Author: Stephanie Bennett, Ross I. Berbeco, Ning Jin, Sonal Josan, Justin Michael Sheetz, Atchar Sudhyadhom 👨🔬
Affiliation: Department of Radiation & Cellular Oncology, University of Chicago, University of Massachusetts - Lowell, Siemens Healthineers, Brigham and Women’s Hospital and Dana Farber Cancer Institute, Harvard Medical School,, Brigham and Women's Hospital 🌍
Purpose: AGuIX, a Gadolinium-based theranostic radiosensitizing nanoparticle, is currently under clinical evaluation in Europe and the US. Using patients from the double-blinded NanoBrainMets trial, uptake patterns of AGuIX in patients with multiple brain metastases were evaluated to determine whether it could replace standard gadolinium-based contrast agents (GBCA) for tumor contouring.
Methods: A cohort of 22 patients (123 tumors) were included in this study; each of which had a standard GBCA MRI scan and, separately, an MRI following AGuIX infusion (recipients were hypothesized using an entropy metric). Using an AI-based auto-contouring tool, respective contours were generated for each patient on high resolution T1-weighted MRIs using GBCA and AGuIX for contrast respectively. For each lesion, contours were compared using Hausdorff distance, Dice coefficient, and contour volume ratios. Physician-drawn GTVs delineated by means of GBCA were considered for a clinical comparison.
Results: When considering the differences between AGuIX and GCBA auto-generated contours, 82.11% of tumors had Dice coefficients greater than 0.8 and the average Hausdorff distance was 1.8445 mm (two-voxels). The average ratio of GBCA to AGuIX GTV volumes was 1.0951 and the average Dice coefficient was 0.8582. 8.04% of tumors had greater than 3 mm Hausdorff distance between the target GBCA and AGuIX auto-contours. This may be indicative of different uptake mechanisms for respective agents. As AGuIX is an intracellular contrast agent with significant retention, differences in contrast could reveal necrosis or hypoxic regions of the target. Compared to physician contours, Hausdorff distances of 3.4816 mm and 3.8652 mm and Dice coefficients of 0.7686 and 0.7521 were found for GBCA and AGuIX, respectively.
Conclusion: Further investigation is needed to determine whether functional differences are correlated with lesion uptake patterns. Results indicate that physicians could contour on AGuIX images without compromising the integrity of the GTV delineation for most patients and most lesions.