Intra-parenchymal injection and delivery of therapeutic agents have been used in clinical trials for brain cancer and other neurodegenerative diseases. The complexity of transport pathways in tissue makes it difficult to envision therapeutic agent distribution from clinical MR images. Computer-assisted planning has been proposed to mitigate risk for inadequate delivery through quantitative understanding of infusion characteristics. We present results from human studies and simulations of intratumoral infusions of immunotoxins in glioblastoma patients. Gd-DTPA and 124I-labeled human serum albumin (124I-HSA) were co-infused with the therapeutic, and their distributions measured in MRI and PET. Simulations were created by modeling tissue fluid mechanics and physiology and suggested that reduced distribution of tracer molecules within tumor is primarily related to elevated loss rates computed from DCE. PET-tracer on the other hand shows that the larger albumin molecule had longer but heterogeneous residence times within the tumor. We found over two orders of magnitude variation in distribution volumes for the same infusion volumes, with relative error ~20%, allowing understanding of even anomalous infusions. Modeling and measurement revealed that key determinants of flow include infusion-induced expansion and loss through compromised BBB. Opportunities are described to improve computer-assisted CED through iterative feedback between simulations and imaging.
Keywords: CED (convection-enhanced delivery); drug delivery planning; predictive flow models; targeted drug delivery.