1. Neuroradiology Department Bellaria Hospital Bologna
University of Bologna
Prof. Marco Leonardi
Neuroradiology Department
Bellaria Hospital Bologna 1 1

2. Considerations on Advanced MRI Techniques in studying brain gliomas.
A.Bacci, F. Menetti, P. Agati, R. Agati, M. Messia, S. Ascanio,
M. Leonardi 2 2

3. Bellaria Hospital 3T Installation - February 2004

4. Advanced MR Techniques
Since 2004 we have used our 3T equipment and the advanced MR Techniques (DWI, PWI, SV Spectroscopy) in studying brain tumours
Our 3T experience: a careful selection, setting up and optimization of the sequences used in order to avoid unsatisfactory results is mandatory
Could advanced MR Techniques increase the proportion of correctly diagnosed cases?

5. 3T 1.5T Higher S/N ratio Longer T1, shorter T2* and T2;
Reduction of r1 relaxivity (contrast media)
Increase of magnetic susceptibility
Increase of geometric distortion
“Chemical Shift” increase
More flow artifacts
More SAR
More acoustic noise
Increase of dieletric effect 5

6. 3T  1.5T Higher spatial resolution Higher temporal resolution
“Complex” studies
Different approach to the sequences
Pathologic Enhancement (SE, Flair or SPGR?): T1
Morphology: T2 (no gap)
Careful selection, setting up and optimization of the sequences used in order to avoid unsatisfactory results

7. 3T  1.5T Higher spatial resolution Higher temporal resolution
“Complex” studies
Different approach to the sequences
Pathologic Enhancement (SE, Flair or SPGR?): T1
Morphology: T2 (no gap)
Careful selection, setting up and optimization of the sequences used in order to avoid unsatisfactory results

8. 3T  1.5T Higher spatial resolution Higher temporal resolution
“Complex” studies
Different approach to the sequences
Pathologic Enhancement (SE, Flair or SPGR?): T1
Morphology: T2 (no gap)
Careful selection, setting up and optimization of the sequences used, in order to avoid unsatisfying results

9. When 3T? Intraaxial Tumours Microadenomas Epilepsy
White matter diseases
Angio-MR (Tricks, Spine..)
fMRI

10. Tumour Protocol 8 channels phased array coil
Sag SE T1-w
Ax Flair T2-w
Ax FRFSE T2-w
Ax SE T1-w
DWI
1H-MRS SV (TR 2000, TE 35)
PWI (SE-EPI)
Sag SE T1-w CE
Ax SE T1-w CE
Cor SE T1-w CE

11. Tumour Protocol 8 channels phased array coil
Sag SE T1-w
Ax Flair T2-w
Ax FRFSE T2-w
Ax SE T1-w
DWI
1H-MRS SV (TR 2000, TE 35)
PWI (SE-EPI)
Sag SE T1-w CE
Ax SE T1-w CE
Cor SE T1-w CE

12. SE T1-w on 3T? (Pathologic enhancement)
On 3T, Flair or SPGR T1-w sequences allow a better grey-white matter contrast than SE
However we have decided to use SE (as 1.5T) because of the improved sensitivity at:
pathologic enhancement after contrast media administration
haemoglobin degradation products

13. SPGR e/o Flair T1-w vs SE
Flair SE
SPGR

14. SE
SPGR

15. FSE T2-w at 3T (Morphology)

16. FSE T2-w on 3T (Morphology)

17. PWI
It is well known that a GE-EPI sequence is more sensitive to the first pass of gadolinium than a SE sequence
However we decided to use a SE-EPI sequence because of:
Less sensitivity to magnetic susceptibility artifacts along the cranial theca
Improved sensitivity only to contrast agent within the capillaries (i.e. neoangiogenesis)
(S.Cha et coll. Radiology 2002; 223:11-29)

18. Magnetic susceptibility artifacts
GE-EPI
SE-EPI

19. The first pass of gadolinium (Signal decreases)
SE-EPI
GE-EPI

20. Spectroscopy Single voxel, TR 2000, TE 35
better S/N ratio and spectral resolution
quantification of variation of the metabolites more than spatial distribution of metabolites (as in CSI)
Relatively long acquisition time (5 min.)

21. SV vs CSI

22. SV

23. Equipment (Complex studies)
2 Workstations RIS-PACS
2 Workstations ADW
2 Functool
1 Brainwave
1 Sage
1 PC Windows (BrainVoyager)
1 PC Windows (Access)
1 PC Windows
Presentation/Stim 2
Hardware for activation studies
1 PC Linux (LC-Model)

24. Work (still) in progress…
For example we are now evaluating:
other software to elaborate PWI studies such as Brainstat and DPTools (correction of the T1 effect)
Arterial Spin labelling
High Order Shimming (Spectroscopy)

25. Use of advanced MR Techniques (Our experience)
Individual cases
More systematic way, in population studies
Glioma grading
Recurrent/residual tumours vs treatment related changes (e.g. radionecrosis)

26. Use of advanced MR Techniques (Our experience)
Individual cases
More sistematic way, in population studies
Glioma grading
Recurrent/residual tumor vs treatment related changes

27. Glioma (grade?)

28. Diffusion
(High cellularity)

29. (High Neoangiogenesis)
Perfusion
(High Neoangiogenesis)

30. (Intralesional Necrosis)
Spectroscopy
(Intralesional Necrosis)

31. ????
Cho
NAA
Lip 31 31

32. Follow-up (Lymphoma) 32 32

33. Radiation injury or recurrent tumour?
Cho Cr
NAA mI
Lip 33

34. Recurrent tumour
17/07/08
Cho Cr
NAA mI
Lip
17/07/08 34

35. Use of advanced MR Techniques (Our experience)
Single cases
More systematic way, in population studies
Glioma grading
Recurrent/residual tumour vs treatment related changes

36. Glioma grading
MRI is capable of providing important morphological information but it is not very accurate in tumour grading (20- 30% of gliomas are misclassified) *
Could DWI, PWI and 1H-MRS studies increase the proportion of correctly diagnosed cases?
Is it possible to define threshold values?
*Neuroradiology 2002; 44: Radiology 1990;174: Radiology 1999:211: Neuroradiology 1992; 34: 36

37. MRI in gliomas
Li X et al. JMRI 16: , 2002 37 37

38. MRI in gliomas
Addition of 1H-MRS and perfusion information to MRI studies significantly increases the proportion of correctly diagnosed cases
Sensitivity, specificity, PPV and NPV to differentiate between high and low-grade gliomas (160 patients) with:
MRI: %, 65%, 86.1%, 44.1%
rCBV+Cho/Cr+Cho/NAA: 93.3%, 60%, 87.5%, 75.0%
Law M et al. AJNR 24: , 2003 38

39. Our Study Grade II (n=25) Grade III (n=21) Grade IV (n=31) : GBM
11 Astrocytomas
12 Oligodendrogliomas
2 Oligoastrocytomas
Grade III (n=21)
11 Anaplastic Astrocytomas
5 Anaplastic Oligodendrogliomas
4 Anaplastic Oligoastrocytomas
1 Gliomatosis
Grade IV (n=31) : GBM
Il nostro studio è stato effettuato tramite l’analisi retrospettiva di 77 pz ( ) con gliomi cerebrali… 39

40. Our experience Histopathologic evaluation
All patients underwent preoperative MR examination on a 3T system:
MRI DWI (ADC ratio) PWI (rCBV ratio) SV MRS
Histopathologic evaluation 40 40

41. ROI (ADC)

42. ROI (rCBV)

43. SV Spectroscopy

44. LC-model Analysis SD = 20

46. II grade vs III+IV grade
Statistical analysis
(Test Brunner Dette Monk)
II grade vs III+IV grade
necrosis
<0.001
Adc min
Adc max
n.s.
rCBV min
<0.01
rCBV max
Enhancement
Lac/Cre
<0.05
mI/Cre
(GPC+PCh)/Cre
(NAA+NAAG)/Cre
(Glu+Gln)/Cre
(Lip13a+Lip13b)/Cre
(MM09+Lip09)/Cre
NAA/Cho 46

47. Evidence of statistically significant differences between two classes (i.e. grade II and grade III+IV gliomas) is of limited practical utility due to overlapping values between different glioma grades
Our aim should be to identify threshold values to define the grade of a glioma in individual cases

48. Cart Software
It is a decision tree that automatically sifts large, complex databases, searching for and isolating significant patterns and relationships.
CART can reveal important data relationships (and threshold values) that could remain hidden using other analytical tools.

49. Grade II vs III+IV MRI (enhancement and necrosis)
Actual Class
Total Cases
Percent Correct
2 N=31
34 N=46 2 25
84,00 21 4 34 52
80,77 10 42
Total:
77,00
Average:
82,38
Overall % Correct:
81.82 49

50. Grade II vs III+IV MRI and Advanced Techniques
Lac/Cre 3.45
Actual Class
Total Cases
Percent Correct
2 N=29
34 N=48 2 25
96,00 24 1 34 52
90,38 5 47
Total:
77,00
Average:
93,19
Overall % Correct:
92.21
ADC min 1.09
rCBV max 2.02
mI/Cre 1.36 50

51. Grade II vs III+IV Advanced Techniques
Lac/Cre 3.45
Actual Class
Total Cases
Percent Correct
2 N=29
34 N=48 2 25
96,00 24 1 34 52
90,38 5 47
Total:
77,00
Average:
93,19
Overall % Correct:
92.21
ADC min 1.09
rCBV max 2.02
mI/Cre 1.36 51

52. Grade II vs III+IV MRI + DWI, PWI, 1H-MRS DWI, PWI, MRI 1H-MRS
Accuracy
0.82
0.92
Sensitivity
0.81
0.90
Specificity
0.84
0.96
PPV
0.91
0.98
NPV
0.68
0.83

53. WARNING
The sample (77 patients) analyzed is too small to be representative of the entire population (all patients affected by glioma)
For this reason these results cannot be generalized
As previously discussed we made precise choices about technical parameters

54. Conclusion I
It is not possible to definitively predict the histological grade of a glioma even if we use DWI, PWI and Spectroscopic parameters
However, our experience shows that the use of these parameters can be useful not only from a statistical point of view but also in the evaluation of individual patients, through identification of threshold values

55. Conclusion II It is necessary to: examine more patients
homogenize acquisition and analysis techniques
The aim is not to state the superiority of one technique above the other, but to improve the “overall diagnostic and therapeutic performance”

56. Thank you for your kind attention…
I would like to thank my colleagues of:
Neuroradiology, Neurosurgery and Neurooncology
and in particular , Prof. Patrizia Agati of Statstical Department, Bologna University
Thank you for your kind attention…

57. Statistical analysis Accuracy: quantity of correct diagnosis
Sensitivity: proportion of lesions determined to be high-grade on MRI of all high-grade lesions at histological evaluation
Specificity: proportion of lesions determined to be low- grade on MRI of all low-grade lesions at histological evaluation
PPV (positive predictive value): proportion of lesions correctly classified as high-grade of all classified as high-grade
NPV (negative predictive value): proportion of lesions correctly classified as low-grade of all classified as low-grade

58. Grade III vs IV MRI (enhancement and necrosis)
Actual Class
Total Cases
Percent Correct
3 N=19
4 N=33 3 21
76,19 16 5 4 31
90,32 28
Total:
52,00
Average:
83,26
Overall % Correct:
84,62
Accuracy
0,85
Sensitivity
0,90
Specificity
0,76
PPV
VPN
0,84 58

59. Grade III vs IV MRI and Advanced Techniques Lip13/Cre 20.66 Enh. 0,1
Actual Class
Total Cases
Percent Correct
3 N=19
4 N=33 3 21
80,95 17 4 31
93,55 2 29
Total:
52,00
Average:
87,25
Overall % Correct:
88,46
Enh. 0,1
Accuracy
0,88
Sensitivity
0,94
Specificity
0,81
PPV
VPN
0,89
ADC min 1.12 59

60. Grade III vs IV Advanced Techniques
Actual Class
Total Cases
Percent Correct
3 N=27
4 N=25 3 21
90,48 19 2 4 31
74,19 8 23
Total:
52,00
Average:
82,33
Overall % Correct:
80,77
Accuracy
0,81
Sensitivity
0,74
Specificity
0,90
PPV
0,92
VPN
0,70 60