112550. 30. Hygino da Cruz LC, Batista RR, Domingues RC, Barkhof F. Diffusion magnetic resonance imaging in multiple sclerosis. Neuroimaging Clin N Am. 2011;21:71---88. 31. Onu M, Roceanu A, Sboto-Frankenstein U, Bendic R, Tarta E, Preoteasa F, et al. Diffusion abnormality maps in demyelinating disease: correlations with clinical scores. Eur J Radiol. 2012;81:e386---91. 32. Nakanishi A, Fukunaga I, Hori M, Masutani Y, Takaaki H, Miyajima M, et al. Microstructural changes of the corticospinal tract in idiopathic normal pressure hydrocephalus: a comparison of diffusion tensor and diffusional kurtosis imaging. Neuroradiology. 2013;55:971---6. 33. Dumas de la Roque A, Oppenheim C, Chassoux F, Rodrigo S, Beuvon F, Daumas-Duport C, et al. Diffusion tensor imaging of partial intractable epilepsy. Eur Radiol. 2005;15:279---85. 34. Kamagata K, Tomiyama H, Hatano T, Motoi Y, Abe O, Shimoji K, et al. A preliminary diffusional kurtosis imaging study of Parkinson disease: comparison with conventional diffusion tensor imaging. Neuroradiology. 2014;56:251---8. 35. Nair SR, Tan LK, Mohd Ramli N, Lim SY, Rahmat K, Mohd Nor H. A decision tree for differentiating multiple system atrophy from Parkinson’s disease using 3-T MR imaging. Eur Radiol. 2013;23:1459---66. 36. Kim MJ, Seo SW, Lee KM, Kim ST, Lee JI, Nam DH, et al. Differential diagnosis of idiopathic normal pressure hydrocephalus from other dementias using diffusion tensor imaging. AJNR Am J Neuroradiol. 2011;32:1496---503. 37. Bisdas S, Koh TS, Roder C, Braun C, Schittenhelm J, Ernemann U, et al. Intravoxel incoherent motion diffusion-weighted MR imaging of gliomas: feasibility of the method and initial results. Neuroradiology. 2013;55:1189---96. 38. Bisdas S, Braun C, Skardelly M, Schittenhelm J, Teo TH, Thng CH, et al. Correlative assessment of tumor microcirculation using contrast-enhanced perfusion MRI and intravoxel incoherent motion diffusion-weighted MRI: is there a link between them. NMR Biomed. 2014:1184---91. 39. Federau C, Sumer S, Becce F, Maeder P, O’Brien K, Meuli R, et al. Intravoxel incoherent motion perfusion imaging in acute stroke: initial clinical experience. Neuroradiology. 2014;62:9---35. 40. Federau C, Hagmann P, Maeder P, Müller M, Meuli R, Stuber M, et al. Dependence of brain intravoxel incoherent motion perfusion parameters on the cardiac cycle. PLoS One. 2013;8: 1---7. 41. Puig J, Sánchez-González J, Blasco G, Daunis-I-Estadella P, Federau C, Alberich-Bayarri Á, et al. Intravoxel incoherent motion metrics as potential biomarkers for survival in glioblastoma. PLoS One. 2016;11:1---14. 42. Federau C, Brien KO, Birbaumer A, Meuli R, Hagmann P, Maeder P. Functional mapping of the human visual cortex with intravoxel incoherent motion (IVIM) MRI. PLoS One. 2015;10:e0117706. 43. Chen Y, Zhao X, Ni H, Feng J, Ding H, Qi H, et al. Parametric mapping of brain tissues from diffusion kurtosis tensor. Comput Math Methods Med. 2012:820847.

Advanced diffusion MRI and biomarkers in the central nervous system 44. Jensen JH, Helpern JA. MRI quantification of nonGaussian water diffusion by kurtosis analysis. NMR Biomed. 2010;23:698---710. 45. Van Cauter S, Veraart J, Sijbers J, Peeters RR, Himmelreich U, De Keyzer F, et al. Gliomas: diffusion Kurtosis MR imaging in grading. Radiology. 2012;263:492---501. 46. Tan Y, Wang XC, Zhang H, Wang J, Qin JB, Wu XF, et al. Differentiation of high-grade-astrocytomas from solitary-brainmetastases: comparing diffusion kurtosis imaging and diffusion tensor imaging. Eur J Radiol. 2015;84:2618---24.

285

47. Gong NJ, Wong CS, Chan CC, Leung LM, Chu YC. Correlations between microstructural alterations and severity of cognitive deficiency in Alzheimer’s disease and mild cognitive impairment: a diffusional kurtosis imaging study. Magn Reson Imaging. 2013;31:688---94. 48. Neto Henriques R, Correia MM, Nunes RG, Ferreira HA. Exploring the 3 D geometry of the diffusion kurtosis tensor --- impact on the development of robust tractography procedures and novel biomarkers. Neuroimage. 2015;111:85---99.