Perfusion MRI of gliomas - comparison of analysis methods

(1)

_________________________________________________________

Supervisors: Elna-Marie Larsson, Professor

Section of Neuroradiology Department of radiology Uppsala University Hospital Uppsala, Sweden

Master Thesis in Medicine

(Second Level, 45 Credits)

Perfusion MRI of gliomas – comparison of

analysis methods

(2)

ii

Perfusion MRI of gliomas – comparison of analysis methods

Keywords

Perfusion MRI, DSC-MR, Hotspot, Histogram, AIF, Glioma List of abbreviations used

 Magnetic Resonance (MR)

 Dynamic susceptibility MR (DSC-MR)

 Arterial Input Function (AIF)

 Relative blood volume (rCBV)

 High Grade Glioma HGG

(3)

iii

Abstract

Purpose: The purpose of this study was to compare analysis of perfusion MRI of gliomas of the brain using the “hot spot” method and the histogram method. A second purpose was to compare the results obtained after manual and automatic selection of region for arterial input function (AIF) determination.

Material and methods Twenty-eight cerebral gliomas patients (15 men and 13 women, mean age 48.4 age range 22- 70 years) were included in this study and examined by dynamic susceptibility contrast enhanced (DSC-MR) 3 T magnetic resonance (MR). Perfusion maps were created and normalized to non-affecting white matter. Tumour perfusion was analysed using the “hot spot” method with drawing of a region of interest in the portion of the tumour with highest perfusion. The value was divided by a value from a region in the contralateral normal appearing white matter. In addition, the entire tumours were segmented and gliomas were graded based on the histogram distribution by two observers. AIF determination was estimated manually and comparison made to the Automatic AIF. Binary Logistic regression was used to asset the diagnosis accuracy of two methods. t-test and Pearson’s correlation were used to compare between AIFs. This study has been approved by the local ethical committee, and informed consent was obtained from all patients. Results Histogram analysis diagnostic accuracy was improved compared to

the hotspot method, the sensitivity (94%) of histogram analysis was higher than of the hotspot (87%) method. Histogram interobserver agreement was almost perfect (_{ĸ = 0.89). AIF determinations showed} no significant difference and high correlation.

Conclusion Histogram analysis in an alternative method to the hotspot method for

glioma grading and automatic AIF determination is an alternative to

(5)

1

Introduction

Gliomas are the most common form of malignant primary brain tumours in adults (1). World Health Organization classified gliomas into four grades (I-IV) based on the malignancy degree and prognosis. Low-grade gliomas (LGGs), classified as astrocytomas, oligodendrogliomas, and oligoastrocytomas World Health Organization Grade II gliomas (2), are infiltrative tumours characterized by a slow progression for several years, and with a high risk of anaplastic transformation into HGGs, which induces poor prognosis and death (3,4). High-grade gliomas (HGGs), classified as astrocytomas, oligodendrogliomas, and oligoastrocytomas, World Health Organization Grade III, and glioblastoma multiform World Health Organization Grade IV, are malignant gliomas and consider the most frequent and lethal tumours of the central nervous system, and grade IV glioma are the most aggressive biological subtypes, characterized by aggressive infiltrative growth and very poor prognosis in spite of intense therapeutic efforts (5). They have an annual incidence of approximately 4-5 per 100.000 (1), and the approximately 5years survival rate is 30% (6).

Glioma treatment relies on the accurate and early diagnosis and classification (5). The treatment of LGG is resection or regular monitoring, while the treatment of HGG mostly is resection followed by radiotherapy and chemotherapy (7,8). Taking into account that the radiological growth rate of low grade gliomas is approximately 4mm/year (9), early and accurate glioma grading is important for patient management and treatment planning.

Histologic analysis is the standard method to grade gliomas. However, this method involves invasive technique, and involves sampling the heterogeneous glioma tissue and this is prone for sampling error (10), furthermore, the continuous tumour growth might demand several biopsies for the same patient, additionally, not all of glioma patients are operable.

Magnetic resonance imaging (MRI) has become the technique of choice for characterization and classification of brain tumours. MRI is useful for diagnostic evaluation of the anatomy (location), size and vascular detailed information of brain tumours. Although morphological MRI is widely used and preferred to detect and diagnose brain tumours, studies showed that morphological MRI is not reliable for the grading of tumour malignancy (11,12). Furthermore, HGG contrast enhancement may fail resulting in limited functional information about the tumour (11,13).

Perfusion-weighted MR includes using the first bolus contrast medium injection during T2*-weighted acquisition to track the change of signal decrease during data acquisition (14),

(6)

2 called Dynamic Susceptibility Contrast MRI (DSC-MRI). DSC-MRI exploits the contrast material to change the magnetic susceptibility of the tissue. Cerebral blood volume CBV, and cerebral blood flow CBF can be calculated and this method can improves tumour grading (15). Furthermore, studies have found that there is a correlation between the maximal CBV and the glioma grade, the higher rCBV the higher tumour grade (16). The reason for increased perfusion parameters in the tumour compared with the unaffected surrounding tissue is the increased metabolism in the tumour, therefore rapid growth of the tumour causing hypoxia resulting increase in CBV and CBF when the physiological auto-regulation is intact.

Evaluation of low-grade gliomas (LGGs) and high-grade gliomas (HGGs) by generating nCBV maps performing is the measurement of the ratio between the highest value of relative CBV of the tumour and the value of relative CBV of unaffected white matter this referred to Hotspot method (17).

Hotspot method or region of interest (ROI) is the most common method to investigate about tumour grade by using DSC-MR images. However, this method is highly user dependent due to the selection of the most elevated rCBV in the tumour tissues which is characterized by infiltrating macro-vessels, and the selection of reference white matter is prone for sampling error due to the assumption that most gliomas are located in the white matter. Arterial input function (AIF) is an important factor of CBF maps generation to determine the contrast media injection arrivals of the intravascular tracer of the tissue. The AIF determination is based on the measurement of signal intensity changes in major supplying artery (i.e. cerebral artery) by selecting a few pixels, and this process usually performed manually(18). The manual determination relies on operator’s knowledge of anatomy of brain vessels and circulation. Automatic AIF determination is an alternative method based on ranking pixels intensity signal variance; earliest, largest, and latest signal intensity by displaying the Deconvolution curve.

The purposes of this paper are; to analyse perfusion scans using the “hot spot” method with drawing of a region of interest in the portion of the tumour with highest perfusion. The value will be divided by a value from a region in contralateral normal appearing white matter. Comparison will be made with analysis using the histogram method based on the histogram analysis of nCBV of the entire glioma tumour. A second purpose is to compare the nCBV ratio after automatic and local AIF determinations.

(7)

3

Material and methods

Imaging data

Imaging data from the neurosurgery department, Uppsala university hospital acquired between February 2010 and September 2012 were used for the evaluation of the diagnostic accuracy of grading the brain tumour glioma. The following criteria were used to include patients in this study: the diagnosis was cerebral glioma, the glioma grade was determined with histopathological analysis, and all the examinations included DSC-MRI of the brain. Consent form was given to all participants, and Regional boards for research ethics have approved this research.

Brain MRI scans of 28 patients were used to assess the diagnostic accuracy of the alternative perfusion evaluation methods; hotspot and histogram; (15 men and 13 women, mean (sd) age 48.4 (15.5) years age range 22 – 70 years at the scan time.

MR imaging and postprocessing MRI Techniques

Scans were obtained on a 3 Tesla (T) scanner (Philips Achieva, Philips Medical Systems, Best, Netherlands). The morphological MRI sequences included sagittal and axial T2-weighted turbo spin echo (SE) (TR/TE = 3,000/80 ms; slice thickness/gap=4-5 mm/0.8-1 mm), coronal and axial T2-weighted fluid attenuated inversion recovery (FLAIR) (TR/TE/TI = 11,000/125/2,800 ms; slice thickness/gap, 4-5 mm/0.8-1 mm), axial T1-weighted SE before and after intravenous gadolinium based contrast injection (TR/TE=600/10 ms; slice thickness/gap, 5 mm/1 mm), sagittal T1-weighted 3D turbo field echo (TFE) after contrast injection (TR/TE: 8.1/3.7 ms; voxel size 1x1x1 mm).

Perfusion MRI

Perfusion MRI was acquired with a gradient echo (GRE) echo planar imaging (EPI) sequence using the dynamic-susceptibility contrast-enhanced (DSC) technique. The following scan parameters were used: TR 1,356 ms, TE 29 ms; in plane resolution 1.7 x 2.3 mm; slice thickness /gap 5 mm/1 mm; 23 slices; 70 dynamic scans. A standard dose of 5 ml gadolinium-based contrast agent (1.0 mol/l, Gadovist, Bayer Healthcare, Berlin, Germany) was injected intravenously at 4 ml/s using a power injector and was followed by a 30-ml bolus of saline at the same injection rate. This injection was performed 8-12 minutes after injection of 0.05 mmol/kg body weight contrast agent for a T1-weighted dynamic contrast

(8)

4 enhanced perfusion sequence. The first injection was also used as contrast agent preload of the extracellular space in tumour regions with possible blood-brain-barrier leakage.

Quantitative image analyses Hotspot analysis

The perfusion MRI had been performed simultaneously with contrast injection. (19). Relative cerebral blood volume (rCBV) maps were calculated based on an established tracer kinetic model applied to first pass data using commercial perfusion analysis software (Nordic Ice, NordicNeuroLab, Bergen, Norway) (20). Deconvolution of the measured signal-time curves was performed using singular value decomposition with arterial input function retrieved from middle cerebral artery branches in the hemisphere contralateral to the tumour. Correction for contrast agent leakage due to possible blood–brain barrier disruption was included in the post-processing (21). Relative CBV maps (rCBV) were coregistered with the T2 weighted images as overlay color images. Regions of interest ROIs were drawn in the highest signal intensity of the tumour on CBV maps normalized by the contralateral normal appearing white matter. Areas with necrosis, cysts, and large vessels were avoided. Tumour margins were defined on T2-weighted images, and for the contrast enhancement T1-weighted images were checked (Figure 1).

Histogram analysis

Delineation of brain tumours was performed on T1-weighted images, T2-weighted images, and/or Flair images synchronically with perfusion images using the NordicTumorEx software package (TumorEx, NordicNeuroLab, Bergen, Norway) DSC-MRI and other MRI

Figure 1: example of the ROI placement on rCBV maps overlaid on T1-weighted post-contrast images for tumour delineation of patient no.8. Image (a) T1-weighted post-contrast image, (b) the white arrow points to the ROI (red circle) placed in the region with maximum rCBF, (c) represents choice of the ROI for the reference region (white matter).

(9)

5 sequences were selected to ensure better tumour boundary visualization. Coregistration followed by segmentation of the brain tissue was applied. The coregistration was accomplished using an automatic algorithm optionally followed by manual corrections. When the DSC-series was selected, the pre-bolus images were used for the coregistration. Automatic coregistration uses a mutual information based algorithm to search an optimal rigid transformation that aligns the two datasets. The implementation is based on an article by Sundar H. et al. (22) Brain tissue was segmented automatically using either seed-growing, threshold, or clustering algorithms. Seed-growing and threshold are based on a logarithm that creates a volume based on the pixel intensity similarity, while clustering uses multiple logarithms with variance pixels intensity similarity. The automatic segmentation is done by multispectral analysis, and evaluation of pixel intensities in all of the structural sequences. Glioma tumours were segmented by applying thresholds at several levels by visual inspection, and then fine tuning and minor corrections was performed. The arterial input function factor was determined automatically by the software. Perfusion images were displayed as overlay mask over the structural sequences. Histogram results were presented together with histogram representative of tumours of high and low grade for comparison, the representative histogram data was included in the software and had been collected by the collaboration of the interventional centre, Oslo university hospital, and Rikshospital in Norway. This method also allows evaluation of the perfusion parameters by displaying volume of interest VOI. Histogram height and tumour size were recorded. Two observers, who were blinded to each other, classified the histogram output results. In order to get anonymous histograms without additional information about the subjects we displayed them anonymously by a program from the Microsoft Access, this program used as image displayer and it only display the histogram result, because we could not hide the patient ID from the patient list on the software.

AIF determination

The hot spot method includes options for AIF determination. AIF is a factor detects the contrast agent bolus shape at arrival. In order to evaluate the influences of applying the AIF factor by different methods, perfusion analyses were repeated twice as follows; the first method for AIF determination was automatic by selecting the Deconvolution curve, this curve must be positive, and has arterial shape with zero baselines, for the automatic method the Deconvolution of the time curve must be enabled so the AIF definition is available to adjust. The second AIF was determined locally (regionally) by looking for the high signal intensity pixel region to place the AIF region of interest over the middle cerebral artery, 5 pixels for determination of AIF were selected for all cases.

(10)

6

Figure 2: (a-c) normalized CBV maps on axial T2-weighted fast spin echo-MR images in patients with (a) grade II oligodendroglioma, (b.) grade III oligoastrocytoma, and (c) grade IV glioblastoma. Patients are 15, 12, and 20 respectively. (d.) the corresponding histogram signature derived from the total tumour volume of the relative patients.

Locations of both ROI of tumour and ROI of non-affected white matter were saved on the local drive with patient ID and the slice number, so when the perfusion analysis was performed again we recalled the ROI position from the local drive for the same patient.

(11)

7

Statistical analysis

Binary logistic regression was used to assess the diagnostic accuracy and sensitivity for the hotspot and the histogram analysis methods, t-test was performed to assess the difference between the AIF ratios (AIF automatic and AIF local), Pearson’s correlation was used to get the correlation between the two AIFs, and Fleiss ĸ statistics agreement was used to assess the agreement between the two observers of the histogram based on whether the observers graded a glioma as grade II, grade III, or grade IV. A ĸ value of less than zero indicated poor agreement; a ĸ value of 0.00–0.20, slight agreement; a ĸ value of 0.21–0.40, fair agreement; a ĸ value of 0.41–0.60, moderate agreement; a ĸ value of 0.61–0.80, substantial agreement; and a ĸ value of 0.81–1.00, almost perfect agreement (23). Statistical analysis was performed by using SPSS13 (SPSS, Chicago, Ill), and Microsoft Excel 2010.

Results

Out of the twenty-eight gliomas, 10 patients were histological graded as HGGs (World Health organization grade III and grade IV), and 18 were histologically graded as LGGs (World Health Organization grade II) (table 1). On average, 12 minutes was needed for analysis of each patient when using the hotspot method, and 16 minutes was needed for each patient when using the histogram analysis method.

Histological diagnosis

The histolopatological diagnosis was based on stereotactic biopsies in 7 patients, and on surgical section specimens in 21 patients. All tumours were graded according to the World Health Organization WHO classification system (24). Astrocytomas were classified as grades II (n=9), and grade III (n=4). Ganglioglioma was classified as grade II (n=1). Oligodendrogliomas were classified as either low-grade (grade II; n=8) or anaplastic (grade III; n=1). Mixed gliomas (oligoastrocytomas) were classified with regard to anaplastic astrocytic components into LGG (n=1), and HGG (n=2). Glioblastoma multiforme was classified as grade IV (n=2). Grade II tumours were designated as low-grade gliomas, and grade III and IV tumours were designated as high-grade gliomas.

Morphological MRI

Conventional contrast material enhanced MR imaging of the entire sample (n=28) are summarized in table 1, all lesions had supratentorial location and involved cortical brain

(12)

8 1,30 1,80 2,30 2,80 3,30 3,80 4,30 4,80 5,30 5,80 HGG LGG nCB V (ml/1 00 g ) 0,00 0,02 0,04 0,06 0,08 0,10 0,12 0,14 0,16 0,18 0,20 0,22 0,24 HGG LGG H isto gr am p eak ( ml /10 0g )

Figure 3: (a.)Box plots showing the distribution of rCBV max ratio from the all subjects. (b.)Box plot shows the distribution of histogram peak height.

areas. Seven tumours (patients are; 13, 20, 21, 25 36, and 37) showed minimal contrast enhancement on T1-wieghted images.

Glioma Grading

Maximum nCBV values of LGG and HGG were 2.20 and 5.70 ml/100 g respectively (figure 3a). Maximum histogram values of LGG and HGG were 0.19 and -0.10 ml/100 g respectively (figure 3b). We found no significant differences between the hotspot and the histogram methods (p=0.13). Diagnostic sensitivity and specificity of the two methods were as follows: sensitivity (94 %) and specificity (84 %) of the histogram method was higher than the sensitivity (87%), and specificity (76%) of the hotspot method.

Hotspot

nCBV values of LGG ranged from (1.40-2.30 ml/100 g), and nCBV of HGG ranged from (3.90-5.70 ml/100 g) (figure 3-a).

The measurements of nCBV as displayed in (figure3-a) show that HGG have higher rCBV value compared with LGG.

Histogram

Histogram peak of LGG ranged from 0.05-0.19, and the Histogram peak for HGG ranged from 0.01-0.10 (figure 3-b). The histogram peak indicated that HGG tumours tend to have a low peak with a broad range of nCBV values along the

x-axis of the histogram, while the histogram peak of the LGG is high with narrow range of nCBV values along the x-axis. The agreement between the two observers of histograms was almost perfect ĸ value was 0.89.

b. a.

(13)

9 0 1 2 3 4 5 6 0 2 4 6 A IF A u to mati c AIF Local AIF automatic vs, AIF local

Grade II Grade IV Grade III

AIF

in order to get result of AIF detremeniation, we grouped our subjects into three groups based on the tumour grade as follow; Grade II Grade III, and Grade IV. There was no significant difference between AIF local and AIF Automatic for any of the groups (paired t-test, PGradeII=0.11, PGradeIII=0.19,

PgradeIV=0.16), demonstrating that both

methods provide similar rCBV ratios. Additionally we also found highly significant correlation between rCBV values when AIF is local and Automatic (r

GradeII=0.84, r GradeIII=0.98, r GradeIV=0.89)

demonstrating that AIF could be set

Automatic (figure4).

Disscussion

In this paper we evaluated the diagnostic accuracy of the alternative methods hotspot and histogram, and the AIF factor determination. The sensitivity and specificity (94 %, 84%) of the histogram method was higher than those of the hotspot method (87% and 76%). The hotspot method was difficult to perform due to the heterogeneity of glioma tissue. Operator dependence in the hotspot method is represented by selection of tumour tissue and reference region in the non-affected white matter keeping in mind that the ROI should not cover intratumoural and extratumoural vessels and necrosis which is highly operator dependent. In histogram analysis, the operator only needs to select the intracranial area (coregistration) and highlight the entire glioma tumour (segmentation) indicating that inexperienced operators may obtain perfusion metric using the histogram analysis that is comparable to those obtained by experienced operator using hotspot analysis. Beside that the agreement of the two observers of histogram was almost perfect.

Figure 4: Pearson’s correlation between the automatic AIF and local AIF determinations.

(14)

10 Several methods have been shown to increase inter-and intraobservers reproducibility when using hotspot methods. (17). Emblem and colleagues (25) reviewed 53 glioma patients of whom 24 were LGGs and 29 were HGGs and four observers, where the histogram method was used to determine tumour grade. Using histological diagnosis as the gold-standard, the methods was found to have a sensitivity and specificity of 90% and 83 %, respectively, and diagnosis accuracy was independent of observer. In comparison, the hot spot method, also tested in the same study, had a sensitivity and specificity of 55 – 76 % and 63 – 88 %, respectively, with much larger inter-observer variation.

rCBV measurement of LGG was (1.40-2.30 ml/100 g), and (3.90-5.70 ml/100 g) for HGG. Our result was a bit higher than in the previous published study. Comparison of rCBV measurements between LGG and HGG has demonstrated LGG to have maximal rCBV values of between 1.11 and 2.14, and HGG to have maximal rCBV values of between 3.54 and 7.32 (20,26-27). Law and colleagues (20) investigated rCBV in 160 patients and found LGG to have maximal 2.14 and HGG to have maximal rCBV of 5.18.

Histogram peaks indicated that HGG tumours tends to have low peak with a broad range of nCBV values along the x-axis while the histogram peak of the LGG is high with a narrow range of nCBV values. Histogram height fail to predict the glioma grade, our result of histogram height found no relationship between the tumour grade and the histogram height. However, the histogram shape with parametric analysis could improve glioma grading rather than histogram height as seen in figure 5.

In our MRI sequences images, gradient echo echo-planner imaging used because it's more sensitive to detect paramagnetic changes in the tissue magnetic susceptibility resulting reduction of T2*, however spin echo sequences are more sensitive to the small capillaries and less susceptibility to the artifacts and excellent signal to noise ratio to overcome the artefacts of gradient echo we reduce the slice thickness and used 0.1mml/kg of contrast to get the same magnitude when using spin echo with 0.2mml/kg of contrast(28).

Arterial input function AIF result: we found no significant differences between the two methods. Additionally, we found highly significant correlation when using AIF automatic determination and AIF local. AIF local determination is based on the selection of few pixels when placing the region of interest over the middle cerebral artery, the image is prone to image noise and pixels spurious pixels value. Furthermore, the local selection of AIF is based

(15)

11 on the assumption that contrast media (bolus injection) reaches all parts of the brain tissue at same time relies on the assumption that the bolus injection does not disperse on its path from the major arteries to the brain tissue.

Our study has some limitation; it would have been preferable to include more patients with histologic analysis to strengthen the statistical power. Other limitation represent by that only two observers evaluate the histogram results.

Conclusion

The present study demonstrates that the histogram method may be more useful to grade gliomas than the hotspot method. The histogram method had higher sensitivity and was easier to use. AIF determination could be set automatically, which is both time-saving and more objective.

Acknowledgment I take this opportunity to express my profound gratitude and deep regards

to my guide my supervisor Prof. Elna-Marie Larsson for her extremely guidance, monitoring and constant encouragement throughout the perfusion project, and for Dr. Anna Falk for being a second observer of histogram.

(16)

12 Figure 5: Histogram analysis with (a.) grade II oligodendroglioma, (b.) grade III oligoastrocytoma, and (c.) grade IV glioblastoma multiform, (d) corresponding histogram of the 3 patient for comparison. Patients are 15, 12, and 20 respectively

d. .

(17)

13 Tab le 1 C lin ic a l c h ar ac te ri sti cs, H o tsp o t an d H isto gr am f in d in gs o f th e e n ti re stud y sampl e ( n =28) Pa ti en t/ sex /a ge ** Hi st ol ogy Tu m or lo ca ti o n Tu mo r vo lu me (c m3 ) n C B V r at io in ROI Hi sto gr am p ea k 2 /M /6 2 OAI I R.P ari et al 1 7 .5 3 1,84 0 .28 3 /M /6 7 AI I L.F ro n tal 3 1 .3 6 1,21 0 .22 4 /M /5 9 GGI I L.Te m p o ral 2 7 .2 3 1,08 0 .14 5 /F /6 6 AI I L.Occup ital 2 2 .4 3 4,45 0 .15 7 /F /4 0 OII R.F ro n tal 2 3 .9 6 2.30 0 .18 1 0/ F/ 2 8 OAI I L.F ro n tal 6 4 .8 7 4,36 0 .24 1 2/ M /3 2 OAI II L.Te m p o ral 4 .40 3,93 0 .20 1 3/ F/ 3 6 AI II L.Parie tal 6 5 .2 4 4,81 0 .22 1 4/ F/ 3 1 OIII R.F ro n tal 6 6 .6 2 2.04 0 .01 1 5/ M /4 3 OII L.Te m p o ral 4 7 .4 4 1.98 0 .28 1 7/ M /2 5 OII L.Parie tal 2 8 .6 6 2.22 0 .24 1 8/ F/ 7 8 AI II L.F ro n tal 1 9 .2 6 3,95 0 .17 2 0/ M /6 9 GBI V R .Te m p o ral 6 .19 5,91 0 .13 2 1/ M /5 4 AI II L.F ro n tal 3 8 .0 0 5,16 0 .18 2 3/ F/ 6 3 AI I R.Te m p o ral 1 3 .8 9 1,44 0 .17 2 4/ F/ 2 2 AI I R.F ro n to -P ari etal 4 9 .4 9 1,39 0 .38 2 5/ M /7 0 AI II R.Te m p o ral 5 5 .7 8 4,62 0 .18 2 6/ F/ 4 0 OII R.P ari et al 3 .29 4,55 0 .46 2 7/ M /5 2 OII R.P ari et al 1 0 .8 0 1 .9 0 0 .18 2 8/ M /4 1 AI I R.Te m p o ral 1 1 .2 9 1,56 0 .24 2 9/ F/ 4 3 AI I L.F ro n to -Te m p o ral 1 9 .6 9 1,67 0 .24 3 0/ F/ 5 8 AI I L.Te m p o ral 2 4 .3 4 1,64 0 .20 3 1/ M /4 4 OII L.Parie tal 4 5 .7 3 2.13 0 .27 3 3/ F/ 5 6 AI I Fr o n tal 3 7 .4 5 1,02 0 .21 3 4/ M /3 3 OII L.F ro n tal 3 0 .9 1 2.20 0. 25 3 5/ M /6 0 AI I L.F ro n tal 1 5 .7 1 1,69 0 .28 3 6/ M /5 0 GBI V L.Te m p o ral 1 0 .5 7 5,48 0 .14 3 7/ F/ 4 3 OII R.F ro n tal 2 2 .2 3 1,99 0 .23 – A II/III astr oc yt o m a gr ad e II/ III, – OAII/ III oli goastr oc yt o m a gr ad e II/ III, – OII /II I oli god en d rogli om a gr ad e II/ III, – GG II gan gli ogli om a gr ad e II, – G B IV gliob lasto m a m u ltif or m . -T u m or lo cat ion : R Righ t, L lef t. ** P atient ’s no . is b ased o n the entire p atient gli o m a p o p u latio n in the n euro sur gical d epar tm ent in Up p sala Un ive rsity h o spi tal.

(18)

14

References

1 Wen PY, Schiff D, Cloughesy TF, Raizer JJ, Laterra J, Smitt M, et al. A phase II study evaluating the efficacy and safety of AMG 102 (rilotumumab) in patients with recurrent glioblastoma. Neuro Oncol. 2011;13(4):437-46.

2 Louis DN, Ohgaki H, Wiestler OD, Cavenee WK, Burger PC, Jouvet A, et al. The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol. 2007;114(2):97-109

3 Cavaliere R, Lopes MB, Schiff D. Low-grade gliomas: an update on pathology and therapy. Lancet Neurol. 2005;4(11):760-70.

4 Wessels PH, Weber WE, Raven G, Ramaekers FC, Hopman AH, Twijnstra A. Supratentorial grade II astrocytoma: biological features and clinical course. Lancet Neurol. 2003;2(7):395-403.

5 Stupp R, Hegi ME, van den Bent MJ, Mason WP, Weller M, Mirimanoff RO, et al. Changing paradigms--an update on the multidisciplinary management of malignant glioma. Oncologist. 2006;11(2):165-80.

6 Dolecek TA, Propp JM, Stroup NE, Kruchko C. CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2005-2009. Neuro Oncol. 2012;14 Suppl 5:v1-49.

7 Danchaivijitr N, Waldman AD, Tozer DJ, Benton CE, Brasil Caseiras G, Tofts PS, et al. Low-grade gliomas: do changes in rCBV measurements at longitudinal perfusion-weighted MR imaging predict malignant transformation? Radiology. 2008;247(1):170-8.

8 Ricard D, Idbaih A, Ducray F, Lahutte M, Hoang-Xuan K, Delattre JY. Primary brain tumours in adults. Lancet. 2012;379(9830):1984-96..

9 Mandonnet E, Delattre JY, Tanguy ML, Swanson KR, Carpentier AF, Duffau H, et al. Continuous growth of mean tumor diameter in a subset of grade II gliomas. Ann Neurol. 2003;53(4):524-8.

10 Jackson RJ, Fuller GN, Abi-Said D, Lang FF, Gokaslan ZL, Shi WM, et al. Limitations of stereotactic biopsy in the initial management of gliomas. Neuro Oncol. 2001;3(3):193-200.. 11 Lev MH, Ozsunar Y, Henson JW, Rasheed AA, Barest GD, Harsh GR, et al. Glial tumor grading and outcome prediction using dynamic spin-echo MR susceptibility mapping compared with conventional contrast-enhanced MR: confounding effect of elevated rCBV of oligodendrogliomas [corrected]. AJNR Am J Neuroradiol. 2004;25(2):214-21.

12 Law M, Oh S, Babb JS, Wang E, Inglese M, Zagzag D, et al. Low-grade gliomas: dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging--prediction of patient clinical response. Radiology. 2006;238(2):658-67..

13 Liu X, Tian W, Kolar B, Yeaney GA, Qiu X, Johnson MD, et al. MR diffusion tensor and perfusion-weighted imaging in preoperative grading of supratentorial nonenhancing gliomas. Neuro Oncol. 2011;13(4):447-55.

14 Westbrook C, Roth CK, Talbot J, eds. MRI in Practice. 3rd ed. Oxford, United Kingdom: Wiley-Blackwell; 2005.

15 Knopp EA, Cha S, Johnson G, Mazumdar A, Golfinos JG, Zagzag D, et al. Glial neoplasms: dynamic contrast-enhanced T2*-weighted MR imaging. Radiology. 1999;211(3):791-8.

16 Morita N, Wang S, Chawla S, Poptani H, Melhem ER. Dynamic susceptibility contrast perfusion weighted imaging in grading of nonenhancing astrocytomas. J Magn Reson Imaging. 2010;32(4):803-8.

17 Wetzel SG, Cha S, Johnson G, Lee P, Law M, Kasow DL, et al. Relative cerebral blood volume measurements in intracranial mass lesions: interobserver and intraobserver reproducibility study. Radiology. 2002;224(3):797-803..

(19)

15 18 Sorensen AG, Reimer P. Cerebral MR perfusion imaging: principles and current applications. New York: Georg Thieme Verlag; 2000. 162 p

19 Ostergaard L, Weisskoff RM, Chesler DA, Gyldensted C, Rosen BR. High resolution measurement of cerebral blood flow using intravascular tracer bolus passages .1. Mathematical approach and statistical analysis. Magnetic Resonance in Medicine. 1996;36(5):715-25.

20 Law M, Yang S, Wang H, Babb JS, Johnson G, Cha S, et al. Glioma grading: sensitivity, specificity, and predictive values of perfusion MR imaging and proton MR spectroscopic imaging compared with conventional MR imaging. AJNR Am J Neuroradiol. 2003;24(10):1989-98.

21 Qu M, Olofsson T, Sigurdardottir S, You C, Kalimo H, Nistér M, et al. Genetically distinct astrocytic and oligodendroglial components in oligoastrocytomas. Acta Neuropathol. 2007;113(2):129-36.

22 Sundar H, Shen D, Biros G, Xu C, Davatzikos C. Robust computation of mutual information using spatially adaptive meshes. Med Image Comput Comput Assist Interv. 2007;10(Pt 1):950-8.

23 Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33(1):159-74.

24 Kleihues P, Burger PC, Scheithauer BW. The new WHO classification of brain tumours. Brain Pathol. 1993;3(3):255-68.

25 K. Emblem, B. Nedregaard, T. Nome, P. Due-Tønnessen, J. Hald, D. Scheie, B. O C, M. Cvancarova and A. Bjørnerud, “Glioma grading by using histogram analysis of blood volume heterogeneity from MR-derived cerebral blood volume maps,” Radiology, vol. 247, pp. 808-817, 2008.

26 Shin JH, Lee HK, Kwun BD, Kim JS, Kang W, Choi CG, et al. Using relative cerebral blood flow and volume to evaluate the histopathologic grade of cerebral gliomas: preliminary results. AJR Am J Roentgenol. 2002;179(3):783-9.

27 Yang D, Korogi Y, Sugahara T, Kitajima M, Shigematsu Y, Liang L, et al. Cerebral gliomas: prospective comparison of multivoxel 2D chemical-shift imaging proton MR spectroscopy, echoplanar perfusion and diffusion-weighted MRI. Neuroradiology. 2002;44(8):656-66.

28 Aronen HJ, Gazit IE, Louis DN, Buchbinder BR, Pardo FS, Weisskoff RM, et al. Cerebral blood volume maps of gliomas: comparison with tumor grade and histologic findings. Radiology. 1994;191(1):41-51.