INTRODUCTION

Malignant tumor tissue develops interstitial hypertension, primarily due to restricted blood supply and impaired lymphatic flow. This results in elevated hydrostatic interstitial fluid pressure (IFP) or change in velocity (IFV) within the tumor compared to normal tissue. The negative pressure gradient between the tumor surface and normal tissue can contribute to treatment resistance and metastatic lymphatic spread¹. However, measuring these parameters is impractical in clinical practice. In this study, we have integrated pharmacokinetic and fluid flow models to assess IFP and IFV in breast cancer patients non-invasively using high-resolution dynamic contrast enhanced (DCE) MRI. We computed these parameters in a subgroup of NAC patients and reported their association with treatment response.

MATERIALS AND METHODS

This retrospective study included breast MRI exams of patients (N =15) with recently diagnosed breast cancer who underwent MRI exam prior to NAC. All imaging exams were performed on a 3.0 Tesla system (Discovery 750, GE Medical Systems, Waukesha, WI) and DCE data was acquired using differential sub-sampling with cartesian ordering, before and after the intravenous administration of gadoterate meglumine (0.2 ml/kg). Nine post contrast timepoints were acquired with a total acquisition time of ~5-7 min. A two-compartment pharmacokinetic model, i.e., extended Tofts model (ETM)², was solved with the aid of a commercial software Olea Sphere®, which computes the microvascular plasma volume transport between intravascular space (IVS) and extravascular extracellular space (EES). When ETM model is applied to DCE MRI (Fig. 1), the change in concentration of contrast agent between IVS and EES of tissue enabled the quantification of the plasma volume transport parameter K^trans (min^-1). Next, tumor and normal tissue volumes were manually segmented on the K^trans map (using 3D slicer) and their binary masks were converted to a 3D computational domain (i.e., STL file) in MATLAB. At the same time, individual voxel values of K^trans and associated coordinate information were extracted from the K^trans maps. Finally, a steady-state fluid flow equation (1) was solved within the 3D computational domain (including tumor and healthy regions) with K^trans as an input ^3-5.
$$-\kappa_{H}\triangledown^{2}p_{i}=\left[\frac{K^{trans}}{K_{m}^{trans}}\right]\frac{L_{po}S}{V}\left[p_{v}-p_{i}-\sigma_{T}\left(\pi_{v}-\pi_{i}\right)\right]-\frac{L_{pl}S_{L}}{V}\left (p_{i}-p_{L}\right)\,\,\mathbf{and}\,\,\,u_{i}=-\kappa_{H}\triangledown p_{i}\,\,\,\,\,\,\,\,\,\,\,\,(1)$$
where, $$$K^{_{m}^{trans}}$$$ is the mean tumor value, p_i is interstitial fluid pressure, u_i = interstitial fluid velocity, $$$\kappa_{H}$$$ is hydraulic conductivity, L_po is vessel permeability, S/V is microvascular surface area per unit volume, σ_T is average osmotic reflection coefficient for plasma, π_i is osmotic pressure in interstitial space, π_v is osmotic pressure in microvasculature, p_L is lymphatic pressure, p_v is microvascular pressure, and L_plS_L/V is lymphatic filtration coefficient. The nominal parameters of Eq (1) for tumor and healthy region are assumed from literature ^3-5. Solving Eq (1) in the 3D domain with (a) no flux or pressure gradient condition external to the normal domain and (b) continuous pressure and flow rate condition at the interface between the tumor and the normal tissue led to p_i (IFP) and u_i (IFV) maps. Finally, volume averages of the K^trans, IFP and IFV were extracted over each lesion for statistical comparison (Fig. 1). K^trans, IFP and IFV values between the pathologic complete response (pCR) and no-pCR subgroups were compared using the non-parametric Kruskal–Wallis test.

RESULTS

We successfully generated K^trans, IFP and IFV maps for all 15 patients. Of these, 6 patients had no-pCR and 9 patients had pCR. We observed a higher inter-subject mean value of tumor IFP for no-pCR group compared to the pCR group. Conversely, we observed higher values for tumor K^trans and IFV in the pCR group compared to the no-pCR group (Fig. 2). However, this difference was not significant (p > 0.05) due to small population size.

DISCUSSION

We presented a study to non-invasively map IFP and IFV on breast MR images in patients with recently diagnosed cancer. Although our small dataset precluded statistically significant results, we observed a trend towards higher IFP and lower IFV in the no-PCR subgroup. This makes physiological sense since high IFP and low IFV may hinder the transport of chemotherapy agents to the tumor, thereby mitigating their therapeutic effect. More work is needed with larger sample sizes to determine if IFP and IFV may serve as valuable biomarkers of treatment response and tumor aggression.

CONCLUSION

Our non-invasive pharmacokinetic and flow modeling of breast MRI quantifies tumor IFP and IFV and could serve as a tool for prediction of early treatment response and tumor aggressiveness.

Acknowledgements

This study is supported in part through the NIH/NCI Cancer Center Support Grant P30 CA008748.