Segmentation of dynamic contrast enhanced magnetic resonance imaging data

Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) allows in vivo characterization of tumour vasculature. As such, it is applicable for monitoring the effects of treatments targeting vasculature. The aims of this study were to evaluate the properties of tumour areas segmented-out by DCE-MRI parameters and to evaluate the changes induced by the vascular disrupting agent (VDA) combretastatin A-4 disodium phosphate (CA4DP), a leading VDA in clinical trials, in these areas.

Two tumour models previously shown to respond differently to CA4DP were chosen. The C3H mammary carcinoma and the KHT sarcoma were grown in the right rear foot of CDF(1) and C3H/km mice, respectively, and treated when at 200 or 800 mm(3) in size. DCE-MRI, using the contrast agent Gd-DTPA, was performed on a 7 T spectroscopy/imaging system before and 3 hours after i.p. CA4DP administration at a dose of 100 mg/kg. From the voxel concentration-time curves, the semiquantitative parameter of initial area under the curve (IAUC), the model parameters transfer constant K(trans), interstitial volume v(e), and blood plasma volume v(p), were calculated. Tumour images were segmented into three groups based on the DCE-MRI model parameters using the K-means algorithm, and the groups were ranked by IAUC.

The resulting voxels of the tumour segments were mainly spatially connected structures. Initial DCE-MRI parameter values showed different dependencies on tumour model and size in the regions. For all regions in all tumour groups, the treatment reduced IAUC by 36-51%, whereas the model parameters showed more dependencies on tumour model and size.

This segmentation technique identifies tumour regions with different microenvironmental characteristics responding differently to CA4DP and may be valuable in the optimization of combined VDA with radiotherapy or chemotherapy. The method may also prove useful for optimization and monitoring of local treatment such as radiotherapy.