A region growing method for tumor volume segmentation on PET images for rectal and anal cancer patients

Abstract

The application of automated segmentation methods for tumor delineation on 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) images presents an opportunity to reduce the interobserver variability in radiotherapy (RT) treatment planning. In this work, three segmentation methods were evaluated and compared for rectal and anal cancer patients: (i) Percentage of the maximum standardized uptake value (SUV% max), (ii) fixed SUV cutoff of 2.5 (SUV2.5), and (iii) mathematical technique based on a confidence connected region growing (CCRG) method. A phantom study was performed to determine the SUV% max threshold value and found to be 43%, SUV43% max. The CCRG method is an iterative scheme that relies on the use of statistics from a specified region in the tumor. The scheme is initialized by a subregion of pixels surrounding the maximum intensity pixel. The mean and standard deviation of this region are measured and the pixels connected to the region are included or not based on the criterion that they are greater than a value derived from the mean and standard deviation. The mean and standard deviation of this new region are then measured and the process repeats. FDG-PET-CT imaging studies for 18 patients who received RT were used to evaluate the segmentation methods. A PET avid (PETavid) region was manually segmented for each patient and the volume was then used to compare the calculated volumes along with the absolute mean difference and range for all methods. For the SUV43% max method, the volumes were always smaller than the PETavid volume by a mean of 56% and a range of 21%-79%. The volumes from the SUV2.5 method were either smaller or larger than the PETavid volume by a mean of 37% and a range of 2%-130%. The CCRG approach provided the best results with a mean difference of 9% and a range of 1%-27%. Results show that the CCRG technique can be used in the segmentation of tumor volumes on FDG-PET images, thus providing treatment planners with a clinically viable starting point for tumor delineation and minimizing the interobserver variability in radiotherapy planning.

Combining multiple FDG‐PET radiotherapy target segmentation methods to reduce the effect of variable performance of individual segmentation methods

“…[32][33][34] For the gradient-based method a 7-mm bilateral filter was applied and the deblurring step was accomplished with 30 iterations of Landweber's algorithm with a symmetrical 6 mm FWHM point spread function (PSF) in accordance with Geets et al 14 The choice of a 7-mm bilateral filter was based on the method outlined by Hofheinz et al 35 as it provides comparable signal-to-noise in background regions to the 5 mm Gaussian filter used for the other segmentation methods.…”

Automated segmentation of tumour changes in temporal PET-CT data

“…Recently, Ballangan et al [8] presented a 'downhill' RG algorithm optimized for lung cancer tumour segmentation, where the intensity gradient of voxels was used to classify the voxels belonging to the tumour. In another study, Day et al [10] introduced an adaptive RG based on the mean and standard deviation of voxel intensities to iteratively delineate the tumours. There are some recent studies that investigated the use of segmentation in temporal PET studies.…”

“…It is very difficult to trace the changes of multiple tumours and perform quantitative assessment of their changes over time. There are many researches done for tumour delineation [4][5][6][7][8][9][10]. Hatt et al [5] developed a fuzzy local adaptive Bayesian (FLAB) clustering algorithm and was shown to be effective for small and heterogeneous tumours.…”

Rationale, Instrumental Accuracy, and Challenges of PET Quantification for Tumor Segmentation in Radiation Treatment Planning

“…(Erdi et al, 1997;Mah et al, 2002;Paulino & Johnstone, 2004); b) adaptive threshold methodse.g. (Erdi et al, 1997;Black et al, 2004;Nehmeh et al, 2009;Nestle et al, 2005;Seuntjens et al, 2011) and c) advanced segmentation methods, which include a large variety of more complex numerical approaches using gradient , statistical (Aristophanous et al, 2007;Hatt et al, 2009;Dewalle-Vignion et al, 2011), region growing (Day et al, 2009); deformable models (Li et al, 2008), texture analysis as well as other supervised or unsupervised learning methods (Zaidi, 2006;Zaidi & El Naqa, 2010;Belhassen & Zaidi, 2010). A similar classification is adopted by the educational task group on PET auto-segmentation within the American Association of Physicists in Medicine (AAPM TG211).…”