Aims and objectives
•A phantom study to select reproducible textural features in 18F-FDG PET/CT image analysis.
•A clinical study where Texture analysis is employed in predicting outcomes of patient affected by Head and Neck Squamous Cell Carcinoma (HNSCC).
Introduction:It has been proven that tumors often show high heterogeneity on microscopic level. Standard clinical procedure for tumor heterogeneity scoring is biopsy but unfortunately this technique is highly invasive and sometimes requires a large specimen of the tumor itself[1][2].
Current medical imaging techniques cannot provide resolutions up to the microscopic...
Methods and materials
Phantom Study:
• NEMA IEC Body Phantom Set™ (Data Spectrum Corporation,
Hillsborough,
NC) see fig.1.
• Spheres of different radius were equally filled varying Target/Background(TB) activity concentration of 18F-FDG and scanned for different acquisition times
• PET images were reconstructed on a 256×256 image matrix with 40 equivalent iteration and varying smoothing filter width
• Spheres contour segmentation was performed on dedicated Workstations using a 40% SUVMAX threshold algorithm
Clinical Study:
• A dataset of 121 patients with HNSCC,
who underwent pre-treatment 18F-FDG PET/CT and...
Results
Phantom study: textural features variation examples(left) and relative correction proposal(right) can be seen in Fig.2 and Fig.3.
•BUSYNESS: linear-scaling with volume (standard behavior)
•ENTROPY: constant behavior with volumes up to 15 ml
•LILRE: artifacts may occur due to segmentation algorithm(circled in red),
instabilities varying TB,
a trend with increasing volume can be recognized
Clinical Study:
SUV based statistics results areshown in Fig.4
Tumor Volume and textural analysis resultsare shown in Fig.5.
Better patient stratification can be obtained using tumor volume and LILRE as predictors of...
Conclusion
Through the phantom study we found out that textural feature can exhibit complex behavior,
in particular:
• Variation with VOI volume analyzed •Artifacts coming from the use of segmentation algorithm •Variation with differences in Target/Background ratio
•Correction proposed in fig.2 and 3 should be taken into account in evaluating patient for classification purposes.
Clinical Study points out that textural features is superior to standard SUV statistics for oncologic patient stratification.
Moreover texture analysis can be used together with classical parameters such as tumor volume to...
Personal information
Giacomo Feliciani
University of Bologna,
Bologna,
Italy
Email:
[email protected]
Phone: +39 3274730398
References
[1].Gerlinger M.
et al.
Intratumor eterogeneity and branched evolution revealed by multiregion sequencing.
N Engl J Med.
2012; 366:883-892
[2].Davnal F.
et al.
Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice? Insight Imaging. 2012; 3:573-589
[3].Tesar L.
et al.
Medical image analysis of 3D CT images based on extension of Haralick texture feautres.
Comp Med Imag and Graph 2008; 32:513-520
[4].Ben- Haim S.
et al.
18F-FDG PET and PET/CT in the Evaluation of Cancer Treatment Response. J Nucl Med.
2009; 50:88-99
[5].El...