Room: Exhibit Hall | Forum 2
Purpose: The heterogeneity of [F-18]-FDG uptake within individual lesions on PET/CT scans can be measured using quantitative image analysis. It has been hypothesized that more heterogeneous, irregularly-shaped lesions respond worse to treatment. The goal of this study was to compare measures of shape and intra-lesion heterogeneity to changes in SUV metrics of lesions over the course of systemic therapy.
Methods: Fourteen patients with advanced lung cancer treated with various systemic therapies received baseline and follow-up FDG PET/CT scans. Each lesion was identified by an experienced nuclear medicine physician and segmented using an in-house automated method. Contours were verified by the physician. Forty-two lesions were identified on both scans. For each time-point, various metrics were extracted for each lesion: SUVmax, SUVmean, SUVtotal, lesion volumes, surface areas normalized to volume, surface areas of sub-volumes of high uptake normalized to volume, variance of SUV values and texture features. Metrics measured on baseline scans were correlated with changes in SUV lesion values (âˆ†SUV) between the baseline and follow-up scans using Pearsonâ€™s correlation. P-values were corrected using the Bonferroni method.
Results: Normalized lesion surface area at baseline (r=-0.50, p<0.001) strongly correlated to Î”SUVtotal. The two measures of intra-lesional heterogeneity also correlated to Î”SUVtotal: variance of baseline SUV values within lesions (r =-0.47, p<0.002) and the ratio of surface area to volume in sub-lesion regions of high uptake (r=-0.43, p<0.001). No higher order baseline texture features significantly correlated with âˆ†SUVmean, âˆ†SUVtotal or âˆ†SUVmax.
Conclusion: Lesion response correlated most strongly with normalized lesion surface area. Two metrics of lesion heterogeneity (surface area of high uptake regions and variance of SUV) also correlated significantly with lesion response. Contrary to initial expectations, heterogeneous lesions responded better to treatment in this population of patients, although these results should be confirmed in a larger patient population.
Funding Support, Disclosures, and Conflict of Interest: Work funded in part by the University of Wisconsin Carbone Cancer Center Support Grant P30 CA014520 and the Wisconsin Oncology Network of Imaging eXcellence.