High resolution, dynamic GdDTPA-enhanced images of proaches have been developed for dynamic contrast-en-MCF7 human breast tumors in immunodeficient mice were anahanced MRI (10-12), assuming a reversible exchange belyzed at pixel resolution. The analysis, based on a physiological tween the intravascular and the extracellular volume in the model, was performed by applying a nonlinear least-square allesion and the rest of the body and clearance via the kidneys. gorithm using a color coded scale. The final output mapped at These models attempted to provide physiological-pharmapixel resolution capillary permeability times surface area and cokinetic and quantitative approaches to contrast-enhanced fraction of extracellular volume, for each tumor slice. In addidata (10-17), including breast carcinoma . Empirition, the output included assessment of the fit to the model by cal approaches of image analysis based on kinetic analysis determining the proportion of variability ( R 2 ) for each pixel.
were also applied and were correlated with histological find-
The spatial variation in the R 2 values served to identify regions where the predominant mechanism of enhancement was leakage ings (20, 21). In most of the contrast-enhanced studies the from the intravascular volume to the extracellular volume ( R 2 recording parameters have been optimized to increase temclose to 1 ) . In regions with low R 2 other mechanisms of enporal resolution, but at the expense of spatial resolution (20hancement appear to be dominating presumably diffusion 23). In the image analysis of such data the spatial resolution within the extracellular space. As expected, in necrotic regions was further reduced by averaging signal intensity over selacking microcapillaries and identified by analyzing T 2lected regions of interest or over the whole tumor (21)(22)(23). weighted images of the same tumors, the model failed to fit the It is clear, however, that most tumors are very heterogeneous dynamic contrast enhanced data. The heterogeneous distribuand vary over space in terms of microscopic histopathologition of the determined pathophysiological features demoncal features (e.g., origin: tumor cells, fibrous cells, endothestrates the importance of recording and analyzing breast tumor images at high spatial resolution. ᭧ 1997 Academic Press lial cells) and macroscopic histological features (such as cellularity-fraction of volume occupied by cells and presence and volume fraction of microvessels) as well as physio-1 To whom correspondence should be addressed.