Introduction

Functional MRI, especially dynamic contrast enhanced (DCE) MRI, has shown some value for staging and treatment response evaluation in certain types of cancer. However, the results in rectal cancer have been highly heterogeneous (1). We examined perfusion related parameters from three functional MRI methods in the primary tumour of rectal cancer patients; DCE MRI, dynamic susceptibility contrast (DSC) MRI and intravoxel incoherent motion (IVIM) MRI, the relationship between them and their value as biomarkers.

Methods

94 patients successfully underwent a multi-echo dynamic MRI with contrast agent injection and an extended diffusion weighted (DWI) sequence with seven b-values in addition to clinical procedure MRI. The scanning was done on a Philips Achieva 1.5 Tesla System (Philips Healthcare, Best, The Netherlands), with a five-channel cardiac coil. The dynamic data was obtained with a 3D multi-shot GRE EPI with three echoes (4.6, 13.9, 23.2 ms) with injection of a dose of 0.2 ml/kg body weight of a gadolinium-based contrast agent (Dotarem® 279.3 mg/ml, Guerbet Roissy, France). The flip angle was 28° and the repetition time was 39 ms. Time resolution varied between 1.9 and 2.5 s, and spatial resolution was 1.96 x 2.00 x 10 mm³. From the multi-echo data, the T₁- and T₂^*-curves were extracted and used for an extended Tofts analysis (2) and a model-free deconvolution(3), respectively. This yielded the parameters K^trans, k_ep, v_p and v_e from the DCE, and relative blood flow (BF) and area under curve (AUC) for 30 and 60 s from the DSC. Individual arterial input functions (AIFs) were extracted from the T₂^* data and applied for both analyses. The relative BF is related to the true perfusion, f, according to the equation:
$$BF = \beta \, x\, f \frac{k_h}\rho$$
where k_h is the haematocrit value, ρ the tissue density and β a scaling constant. To compare these values, it was therefore assumed that k_h was the same for all patients, ρ the same for all tumours and β the same for all examinations. It was also assumed that the leakage was negligible for the estimation of BF from the initial part of residual function after deconvolution. The DWI was obtained with an SE EPI with b-values 0, 25, 50, 100, 500, 1000 and 1300 s/mm². The repetition time was 3000 ms, the echo time was 75 ms, and the spatial resolution was 2.0 x 2.67 x 4 mm³. DWI-data were fitted to a bi-exponential IVIM equation, yielding the parameters perfusion fraction f, pseudo-diffusion D^* and diffusion D. Tumour delineations were done on T₂ weighted images with DWI as extra guidance by two radiologists with 7 and 14 years of experience with abdominal MRI. Tumour regions were then semi-automatically co-registered to the parametrical images, and the median values were extracted. Correlation analysis was done with Spearman's rank correlation test. Survival analysis was done with a univariable Cox regression, where progression free survival was estimated from date of inclusion to local recurrence, metastatic disease or death for patients without metastasis at time of diagnosis. Differences in treatment response were examined with Student’s t-test for variables that were normally distributed and a Mann-Whitney U-test for variables that were not, this was determined by the Shapiro-Wilk test. The latter was done on the group of patients receiving neoadjuvant treatment (n = 43) where the resected specimen was examined by a pathologist and given a ypT score (pathologically assessed T stage after treatment).

Results

We found a correlation between BF and D^* (r_s = 0.47, p < 0.001), BF and K^trans (r_s = 0.29, p = 0.004), as well as BF and v_p (r_s = 0.44, p < 0.001). Low BF was associated with short progression free survival and overall survival for all patients, as well as poor treatment response for the patients receiving preoperative treatment before surgery, Table 1. Example images are given in Figure 1, and a Kaplan Meier plot for progression free survival is given in Figure 2. D^* was also related to progression free survival and v_p to treatment response. We also observed a sex difference in BF, at higher tumour stages, where women had higher BF than men (125 ± 27 ml/min/100 g for women versus 74 ± 26 ml/min/100 g for men at stage 4, p < 0.001), Figure 3.

Discussion

This comparison of DSC, DCE and IVIM MRI in rectal cancer patients suggests that BF from DSC was the best biomarker for progression free survival, overall survival and treatment response. A low BF was related to poor outcome and may reflect poor oxygen delivery to the tissue and possibly hypoxia, a known marker of aggressive disease, metastatic development and poor treatment response. The assumptions for comparing the relative BF between patients, including the reproducibility of the AIF, are debatable, but our results suggest that they are within reason. DSC is not commonly used in extracranial cancers and these results supports further investigations of this method outside the brain.

Conclusion

DSC should be further investigated as a method for obtaining biomarkers in extracranial cancers.

Acknowledgements

No acknowledgement found.