‘REDUCTION OF SYSTEMATIC ERROR IN
MRSI BASED ABSOLUTE BRAIN
TEMPERARTURE MAPPING
Jehill Parikh1, Michael Thrippleton1 and Ian Marshall1,2
1.
2.
Brain Research Imaging Centre, Dept. of Clinical Neurosciences,
(part of SINAPSE collaboration),
Centre of Clinical Brain Science, University of Edinburgh
COMPUTER 61
Date 10 May 2012
Time 13.30
MR THERMOMETRY
•
Brain Temperature
Stroke, Traumatic Brain injury, Birth Aphasia
(neonates), Schizophrenia and Brain Cooling
•
Proton frequency Shift
Linear Temperature coefficient 0.01ppm/oC
Internal reference
MRS
Phase mapping
Absolute temperature
Possible
Changes
Spatial
Resolution
Low
High
Susceptibility errors
Low
High
Field inhomogeneity
errors
Low
High
•
MRSI TECHNICALITIES
–
–
FREQUENCY SELECTIVE WATER
SUPPRESSION (CHESS)
CHEMICAL SHIFT MIS-REGISTRATION
T =Tref - k (CSH2O -CSref)
METHODS
•
MRSI DATA ACQUIRED FROM
HOMOGENENOUS ROOM TEMPERATURE
PHANTOM AND VOLUNTEERS
•
GE 1.5T SIGNA scanner (GE Healthcare Slough UK)
•
2D PRESS-CSI SEQUENCE, TE-144ms Tr-1s,
imaging matrix 24*24 (interpolated to 32*32),
spectral width=1000Hz, acquisition time 9min
40sec, Axial slice 10 mm, FOV 320,voxel
volume-1ml was employed
•
Automatic first order shimming was used
•
CHESS based water suppression
–
–
Processing
3 fixed Bandwidth (BW) 3 RF pulses
WS level changed by varying the delay τ between
the CHESS and PRESS
Details Marshall et al, MRI 2006 pg 699-706
Systematic error in Temperature maps
Default water suppression settings BW=75Hz and τ=48ms, WS factor 90
Temperature estimated using T=37-100(CSwater-CSnaa)
A systematic error of 0.6C can be identified in Temperature maps
Date
19oct2010
1Nov2011
4jan2012
Tmean (0C)
20.3
22.6
21.93
TSD (0C)
0.27
0.18
0.20
Tphantom (0C)
20
22.0
22.0
Effect of varying BW of the CHESS pulses
BW:75Hz
BW:85Hz
BW:95Hz
BW:105Hz
Temperature (oC)
WS parameter
WS factor
scan#
τ (ms)
BW (Hz)
Mean
SD
Mean
%COV
1
48
75
21.93
0.2
84.7
5.25
2
48
85
22.07
0.14
51.1
3.46
3
48
95
22.15
0.14
38.5
3.91
4
48
105
22.24
0.13
32.7
3.88
The systematic error in temperature maps decreases by
increasing the BW of the RF pulse employed for the CHESS
Effect of varying delay τ between CHESS and
PRESS
τ :48ms
τ :68ms
τ :58ms
τ :78ms
τ :88ms
MRSI Temperature (oC)
WS parameter
WS factor
scan#
τ (ms)
BW (Hz)
Mean
SD
Mean
%COV
5
48*
75
22.24
0.18
88.6
4.57
6
58
75
22.36
0.09
24.6
1.19
7
68
75
22.45
0.09
14.5
1.25
8
78
75
22.51
0.08
10.3
1.04
9
88
75
22.61
0.07
8.1
0.94
Weakening the water suppression by increasing τ reduces the systematic
error and temperature maps acquired using weak WS are uniform.
No Water Suppression
Temperature (oC)
WS parameter
τ (ms)
BW
(Hz)
Mean
SD
7
68 (weak WS)
75
22.45
0.09
10
n-WS
75
22.91
1.07
11
48 (strong WS)
75
22.67
0.17
12
n-WS
75
22.72
1.03
scan#
Temperature maps acquired using no water suppression are not uniform
The presence of the side bands and baseline variations in non WS data
introduces errors in estimate of the frequency of the internal reference
(i.e. NAA) using AMARES.
Thus it is necessary to employ water suppression
Volunteer Scans
Volunteer
Volunteer 1
Volunteer 2
Volunteer 3
WS level
Temperature
WS
factor
mean
SD
mean
Strong (τ =48ms)
37.1
2.24
36.7
Weak (τ =68ms)
37.0
2.02
15.1
Strong (τ =48ms)
38.1
1.87
39.6
Weak (τ =68ms)
38.3
1.26
15.9
Strong (τ =48ms)
38.1
1.47
35.8
Weak (τ =68ms)
38.0
1.50
16.3
The systematic error cannot be identified in the
temperature maps acquired from volunteers
The SD (~1.5-2.240C) of the in-vivo temperature
estimate, is higher compared to systematic error
(~0.50C) as seen in the phantom data
Chemical Shift Mis-registration
Small but proportional to chemical shift and different for different metabolites
Temperature maps estimated by using different metabolites as internal reference
Metabolite
Temperature (oC)
Amplitude (I.U.)
Linewidths (Hz)
Choline
22.01 (0.20)
199.3 (2.99)
2.99 (0.26)
Creatine
21.7 (0.19)
231.2 (34.5)
2.07 (0.24)
NAA
21.94 (0.21)
287.9 (34.7)
1.37 (0.2)
CONCLUSION
• Water suppression is necessary for accurate temperature estimation
using MRSI data
• Systematic error is present temperature maps acquired with strong
CHESS based water suppression
• Weakening CHESS based water suppression reduced the systematic
errors in MRSI temperature maps.
• Systematic error in temperature maps cannot be identified in vivo
data from volunteers and is likely to be masked by the random in the
temperature estimates.
• Chemical shift mis-registration has little influence on MRSI
temperature mapping.
Acknowledgments
Radiographers team @ ‘BRIC’
Funding
Medical
Physics