Purpose: This work looks to examine the potential for an automated contouring algorithm to delineate gadolinium-based (Gd) nanoparticle uptake in patient metastases, allowing for voxel-to-voxel determination of its distribution.
Methods: Data have been acquired from four patients with brain metastases enrolled in a phase 2 clinical trial (NCT04899908) examining uptake of Gd-based nanoparticles and its theranostic effects. Each patient was scanned with an MRI (3T) MP2RAGE sequence pre- and post- intravenous infusion of nanoparticles. Sequence outputs (UNI volumes) were registered to planning volumes. Nanoparticle concentrations within physician defined volumes were calculated from T1 maps using the measured change in T1 values and previously determined relaxivity constant (6 mM⁻¹s⁻¹). Physician drawn contours were input to a Chan-Vese segmentation algorithm which converges (100 iterations) to a level-set based solution. The resulting regions were subject to refinement via morphological opening and binary overlap operations. Mean physician-defined nanoparticle uptake was compared to mean automated contour defined uptake.
Results: The segmentation algorithm resulted in automated contours that were visually confirmed to be in good agreement (large overlap) with physician drawn contours. Physician-defined uptake was compared to automated contour defined uptake. There was a notable increase in mean nanoparticle concentration within patient tumors when using the automated contours. Additionally, an expected pattern of tumor uptake was observed more clearly with automated contours; increasing uptake moving from tumor center to surface. A relationship between tumor uptake and tumor volume was non-existent using the physician contours but present using the automated contours.
Conclusion: Results showed a notable difference in the measurement of nanoparticle uptake when an automated segmentation tool is introduced. Further, the introduction of automated contours allowed for the observation of uptake patterns not captured by physician drawn contours. This work suggests automated contours may result in a more accurate 3D representation of nanoparticle uptake and distribution.
Not Applicable / None Entered.