Diffusion-weighted MR imaging of metastatic abdominal and pelvic tumours is sensitive to early changes induced by a VEGF inhibitor using alternative diffusion attenuation models

Abstract

ObjectivesTo assess the utility of diffusion weighted imaging for monitoring early treatment effects associated with a VEGF inhibitor.Materials and methodsTwenty-nine patients with metastatic abdominal and pelvic tumours were recruited and imaged with DW-MRI: twice at baseline, and after 7 and 28 days of treatment with cediranib. Tumour measures were derived using mono-exponential, bi-exponential and stretched-exponential models, and parameter repeatability and treatment effects seen after 7 and 28 days were assessed. Correlations with volume changes and DCE-MRI metrics were also assessed.ResultsDiffusion coefficient repeatabilities from all models were < 6 %; f and D* (bi-exponential) were 22 % and 44 %; α (stretched-exponential) was 4.2 %. Significant increases in the diffusion coefficients from all models were observed at day 28 but not day 7. Significant decreases in D* and f.D* were observed at day 7 and in f at day 28; significant increases in α were observed at both time-points. Weak correlations between DW-MRI changes and volume changes and DCE-MRI changes were observed.ConclusionDW-MRI is sensitive to early and late treatment changes caused by a VEGF inhibitor using non-mono-exponential models. Evidence of over-fitting using the bi-exponential model suggests that the stretched-exponential model is best suited to monitor such changes.Key points• Non-mono-exponential diffusion models widen sensitivity to a broader class of tissue properties.• A stretched-exponential model robustly detects changes after 7 days of VEGF-inhibitor treatment.• There are very weak correlations between DWI-IVIM perfusion and similar DCE-MRI measures.• Diffusion-weighted MRI is a highly informative technique for assessing novel tumour therapies.

Assessment of early treatment response to neoadjuvant chemotherapy in breast cancer using non-mono-exponential diffusion models: a feasibility study comparing the baseline and mid-treatment MRI examinations

“…These preliminary results are similar to those found by Orton et al in abdominal and pelvic tumours where changes in α were substantially different between response groups when assessed before the start of therapy, 7 days and 28 following treatment [5]. …”

Separation of type and grade in cervical tumours using non-mono-exponential models of diffusion-weighted MRI

“…It is, however, important to note that the mono-exponential model can be considered a special case of the other models applied here (using e.g. α = 1, K = 0, f  = 0), as has been pointed out in previous studies [13], suggesting that tumours that prefer the mono-exponential model are not disadvantaged by the use of other models, although the additional parameters would be redundant if the mono-exponential model were preferred in all tumours. The stretched exponential, kurtosis, and statistical models describe the departure of the DW-MRI signal decay from mono-exponential behaviour, which may be due to intra-voxel heterogeneity or non-Gaussian diffusion processes, either of which may be related to the observed separation of types and grades of tumours.…”

“…It was not possible to carry out repeated examinations on the patients in this study owing to the invasive nature of the MRI examination using an endovaginal receiver coil and it was, therefore, not possible to assess the test-retest repeatability of the parameter estimates. However, the repeatability of fitted parameters of the bi-exponential and stretched exponential models, which were preferred by largest numbers of tumours in this study, have been investigated in several previous studies that use similar DW-MRI protocols to the one used here [9, 11, 13, 31]. The parameters of the stretched exponential model have been shown to have very good repeatability (coefficient of variation (CV) ∼ 4 to 7 %), comparable to the repeatability of ADC estimates [9, 11, 13], while the repeatability of parameters from the bi-exponential model, particularly f and D* are significantly worse (CV > 20 %) [9, 11, 13, 31].…”

“…The use of non-mono-exponential models provides a better description of the DW-MRI signal, and parameters derived from these models allow more detailed investigation of differences between tumour sub-types or inter-tumour heterogeneity [6–11] and may also provide an earlier indication of response to treatment [12, 13]. However, use of a model with a large number of additional parameters risks over-fitting the data and may be sensitive to noise characteristics of the system rather than structural properties of the tumour or normal tissue.…”

DCE-MRI, DW-MRI, and MRS in Cancer

“…Although estimates of ADC and some other fitted parameters have shown good repeatability, estimates of f and D ∗ from the bi-exponential model generally exhibit poorer repeatability. 66,67 …”