Vall Hebron Research Institute

Author Of 1 Presentation

Imaging Poster Presentation

P0540 - Assessing the relationship between the ratio of T1- T2-weighted images and MRI sequences by using the structural similarity index (ID 1260)

Presentation Number
Presentation Topic



Background: The ratio of T1- T2-weighted images (T1/T2) has been proposed as a biomarker of tissue integrity, although comparison with MRI-derived tissue metrics has not been perfomed.


Objectives: In this study, T1/T2 has been compared to the T1/FLAIR, Phase-sensitive inversion recovery (PSIR), the fractional anisotropy (FA) and the orientation dispersion (ODI) in normal appearing grey, white matter and lesion mask, by using the structural similarity (SSIM) index. The SSIM is the product of 3 terms that take into account the luminance, the contrast and the structural information of the images compared. It ranges between [-1,1], with 1 indicating that the two images compared are identical.


Material and methods: MRIs of patients with a clinically isolated syndrome were included in this study. Images were acquired in a 3.0 T sytem (Trio, Siemens) using established acquisition protocols. The FA and ODI images were generated with the TBSS/FSL and NODDI/SPM toolboxes, respectively. The SSIM maps were calculated with the function implemented in Matlab. T1-weighted images were segmented with SPM12 and the lesion masks were generated from the FLAIR with the LST tool. Mean SSIM values were calculated over the normal-appearing grey (NAGM) and white (NAWM) matter, as well as in the lesion masks. The contribution of each term and the total SSIM were computed separately.


Results: When T1/T2 images were compared to T1/FLAIR, in NAGM and NAWM the luminance, contrast and structural terms ranged between [0.81-0.99]; for the lesion mask, between [0.61, 0.92]. When T1/T2 was compared to FA, both the luminance and contrast terms ranged between [0.72, 0.92] in the three compartments; while the structural term ranged between [0.49, 0.61]. When T1/T2 was compared to ODI, the luminance term was quite high for the 3 masks [0.72, 0.93], intermediate for the contrast term [0.58, 0.67] and very low and negative [-0.19, -0.11] for the structural term. Finally, regarding the PSIR comparison, the structural term showed a high correspondence in the 3 compartments [0.72, 0.79], while the luminance and contrast terms were below 0.01 for the 3 compartments.


Conclusions: As expected, T1/T2 and T1/FLAIR are highly equivalent, in all aspects. When T1/T2 was compared to FA and ODI, a poor structural correspondence was found, whereas when T1/T2 was compared to PSIR, the structural correspondence was very high, suggesting that the anatomical features are reproduced in a similar way in both sequences. The SSIM index could be used as a measure to assess the equivalency between MRI sequences in a more thorough way.