Measurement of the Effect of Tissue Fixation on Tumour Microstructure using VERDICT Diffusion-MRI
Ben Jordan1, Tom Roberts1, Angela D'Esposito1, John Connell1, Andrada Ianus2, Eleftheria Panagiotaki2, Daniel Alexander2, Mark Lythgoe1, and Simon Walker-Samuel1

1Centre for Advanced Biomedical Imaging, University College London, London, United Kingdom, 2Centre for Medical Image Computing, University College London, London, United Kingdom

Synopsis

It has previously been shown that compartmental models of tissue diffusion such as VERDICT can enable access to useful measures of in-vivo tumour microstructure such as cell radius. However, comparing the in-vivo values with those measured from histology showed that a discrepancy exists between the two; histological values were consistently smaller. In this study, we assess the ability of VERDICT MRI to detect this change in cell radius by acquiring data (9.4T MRI) both in-vivo and post-fixation. A significant decrease in cell radius was detected post-fixation, which was supported by a decrease in the intra-cellular volume fraction.

Purpose

The purpose of this study was to evaluate the ability of the VERDICT MRI model of compartmental diffusion to quantify tumour microstructure and use it to assess the effects of tissue fixation. Previously it has been demonstrated that the VERDICT model can be successfully used in-vivo to access histologic features such as cell size, vascular volume fraction, intra- and extracellular volume fractions and microvascular pseudo-diffusivity1. However, the gold standard cell diameter values attained through histological analysis were on average 12.3% smaller than the VERDICT measurements. This discrepancy was attributed to the shrinkage of cells during the fixation process. In this study both in-vivo and ex-vivo data acquired from the same subjects are compared to ascertain whether the expected cell shrinkage can be detected in the VERDICT estimate of cell size.

Methods

Animal Model: The LS174T human colon adenocarcinoma cell line was injected subcutaneously at a concentration of 5x106 cells per 100ml of serum-free media (~106 cells per animal) into 5 female MF1 nu/nu mice (age ~6wks)2. Daily checks for tumour growth were carried out using callipers. After a period of 18-19 days tumours were ~1cm3 and thereby suitable for imaging.

Data Acquisition: Animals were scanned on a 9.4T Varian 20cm horizontal bore scanner (Varian Inc. Palo Alto, CA, USA) with a maximum gradient strength of 400mT/m, and a 39mm birdcage RF coil (Rapid Biomedical, Rimpar, Germany). Dental paste was used to help secure the tumour, and prevent excess respiratory motion. In-Vivo VERDICT data was acquired using a steady-state respiration-gated PGSE sequence with 46 b-values in total. Gradient separation times Δ = 10, 20, 30, 40ms with duration δ = 3ms for all Δs and δ = 10ms for Δ = 30, 40ms. Gradient strength G was stepped from 40 to 400mT/m in steps of 40mT/m for δ = 3ms and 40, 80, 120mT/m for δ = 10ms. The in-plane field-of-view was 25mm x 25mm, slice thickness 0.5mm, minimum TE, 2 averages, data matrix 64 x 64 and 3 slices per acquisition. The total scan time for in-vivo data was ~5hours. Ex-vivo data was acquired using the same protocol, except with the number of averages increased to 6 to improve SNR and TR was minimised. Total scan time was around 8.5 hours.

Tissue Fixation: Perfusion fixation was carried out through the left ventricle using heparinised saline followed by 4% paraformaldehyde (PFA). This was followed by immersion fixation for 2 weeks in 4% PFA at 4°C. 1 week prior to imaging the fixed animals were rinsed and washed and transferred to 0.9% saline to rehydrate the tissue.

Image Analysis: Tumour ROIs were manually segmented from each data set. The signal was averaged over the whole ROI, and modelling performed on the averaged signal. Model fitting was performed using an iterative optimisation scheme in the Camino toolkit3. The “BallSphereStick” model (using the taxonomy of Panagiotaki et al 2012)4 was fitted to in-vivo data, where the “Ball” and “Sphere” compartments correspond to the free and restricted (intracellular) diffusion, respectively, and the “Stick” compartment corresponds to the pseudo-diffusion within the microvasculature. The intra-cellular diffusivity was fixed (diso=3e-9m2s-1) For ex-vivo data the “BallSphere” model was fitted as there is no pseudo-diffusion present.

Results

The VERDICT model provided an accurate fit to the diffusion data for both in-vivo and ex-vivo samples. Figure 1 shows typical fits produced for the data in this study, where the VERDICT model is capable of delivering an accurate fit over the entire range of b-values. The cell-radius parameters produced by the model-fitting are shown in figure 2. A significant reduction (Wilcoxon rank-sum, p=0.0079) in cell radius was detected between the in-vivo and ex-vivo datasets, in agreement with previous findings1. The measured decrease in cell size was supported by a significant (p=0.0079) decrease in the intra-cellular volume fraction and an increase in the extracellular volume fraction.

Discussion and Conclusions

This study aimed to evaluate whether a decrease in tumour cell size caused by chemical fixation could be measured using VERDICT MRI. We successfully measured a significant decrease in cell radius and intracellular volume fraction in a subcutaneous colorectal tumour model. This result demonstrates that there is a potential for VERDICT MRI to access measurements of ex-vivo histological parameters from within tumours non-invasively. A direct comparison between in-vivo, ex-vivo and histology with a variety of fixation methods is required to validate this. One potential limitation of this study is that different models were used to represent the in-vivo and ex-vivo data, meaning direct comparison of the fitted parameters is more difficult. Future in-silico work will aim to validate this.

Acknowledgements

This work was supported by a Wellcome Trust Senior Research Fellowship (grant WT100247MA), King’s College London and UCL Comprehensive Cancer Imaging Centre CR-UK & EPSRC, in association with the DoH (England). This work was also supported by the EPSRC-funded UCL Centre for Doctoral Training in Medical Imaging (EP/L016478/1) and the Department of Health’s NIHR-funded Biomedical Research Centre at University College London Hospitals.

References

1. Panagiotaki E, Walker-Samuel S, et al. Non-invasive quantification of solid tumour microstructure using VERDICT MRI. Cancer Research. 2014;74(7):1902-12.

2. Folarin A, Konerding M, Timonen J, et al. Three-dimensional analysis of tumour vascular corrosion casts using stereoimaging and micro-computed tomography. Microvascular Research. 2010;80(1):89-98.

3. Cook PA, Bai Y, Alexander DC, et al. Camino: Open-Source Diffusion-MRI Reconstruction and Processing. Proc ISMRM. 2006:2759.

4. Panagiotaki E, Schneider T, Siow B, et al. Compartment models of the diffusion MR signal in brain white matter: a taxonomy and comparison. Neuroimage. 2012;59(3):2241-54.

Figures

Figure 1 – Typical manually-segmented ROI overlayed on b0 diffusion image. ROIs were drawn conservatively - keeping clear of the edge of the tumour in order to minimise partial-volume effects.

Figure 2 – VERDICT fits to typical (left) in-vivo and (right) ex-vivo (fixed) data, acquired in a colorectal tumour xenograft model. Data points (symbols) represent the normalized signal intensity and solid lines show corresponding optimized fits by the VERDICT model.

Figure 3 – Cell radius and intra-cellular volume fraction parameter estimates produced by the VERDICT model for the in-vivo and ex-vivo datasets.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0012