A Novel Framework for Automated Segmentation and Labeling of Homogeneous Versus Heterogeneous Lung Tumors in [18F]FDG-PET Imaging
- 380 Downloads
Determination of intra-tumor high-uptake area using 2-deoxy-2-[18F]fluoro-d-glucose ([18F]FDG) positron emission tomography (PET) imaging is an important consideration for dose painting in radiation treatment applications. The aim of our study was to develop a framework towards automated segmentation and labeling of homogeneous vs. heterogeneous tumors in clinical lung [18F]FDG-PET with the capability of intra-tumor high-uptake region delineation.
We utilized and extended a fuzzy random walk PET tumor segmentation algorithm to delineate intra-tumor high-uptake areas. Tumor textural feature (TF) analysis was used to find a relationship between tumor type and TF values. Segmentation accuracy was evaluated quantitatively utilizing 70 clinical [18F]FDG-PET lung images of patients with a total of 150 solid tumors. For volumetric analysis, the Dice similarity coefficient (DSC) and Hausdorff distance (HD) measures were extracted with respect to gold-standard manual segmentation. A multi-linear regression model was also proposed for automated tumor labeling based on TFs, including cross-validation analysis.
Two-tailed t test analysis of TFs between homogeneous and heterogeneous tumors revealed significant statistical difference for size-zone variability (SZV), intensity variability (IV), zone percentage (ZP), proposed parameters II and III, entropy and tumor volume (p < 0.001), dissimilarity, high intensity emphasis (HIE), and SUVmin (p < 0.01). Lower statistical differences were observed for proposed parameter I (p = 0.02), and no significant differences were observed for SUVmax and SUVmean. Furthermore, the Spearman rank analysis between visual tumor labeling and TF analysis depicted a significant correlation for SZV, IV, entropy, parameters II and III, and tumor volume (0.68 ≤ ρ ≤ 0.84) and moderate correlation for ZP, HIE, homogeneity, dissimilarity, parameter I, and SUVmin (0.22 ≤ ρ ≤ 0.52), while no correlations were observed for SUVmax and SUVmean (ρ < 0.08). The multi-linear regression model for automated tumor labeling process resulted in R 2 and RMSE values of 0.93 and 0.14, respectively (p < 0.001), and generated tumor labeling sensitivity and specificity of 0.93 and 0.89. With respect to baseline random walk segmentation, the results showed significant (p < 0.001) mean DSC, HD, and SUVmean error improvements of 21.4 ± 11.5 %, 1.4 ± 0.8 mm, and 16.8 ± 8.1 % in homogeneous tumors and 7.4 ± 4.4 %, 1.5 ± 0.6 mm, and 7.9 ± 2.7 % in heterogeneous lesions. In addition, significant (p < 0.001) mean DSC, HD, and SUVmean error improvements were observed for tumor sub-volume delineations, namely 5 ± 2 %, 1.5 ± 0.6 mm, and 7 ± 3 % for the proposed Fuzzy RW method compared to RW segmentation.
We proposed and demonstrated an automatic framework for significantly improved segmentation and labeling of homogeneous vs. heterogeneous tumors in lung [18F]FDG-PET images.
Key wordsHeterogeneous tumor delineation Automated PET image segmentation Random walk Fuzzy logic
Compliance with Ethical Standards
Conflict of Interest
There authors declare they have no conflict of interest.
- 2.Rahmim A, Wahl R (2006) An overview of clinical PET/CT. Iranian. J Nucl Med 14:1–14Google Scholar
- 7.Basu S, Kwee T, Gatenby R et al (2011) Evolving role of molecular imaging with PET in detecting and characterizing heterogeneity of cancer tissue at the primary and metastatic sites, a plausible explanation for failed attempts to cure malignant disorders. Eur J Nucl Med Mol Imaging 38:987–991CrossRefPubMedGoogle Scholar
- 18.Onoma DP, Ruan S, Gardin I, et al. (2012) 3D random walk based segmentation for lung tumor delineation in PET imaging. Biomedical Imaging (ISBI), 2012 9th IEEE International Symposium on; p. 1260–3Google Scholar
- 19.Soufi M, Kamali Asl AR, Geramifar P (2015) Random walk algorithm seed localization parameters in lung positron emission tomography (PET) images. Med Phys 42Google Scholar
- 21.Hui C, Xiuying W, Fulham M, Feng DD (2013) Prior knowledge enhanced random walk for lung tumor segmentation from low-contrast CT images. Eng Med Biol Soc (EMBC), 2013 35th Annual International Conference of the IEEE:6071–6074Google Scholar
- 23.Bagci U, Udupa J, Yao J, Mollura D. (2012) Co-segmentation of functional and anatomical images. In: Proc. Med Image Computing and Computer-Assisted Intervention:459–67Google Scholar
- 24.Bagci U, Yao J, Caban J, et al. (2011) A graph-theoretic approach for segmentation of pet images. In: Engineering in Medicine and Biology Society, EMBC, 2011 Annual International Conference of the IEEE. IEEE:8479–82Google Scholar
- 25.Kaur EK, Mutenja EV (2010) Fuzzy logic based image edge detection algorithm in MATLAB. Intl J Computer Appl 1:55–58Google Scholar
- 26.Kumar D J, Mohan V. (2014) Edge detection in the medical MR brain image based on fuzzy logic technique. Information Communication and Embedded Systems (ICICES), 2014 International Conference on; p. 1–9Google Scholar
- 27.Rashmi KA, Kusagur DA (2012) An improved fast edge detection for medical image based on fuzzy techniques. Fuzzy Systems 4:147–150Google Scholar
- 49.Grkovski M, Apte A, Schwartz J, et al. (2015) Reproducibility of F-18-FMISO intratumor distribution and texture features in NSCLC. J Nucl Med 56Google Scholar
- 50.van Velden F, Kramer G, Frings V, et al. (2016) Repeatability of radiomic features in non-small-cell lung cancer [18F]FDG-PET/CT studies: impact of reconstruction and delineation. Molec. Imag. Biol. In PressGoogle Scholar
- 51.Ashrafinia S, Gonzalez EM, Mohy-ud-Din H et al (2016) Adaptive PSF modeling for enhanced heterogeneity quantification in oncologic PET imaging. Nuc Med Med 57(suppl. 2):479Google Scholar