Edge detection for biological tumor volume definition based on FDG-PET/CT fused imaging: Evaluation of diffusion processes by using an agar phantom

Abstract

Purpose: Fused positron emission and computed tomography (PET/CT) images have an important role in external beam radiotherapy (RT), especially for target volume delineation. For contouring on PET/CT images, the sourcebackground algorithm is the method currently used, but its sensitivity to partial volume effects may produce inaccuracies when applied on small lesions. The goal of this research was to develop a computational algorithm based on Canny's edge detection tool for processing PET/CT images. Materials and Methods: The software was implemented using the data analysis framework ROOT and the Grassroots DiCoM GDCM libraries (CERN, Geneva, Switzerland). First, pre-processing by Gaussian smoothing was applied to PET/CT image data. Then directional derivatives and gradients were computed on the basis of multivariate calculation tools. Upon completion of this process, within an agar phantom the boundaries of uptake regions having cylindrical shape were automatically sought by searching for the gradient magnitude and Laplacian above a pre-defined gradient threshold. Results: The mean diameters calculated by using the developed software show relative differences below 3% with respect to the measured value (69 mm) for all gradient thresholds between 15% and 55%. The diameters identified with the source-background algorithm differ from the measured value and only are similar at a defined source-background threshold of 35%. Conclusion: Gradient magnitude and Laplacian images obtained after processing the 18F-Fluorodeoxyglucose (FDG) PET-CT agar phantom images are associated with the diffusion processes explained by Fick’s laws. According to the obtained results, the applicability of the source-background algorithm is limited in the presence of diffusive phenomena, but it is still useful. The developed software is an alternative for improving the conventional methods for the delineation of biological tumor volume (BTV) boundaries in RT planning based on PET/CT.