Книги по МРТ КТ на английском языке / MR Imaging in White Matter Diseases of the Brain and Spinal Cord - K Sartor Massimo Filippi Nicola De Stefano Vincent Dou

422

Proton MR spectroscopic imaging (1H-MRSI) is a technique that combines the excellent spatial localisation capabilities of MRI with the chemical information of MR spectroscopy. Since the chemical environment influences the resonant frequency of the atomic nuclei, protons from different compounds have different chemical shifts and appear as distinct peaks in the acquired MR spectrum. The main peaks observed in the healthy human brain at field strengths of 1.5 T or 3.0 T are, in order of resonance, myo-inositol (3.55 ppm), choline (3.2 ppm), creatine (3.02 ppm), glutamine and glutamate (2.1 ppm) and N-acetyl- aspartate (NAA) at 2.02 ppm. NAA is found in neurons, axons and dendrites; creatine (Cr) is found in neurons and glia; choline (Cho) is a building block of cellular membranes. Additional peaks may appear in pathologic areas of the brain: lactate (1.44 ppm) is a sign of hypoxia or necrosis; mobile lipids (1.4 ppm and 0.9 ppm) are found in areas of necrosis.

A multitude of 1H-MRS studies have been published in the literature in the last 15 years (Bruhn et al. 1989; Alger et al. 1990; Demaerel et al. 1991;

Fulham et al. 1992; Negendank 1992; Preul et al. 1996; de Edelenyi et al. 2000; Tamiya et al. 2000). These studies have consistently shown that choline signal is elevated in all tumour types because of altered membrane metabolism (Podo 1999; Ackerstaff et al. 2003). Choline signal increases with cellular density, and, according to some authors, also correlates with cellular proliferative activity (Ki-67) (Shimizu et al. 2000; Tamiya et al.2000). NAA signal loss occurs following substitution of neurons and its prolonging by neoplastic-cell invasion.Changes in creatine signal may vary with the tendency towards a mild increase in low-grade astrocytomas and depletion in the most undifferentiated types.

Several hypotheses have been tested by multiple investigators. The hypothesis that accumulation of lactate may correlate with tumour grade was one of the first investigated. As originally described by Warburg, neoplastic cells may develop bioenergetic aberrations such as elevated anaerobic glycolysis (Warburg 1956). This phenomenon is mainly characteristic of higher grade tumours that have lost aerobic cell respiration, with their metabolism depending mostly on inefficient “anaerobic glycolysis” with relatively higher production of lactate. 1H-MRS studies have shown that lactate accumulation may occur in gliomas; however, it is found only in a minority of tumours, irrespective of grade (Alger et al. 1990). This disappointing finding can be explained with the consideration that efficient lactate clearance in the venous blood stream can prevent lactate accumula-

A. Bizzi et al.

a

b

c

Fig. 28.7.2 a–c

Neoplastic Disorders

d

423

Fig. 28.7.2a–d. Axial SE PD-weighted (a) and coronal FLAIR (b) MR images showing a mass located in the right inferior frontal, temporal and insular region of this 47-year-old male. There is a moderate mass effect, with shifting over the midline. There was no enhancement after I.V. injection of contrast medium on SE T1-weighted MR image (c). H-MRSI (d) (multivoxel PRESS sequence; TR/TE=1,500/136 ms; 32×32 matrix; FOV=200×200×20 cm3) shows a homogeneous metabolic lesion with borders well-defined by T2weighted MR signal abnormalities. The spectra show marked choline signal elevation with moderate loss of creatine and almost complete depletion of NAA. There is mild lactate accumulation in a few voxels in the centre of the mass. There is no accumulation of lipids. The volume of spectra displayed (d) are indicated in the reference T2-weighted SE MR image. The diagnosis of anaplastic astrocytoma (WHO III) was confirmed on the neuropathologic specimen after surgery

424

tion. Increased lactate is often found within necrotic pseudocysts and in areas of the tumour where venular outflow is obstructed. Lactate is sometimes detected inside a cyst or a necrotic core of glioblastoma and metastasis.

The choline signal has become a novel, important biochemical indicator of tumour progression (Fulham et al. 1992) and response to therapy (Bizzi et al. 1995). Precursors and catabolites of phospholipid metabolism are altered in tumours. The“choline peak”detected by clinical 1H-MRS is composed of free choline and several choline-containing compounds, such as phosphocholine (PC) and phosphoethanolamine (PE), glycerol 3-phosphocholine (GPC) and glycerol 3-phosphoethanolamine (GPE). Elevation of intracellular phosphomonoesters (PME) PC and PE suggests enhanced cell-membrane synthesis, cellular growth and nutritional state, while increased phosphodiesters (PDE), GPC and GPE represent altered rates of membrane synthesis, catabolism and metabolic turnover. Diverse molecular alterations such as oncogene expression and malignant transformation arrive at common endpoints in choline phospholipid metabolism. They often determine a shift in GPC and PC that results in increasing PC/GPC ratio and total choline-containing metabolite level. Experimental data suggest that increased PC/GPC is primarily related to malignant degeneration. Conversely, reduced PC/GPC is related to growth arrest (Bhakoo et al. 1996). Extract studies have also shown that PC elevation is likely related to malignancy, since it is found at a twofold greater level in high-grade glioma compared with low-grade and normal tissue (Usenius et al. 1994). Therefore, MRS appears to be more sensitive to up-regulation of the anabolic pathway than to acceleration of catabolic pathways.

In vivo 1H-MRSI has shown that choline signal is increased in the solid portion of brain tumours. One study in 18 glioma patients demonstrated a significant linear correlation between normalised choline (nCho) signal and cell density, not with proliferative index (Gupta et al. 2000). In the same study a significant inverse linear correlation between cell density and apparent diffusion coefficient (ADC) measured by diffusion MR was also demonstrated. Choline is consistently low in areas of necrosis. A progressive increase in choline signal as a tumour undergoes malignant degeneration has been reported (Tedeschi et al. 1997). The finding of elevated choline signal in the enhancing rim of a mass does not allow solving the differential diagnosis between GBM, metastases or lymphomas. On the contrary, the identification of high choline in the T2-weighted hyperintense peri-

A. Bizzi et al.

a

b

c

Fig. 28.7.3 a–c

enhancing area is suggestive of GBM rather than metastases (Law et al. 2002).

The finding of abnormal elevated MRS signals resonating at 1.4 ppm and 0.9 ppm indicates presence of lipid droplets in areas of extracellular necrosis (Zoula et al. 2003). Mobile lipids are often found in the necrotic core of GBM, lymphomas and metastases (Kuesel et al. 1994; Poptani et al. 1995; Sijens et al. 1996). Primary cerebral lymphoma in immunocompetent patients mimics the infiltrative behaviour of glial neoplasms. The recognition of lymphoma is difficult with conventional MRI, and it has important diagnostic and therapeutic implications, since chemotherapy and radiotherapy are the treatment of choice and surgery is not effective. The demonstration of high choline with massively elevated lipid resonances associated with absent creatine and NAA signals is the hallmark of cerebral lymphoma (Bizzi et al. 1995; Harting et al. 2003), but it is not found in all cases. Lipids can also be found in large amounts in areas that have been treated with radiotherapy and have evolved into areas of delayed radiation necrosis.

Questions remain whether 1H-MRSI is a useful technique to determine in vivo the histopathological grade of gliomas. Most multivoxel (1H-MRSI) studies have shown that higher grade tumours have a tendency to show higher choline levels compared with low-grade tumours (Fulham et al. 1992; Preul et al. 1996; Li et al. 2002). This is especially true if the voxel with the maximum Cho/NAA ratio is considered in the analysis. In the evaluation of tumour grade we believe that the multivoxel (2D and 3D MR spectroscopic imaging; Li et al. 2002) is superior to the single-voxel technique for at least three reasons. Spectroscopic imaging evaluates spatial heterogeneity, offers a definition of the macroscopic boundary of the mass and is less subjected to partial volume effects. Spectroscopic imaging is helpful in characterising areas with T2-weighted signal hyperintensity: it can point to areas of higher cellular density and it can distinguish areas with prevalent vasogenic oedema from areas with neoplastic invasion or necrosis.

However, in the individual case, assignment of grade becomes difficult, mainly because of overlapping between grade II and III. The biology of glioma during the progression from diffuse astrocytoma to anaplastic then glioblastomas is a spectrum. In neuropathology the arbitrary separation in grades is practical due to clear and simple histopathological criteria that can be applied by all medical centres. So far, with MR spectroscopy it has not been possible to set cut-off ratios that could be reliably used and compared by multiple groups.

a

b

Fig. 28.7.3 a,b

It is then useful to consider that during the transition from diffuse to anaplastic astrocytoma the Cho signal progressively increases while NAA is decreasing, until the Cho/NAA reaches a plateau at blown anaplastic astrocytoma. Glioblastoma tends to show the highest Cho/NAA levels when they are still well perfused and oxygenated. Then the formation of hypoxic areas within the mass causes a significant drop of Cho signal and the appearance of lipid signals. In this spectrum the creatine signal also has a say. In well-differentiated and oxygenated astrocytomas creatine shows usually normal or slightly elevated levels, then its signal drops significantly. This depletion in creatine usually precedes choline signal drop.

Spectroscopic imaging is also particularly useful for defining target volumes for radiotherapy and for evaluating heterogeneous tumour response to therapy. Incorporation of MRSI into the treatmentplanning process may have the potential to improve control while reducing complications (Pirzkall et al. 2001).

In patients with high-grade gliomas the most common and significant clinical problem is the definitive

diagnosis of tumour recurrence as opposed to delayed radiation necrosis (DRN). 1H-MRSI has been shown to be useful for improving diagnostic acumen, when MRI cannot reliably differentiate between these entities (Taylor et al. 1996; Rock et al. 2002; Schlemmer et al. 2002). In cases of suspected recurrent tumours that have recently changed their MRI signal characteristics since the previous examination, the finding of increased choline suggests the di-

agnosis of recurrent tumour. Conversely, the absence of voxels with elevated choline within the volumes of T2-signal abnormality or contrast enhancement suggests the diagnosis of DRN. The lipid signal tends to be most elevated in areas of DRN or mixed tumour and necrosis than in pure tumour. 1H-MRSI cannot distinguish either pure tumour or pure necrosis from areas of mixed tumour and necrosis.

28.5

Perfusion MR Imaging

At the time of this writing, perfusion MR imaging is increasingly being used as a diagnostic and research tool. However, it is important to remark that it is still

relatively new and promising rather than a standard technique for evaluating tumour grading and malignancy. With very short imaging and data processing times and the use of a standard dose of gadolinium chelate, the differential scanning calorimetry (DSC) method can easily be incorporated into the routine clinical evaluation of brain tumours. Cerebral blood volume (CBV) measurements are a relative rather than an absolute quantification of blood volume. Relative CBV maps can be generated after selecting the arterial input function (AIF) in one of the main feeding vessels (i.e., middle cerebral artery).

In the interpretation of rCBV maps it is very important to correlate them with conventional MR images that may show areas of blood–brain barrier (BBB) disruption or T2* signal loss due to susceptibility effects within the tumour. A severe breakdown

of the BBB may cause inaccurate rCBV estimation in GBM, metastasis, meningioma or lymphoma. In the presence of these lesions it is appropriate to give the patient a small baseline dose of contrast agent before the perfusion study. While conventional T1-weighted MR images show areas of disrupted or absent BBB breakdown, variations in rCBV demonstrate geographic differences within the tumour and are a sign of heterogeneous vascularity.

Several recent studies have found a significant correlation between tumour rCBV and glioma grade (Aronen et al. 1994; Knopp et al. 1999; Roberts et al. 2000; Lev et al. 2004). In a study of 22 patients, all high grade (III and IV) gliomas had normalised CBV values greater than the defined cut-off ratio of 1.5 (Lev et al. 2004). High CBV value was found in a significant number of grade II gliomas that histopa-

thology later showed to be oligodendrogliomas. In the same study correlation with survival was better for CBV than enhancement on T1-weighted MR images. The value of maximum rCBV in a preoperative study of histologically proven GBM may vary over a wide range, but the mean is usually higher than maximum rCBV measured in grade III and II gliomas. This is explained by the larger vascular heterogeneity of GBM with frequent areas of necrosis. The absence of necrosis and the lower vascular heterogeneity explains the lesser variation in maximum rCBV in anaplastic astrocytomas. The finding of low rCBV values in non-enhancing gliomas would add evidence to the preoperative diagnosis of a welldifferentiated low-grade II astrocytoma (Fig 28.8.1), while high rCBV values will suggest an undifferentiated anaplastic astrocytoma (Fig. 28.8.2). However,

it is important to emphasize once again that the transition from low-grade to high-grade gliomas is a spectrum without well-established boundaries and that there may be overlap of rCBV values between different glioma grades. In another recent study Law et al. (2003) evaluated sensitivity, specificity and predictive values of perfusion, proton spectroscopic and conventional MR imaging in 106 patients with known histopathologic diagnosis. They showed that rCBV and metabolite ratios (Cho/NAA and Cho/Cr) both individually and in combination can increase the sensitivity and positive predictive values (PPV) in determining grade when compared with conventional MRI alone. The best diagnostic predictor was rCBV. High-grade and low-grade gliomas can be distinguished at the same level of significance also with the arterial spin labelling (ASL) technique (Warmuth et al. 2003).A close linear correlation between DSC-MR and ASL in the tumour region of interest has been demonstrated. Blood flow is underestimated with ASL at low flow rates.

Another application of rCBV maps is guidance of biopsy sampling during stereotactic and open surgery. Foci of increased vascularity within nonenhancing brain glioma can be identified. Zones of higher vascularity are likely to represent the area of the tumour with the highest grade.

Soon et al. have shown that DSC-MRI is a valuable adjunct to conventional imaging also in assessing tumour response during therapy. They have shown that rCBV maps correlate better with clinical follow-up than do gadolinium enhanced T1-weighted MR images, in 18 patients with recurrent GBM receiving an antiangiogenic drug (thalidomide) and carboplatin (Cha et al. 2000).

Preliminary studies have shown that rCBV maps can demonstrate differences in vascularity between delayed radiation necrosis (DRN) and recurrent tumour. These two conditions are easily distinguished by histopathology, but they are often indistinguish-

able on clinical and conventional imaging grounds. Neovascularity is a distinguished feature of most recurrent GBM and will show as an area of increased rCBV on DSC-MR images. Conversely, the presence of coagulative necrosis mixed with signs of extensive vascular injury is characteristic of DRN that will show up as areas of low rCBV on perfusion images. The ASL method has also shown promising results on the differential diagnosis of DRN vs recurrent GBM, especially in those lesions with areas of T2* magnetic susceptibility due to deposits of hemosiderin or calcium and in lesions near the base of the skull at the air–bone interface.

The contribution of perfusion MRI in the differentiation of GBM, metastasis and lymphoma is still controversial, and additional studies need to be done. As with MR spectroscopic imaging, the finding of increased rCBV in the area around an enhancing mass suggests the diagnosis of glioblastoma rather than a metastasis. In a study of 51 patients (33 gliomas, 18 metastases) the measured rCBV in the peri-tumoral region of high-grade gliomas and metastases was statistically significant: 1.31±0.97 (mean±SD) and 0.39±0.19, respectively (Law et al. 2002). The finding of low rCBV in an enhancing mass suggests the diagnosis of lymphoma, especially when the mass is located in the deep grey matter, subependymal periventricular regions or corpus callosum, which are common locations for primary cerebral lymphoma (Sugahara et al. 1999). On neuropathology lymphoma shows low vascularity despite invasion of the vessel lumina and around endothelial cells of the host. A study of 24 patients demonstrated that rCBV and analysis of the intensity-time curves may be useful in distinguishing primary lymphoma from glioblastoma (Hartmann et al. 2003). The maximum rCBV ratio in lymphoma was significantly lower than that for GBM (p<0.0001). Lymphoma showed a characteristic type of curve with a significant increase in signal intensity above the baseline due to massive leakage of contrast media into the interstitial space.