2398
A novel imaging biomarker for cancer from multicomponent T1 relaxometry1Institute of Neuroscience and Medicine, Research Centre Juelich, Juelich, Germany, 2Institute of Neuroscience and Medicine (INM-4), Research Centre Juelich, Juelich, Germany, 3JARA-BRAIN-Translational Medicine, Research Centre Juelich, Aachen, Germany, 4Institute of Neuroscience and Medicine (INM-11, JARA), Research Centre Juelich, Juelich, Germany, 5Department of Neurology, RWTH Aachen University, Aachen, Germany
Synopsis
We hypothesised that T1 relaxation curves sampled with very high temporal resolution reflect the existence of several environments in healthy and brain tumour tissue. Relaxation properties of healthy as well as tumour tissue, identified by FET-PET in a hybrid MR-PET environment, were investigated using a Look-Locker inversion recovery sequence sampled with a 17ms time resolution and 460 time points. The properties of normal appearing tissue were very similar in patients and healthy volunteers. In addition, a novel component was identified in brain tumour patients, which seems characteristic of the presence of tumour and oedema.
Introduction
Despite their diverse histological types, most brain tumours cause brain oedema, which is a significant cause of patient morbidity and mortality. Due to its high water content and long relaxation times, oedema can be well distinguished visually from normal tissue by mixed contrast MRI. However, the exact (quantitative) properties within and surrounding the tumour tissue are seldom measured in vivo. We hypothesized that: 1. analysis of T1 relaxation curves sampled with very high temporal resolution will reflect the existence of several environments in healthy tissue; 2. performing this analysis in tumour and oedema tissue will separate the properties of underlying tumour cells from those of still intact tissue and visualise tumour cells infiltration. To this aim we investigated the relaxation properties of tumour tissue, identified by FET-PET in a hybrid MR-PET environment, using a Look-Locker inversion recovery sequence sampled with a 17ms time resolution and 460 time points.Methods
Fourty seven patients (30 female, 46.5±0.8 years old) with a variety of malignant brain tumours underwent MRI and dynamic FET PET in a hybrid 3T MR-PET scanner. In addition, 10 healthy volunteers (5 female, 32.4±0.3 years old) were measured in the same setup, but without PET tracer injection. Quantitative MRI was performed using a Look-Locker sequence, TAPIR6,7, including mapping of the inversion efficiency. The inversion-recovery curved was sampled in steps of 17ms with 460 time-points within a single-slice TA of ~4 min (Table I). Multi-contrast magnitude and phase data were saved and processed off-line. A mono-exponential fit to the signal recovery as a function of inversion time7 delivered the longitudinal relaxation time T1 and signal intensity at TI=0. NNLS analysis using 200 logarithmically spaced T1 values between 50ms and 6s and Tikhonov regularisation was performed. Denoising of magnitude and phase data was performed using a PCA decomposition of the data and keeping only 3-5 components out of 460, as required by random matrix theory analysis of the distribution of eigenvalues, as well as comparison of eigenvalue spectra from signal and noise regions. Masks for different tissue types were defined as follows: wm, gm and CSF via SPM using simultaneously acquired MPRAGE; tumour: using PET data, thresholded for intensities 1.6 times higher than average healthy white matter; oedema: from MPRAGE with and without contrast agent, T2-SPACE and FLAIR, using a random forest algorithm. For comparison of maps and distributions, an age and gender-matched volunteer-glioblastoma patient pair was selected (female, 29y.o.).Results
The low SNR of the original data was dramatically increased by performing PCA-based denoising of the complex data. Subsequent NNLS analysis revealed the existence of several tissue specific components. Due to the extremly high temporal sampling, NNLS analysis of the original, low-SNR data, was also possible and consistent with the results on the denoised curves. The analysis of voxels within tissue apparently unaffected by the presence of tumour revealed the existence of a short T1 component in healthy white and grey matter, of components reflecting the characteristic WM and GM T1 values, and of components with very long relaxation times (similar to CSF). Figures 1 - 3 illustrate similarities and differences between healthy tissue and brain tumour. The maps corresponding to different T1 components, shown in Fig.1 (volunteer) and 2 (patient), were obtained by splitting the T1 interval 50ms-6s into several bins, labeled ‘Myelin’, ‘WM’, ‘GM’, ‘high GM’ and ‘CSF’. The number of peaks in the T1 relaxogram identified in the ‘high gray matter’ region, which seems to be stronger represented in the tumour area, is compared in Fig. 3. The percentage of this ‘high grey matter’ structure in different tissue classes is listed in Table 2 for all patients and volunteers.Discussion and conclusions
The properties of normal appearing
tissue were very similar in patients and healthy volunteers. A low-amplitude
multi-peak T1 distribution in healthy tissue seems to converge into a single
component in the tumour and oedema area. Whether the underlying tissue
structure is similar and the differences in T1 distribution reflect, for
example, altered exchange between compartments (due to e.g. changes in pH,
temperature, membrane permittivity etc in tumours), or whether the new T1
component is indeed a specific marker for cancer cells remains to be
elucidated.
As amply demonstrated by diffusion
parameters previously, quantitative MRI offers a unique possibility to
investigate tumour structure in vivo. In this contribution, multicomponent
T1 values are investigated for the first time. A unique component associated
with tumour and oedema is identified. A better classification of the tumour environment
offers hope for a better characterisation of the region of tumour infiltration.Acknowledgements
No acknowledgement found.References
[1] C. Labadie, J.H. Lee, W.D. Rooney, S. Jarchow, M. Aubert-Frecon, C.S. Springer Jr., H.E. Moller. Myelin water mapping by spatially regularized longitudinal relaxographic imaging at high magnetic fields. Magn. Reson. Med., 71 (1) (2014), pp. 375-387
[2] R. Barta, S. Kalantari, C. Laule, I.M. Vavasour, A.L. MacKay, C.A. Michal. Modeling T1 and T2 relaxation in bovine white matter. Journal of Magnetic Resonance 259, 2015, p56-67, https://doi.org/10.1016/j.jmr.2015.08.001.
[3] Shah, N. J., Neeb, H., Zaitsev, M., Steinhoff, S., Kircheis, G., Amunts, K., . . . Zilles, K. (2003). Quantitative T1 mapping of hepatic encephalopathy using magnetic resonance imaging. Hepatology, 38(5), 1219-1226. doi:10.1053/jhep.2003.50477
[4] Zaitsev, M., Steinhoff, S., & Shah, N. J. (2003). Error reduction and parameter optimization of the TAPIR method for fast T1 mapping. Magn Reson Med, 49(6), 1121-1132. doi:10.1002/mrm.10478
[5] I.T. Jolliffe. Principal Component Analysis (Springer Series in Statistics) 2nd Edition ISBN-13: 978-0387954424
[6] Veraart, J., Novikov, D. S., Christiaens, D., Ades-Aron, B., Sijbers, J., & Fieremans, E. (2016). Denoising of diffusion MRI using random matrix theory. NeuroImage, 142, 394-406. doi:10.1016/j.neuroimage.2016.08.016
Figures
