Keywords:
MR physics, MR, Imaging sequences, Tissue characterisation
Authors:
D. Cicolari1, D. Lizio2, P. Pedrotti2, R. Sironi2, M. T. Moioli2, A. Lascialfari2, M. Mariani1, A. Torresin2; 1Pavia/IT, 2Milan/IT
DOI:
10.26044/ecr2019/C-1108
Results
The NMR spectrometer results at 23°C are listed in Tab.
1.
Tab.
1. Relaxation times values of the Eurospin phantom inserts measured with the NMR spectrometer at 23°C.
Insert ID |
SR - T1 (ms) |
IR - T1 (ms) |
SE - T2 (ms) |
CPMG - T2 (ms) |
2 |
348 ± 6 |
294 ± 10 |
20 ± 1 |
50 ± 2 |
4 |
507 ± 9 |
427 ± 12 |
18 ± 1 |
33 ± 1 |
11 |
1039 ± 20 |
886 ± 44 |
17 ± 1 |
71 ± 1 |
13 |
2923 ± 78 |
2491 ± 79 |
21 ± 1 |
156 ± 3 |
14 |
2990 ± 50 |
2311 ± 53 |
21 ± 1 |
89 ± 1 |
15 |
1371 ± 19 |
1302 ± 65 |
20 ± 1 |
113 ± 2 |
From the NMR spectrometer measurements (Tab.
1),
IR T1 values showed on average an underestimation of 16% compared to SR T1 values.
T2 values measured with the spectrometer Spin-Echo sequence (Tab.
1) were affected by a high diffusion,
which suppressed the signal from the vials and it shortened the relaxation of every scanned vials to a value near 20 ms.
Thus,
values from SE sequences of the NMR spectrometer will not be taken as references: as reference values for the T2relaxation time we considered the CPMG results.
As can be seen in Fig.
2,
a Bland Altman analysis stated that results obtained with Cvi42 and Segment had a confidence level of ±0.7%,
meaning that the algorithms to generate maps were similar for the two software.
In order to compare T1 and T2 values of MRI results with those estimated with the NMR spectrometer (Tab.
1),
which are standard reference values,
the temperature difference of the various sites in which acquisitions were performed must be considered: room temperature of spectrometer site was 23°C,
Aera site was at 20°C and Signa compartment was at 21°C.
The comparison (Fig.
3) was performed by reconducting values measured from cvi42 maps (Aera 20°C,
Signa 21°C) at the same temperature of the standard reference values obtained by NMR spectrometer analysis (23°C),
assuming a linear dependence.
The comparison was not possible for T1 values of inserts 13 and 14 because of inaccuracy and imprecision due to the chosen sampling of the response curves (the image acquisition parameter TI should be distributed in a temporal range which must be at least three times longer than the T1 value of the scanned object).
Gnuplot manual fit procedure allowed the measurement of real fit errors of estimated relaxation times values.
In mapping procedure all the information about fit errors is lost,
as it consists in assigning a bit per pixel (BPP) value corresponding only to the estimated relaxation times values in a certain position: errors measured with Cvi42 and Segment are related only to the variability of fitted relaxation times values assigned to each pixel in a selected ROI of the maps,
while errors estimated with gnuplot reflect the goodness of the sampling of the response curves.