Optimizing therapeutic agent delivery to malignant gliomas via convection-enhanced delivery

Convection-enhanced delivery (CED) is a promising investigational strategy for locoregionally administering therapeutic agents directly into the brain parenchyma to treat intracranial malignant gliomas. CED permits high molecular weight anti-tumoral agents transport across the selectively permeable blood-brain barrier into the central nervous system. The agents may be delivered directly to diseased tissue regions in concentrations clinically effective against tumors while minimizing toxicities associated with systemic drug administration methods. Despite CED's advantages, several challenges result in lower-than-desirable coverage of the target tumor. This work aims to address the challenge of matching drug distribution to the target tumor contours. This dissertation attempts to address the challenge of tumor contour matching by formulating the tumor coverage challenge as a coverage path planning challenge. Then presents the results of a coverage path plan that uses MRI imaging feedback to determine catheter positioning parameters for specifically targeting tumor contours yet to be perfused by a therapeutic agent.

First, we evaluated the MRI-compatible CED device's performance to verify its therapeutic agent delivery functionality. Next, we utilize existing image segmentation software and techniques to segment MRI images of a canine patient's brain. The segmentation identified the target tumor regions to be covered, the ventricles to be avoided, and the brain, which defined the planning environment's boundaries. Segmented regions were reconstructed into three-dimensional volumes in a suitable reference frame to provide coordinate information for the robotic CED device. After segmentation, a model for the morphology of the infused agent's spatial distribution from infusion points was developed. This model was used to define the spatial coverage footprint from each microneedle. Finally, we developed a catheter repositioning algorithm. The repositioning algorithm predicted the CED device's configuration that increased the target tumor's perfused volume, i.e., the tumor volume perfused by the therapeutic agent, subject to the relative importance of two control objectives: maximizing the tumor volume perfused and avoiding infusing the drug into the ventricles.

The MRI-compatible CED device delivered infused agents within acceptable specifications. Existing MRI image segmentation techniques were sufficient to identify the regions of interest from MRI images. The segmented regions were effectively reconstructed in the desired reference frame to provide target region coordinates for the planning algorithm. The algorithm successfully predicted device configurations that increase tumor perfused volume by repositioning the catheter only once. The perfused volume significantly depended on the infused agent's spatial coverage footprint model and the relative importance of avoiding the ventricles. The results from this algorithm suggest promising physical experimental results in further development.

Results from this work suggest that a refined model for the spatial coverage footprint could improve the target tumor's perfused volume. Real-time intra-operative MRI imaging feedback can address the tumor contour matching challenge. Real-time control of the catheter positions during CED can improve region targeting accuracy by compensating for differences between predicted and actual tumor perfusion and therefore improving the efficacy of CED.