1. DW-MRI and MRS to Differentiate Radiation Necrosis and Recurrent Disease in Gliomas P100 Exams 4224, SV, 2x2x2cm MV Thomas Chong

2. Scans Conducted to Observe MRS Voxel Size Influence on S/N Single Voxel (SV) Comparisons  1x1x1 (1cm 3 )‏  3x3x1 (9cm 3 )‏  2x2x2 (8cm 3 )‏  3x3x3 (27cm 3 )‏ Larger Multi-Voxel (MV) Grid  2x2x2cm voxels  6x6 grid

3. Single Voxel MRS 1x1x1cm This and subsequent SV positions selected to be in physical middle of hemisphere, away from voids and boundaries. No discernible metabolite peaks. Why? Seems inconsistent with MV results. Could not get results with metabolite peaks using protocol settings on SV scan. 2-3 attempts

4. 3x3x1cm (9cm 3 ) Single Voxel, S36.6 Exam 4135

5. 3x3x1cm (9cm 3 ) Versus 2x2x2cm (8cm 3 ) Single Voxel, ~S30 Exam 4139Exam 4135

6. 3x3x3cm (27cm 3 ) Single Voxel, ~S26.0

7. 2x2x2cm (8cm 3 ) Versus 3x3x3cm (27cm 3 ) Single Voxel 2x2x2cm SV, Exam 4139 3x3x3cm SV, Exam 4224

8. Multivoxel Scan with 2x2x2cm Voxels, Exam 4224, S36.6, Vox30 Note: Spectra shows less background noise that 2x2x2cm SV spectra. Why?

9. Multivoxel Scan with 2x2x2cm Voxels, Exam 4224, S36.6, Vox 10

10. Multivoxel Scan with 2x2x2cm Voxels, Exam 4224, S36.6, Vox 12 & 27

11. INTREPRET Study Protocol INTERPRET group MRS data protocol  1.5T GE, Phillips, and Siemens scanners  Both short and long TE  SV volume 4-8cm 3 ; equivalent to cubes of widths 1.6 – 2cm  “whole study protocol took less than 30min” including MRI set for voxel placement, therefore number of scan averages were higher than in our protocol (2). Averaging helps reduce noise.  “N averages metabolites = 192-128” ?  “N averages water = 8-32” ? Tate A, et al 2006

12. INTREPRET Study Protocol Voxel Placement: “Whenever possible voxels were placed entirely within the lesion... avoiding contamination from normal tissue and oedema.” Data Processing  Some SV spectra created by combining MV voxels.  Custom program based on MRUI software package 1) Lineshape correction and zero-order phasing using water reference with Klose method 2) 0.8Hz exponential line-broadening 3) FFT processing 4) Water removal by HLSVD; five components removed within +-0.37ppm of water resonance

13. INTREPRET Study Protocol Data Processing, cont'd 4) Water removal by HLSVD; five components removed within +-0.37ppm of water resonance 5) Residual water suppression; points at 4.2-5.1ppm set to zero 6) Linear interpolation to 512 points over 1000Hz of Siemens and Phillips data 7) Spectrum alignment; maximum of choline peak shifted to 3.21ppm 8) Normalization of spectrum to Euclidian norm of peak heights.

14. Data-collection Difference Between INTREPRET and Our Protocol Highest impact difference between protocols appears to be that their scan focused only on region of lesion  INTERPRET used single voxel scans, more averages, larger voxel sizes  These contribute to improve S/N (except maybe the SV vs MV)‏

15. INTERPRET considered mainly glioblastomas, meningiomas, metastases, and astrocytomas grade II

19. Re-consideration of Current Scan Protocol Fact: Based on observations of S/N trends with voxel size and theory, we can say that S/N is improved by:  Larger voxel sizes, more averages Fact: Current protocol gives mostly unusable, low S/N data. We must change it to get data. Suggestion:  Increase voxel size, increase scan averaging  Reallocate scan time to focus on region of interest; can reduce grid size, # of slices; consider SV

20. Questions to Discuss When Considering Updating Protocol If switch to SV's, brains with multiple lesions would require multiple SV scans If switch to SV's, would be useful to collect data for a non-lesioned “control” voxel at consistent location, e.g. contralateral to lesion region. Switching to SV's would lose spatially changing metabolite ratio information Comprise with 3x3 MV grid centered at single lesion? E.g. 1.8x1.8x1.8cm, 3 slices, max avgs to fill protocol scan time. Have to consider scan time tradeoffs. Must answer why the 1x1x1cm SV test scan did not show any peaks

21. References Tate, et al, Development of a decision support system for diagnosis and grading of brain tumours using in vivo magnetic resonance single voxel spectra, NMR in Biomedicine, 2006: 19: 411-434. Tate, et al, Classification of brain tumours using short echo time 1H MR spectra, J of Magn Reson, 2004: 17: 164-175.