(E-187) Heterogeneity Analysis in PET/CT Imaging: A Simulation and Phantom Study for Parameter Selection Optimization

Introduction:: PET-CT is a powerful imaging modality that combines the functional information provided by positron emission tomography (PET) with the anatomical information provided by X-ray computed tomography (CT), allowing for a more comprehensive assessment of tissue architecture and disease progression. Conventional analysis of PET-CT images involves identifying specific volumes of interest (VOIs) and measuring the intensity of voxels within those regions. This technique has proven valuable for assessing tumors, but recent studies have underscored the importance of tumor heterogeneity in understanding their invasiveness. In light of this, our group is exploring the potential of image heterogeneity as a valuable tool for molecular imaging of tumors. By evaluating the distribution and diversity of voxel intensities across a tumor, we hope to gain deeper insights into its characteristics and behavior. Currently, there is a lack of standardization and consensus on the most appropriate heterogeneity parameters to use in different imaging applications. Additionally, the impact of image acquisition and processing parameters on heterogeneity analysis is not fully understood. The goal of this study is to optimize heterogeneity parameter selection in PET-CT.


Materials and Methods:: To address the aim of this study, we used a combination of simulated and phantom images processed by applying various levels of noise and reconstructed with different algorithms. In this study, we generated three images of different levels of heterogeneity and acquired similar heterogeneous PET-CT images of a 3D-printed phantom filled with a Copper-64 isotope. These images were modified by obtaining images at (30,60,120,180,240,300,600,900,1800 and 3600 seconds) and reconstructed with a 2D FBP or iterative 3D OSEM-SD algorithm with various numbers of iterations (18, 50 and 100), filter levels (Colsher, Colsher+Hann, Hamm, Shepp, Ramp, Butterworth). A total of 109 textural features have been extracted from these images using LifeX software. The statistical analysis was performed with the SAS software package.

Results, Conclusions, and Discussions:: Heterogeneity scores from our preliminary results reveal that both processing and reconstruction parameters, singly or overlapping with other parameters, influence the quantification of heterogeneity of images. By checking for repeatability of the textural features using coefficient of variation (CoV) and performing ANOVA for the 109 textural parameters, we have identified a subset of the most important parameters which we can use to build a computational model that accurately predicts heterogeneity. This early evaluation also determines the proper imaging workflow and the optimal choice of image acquisition and processing settings when quantifying heterogeneity in PET-CT images.
Our study aimed to improve the accuracy and reliability of PET-CT imaging by evaluating a range of heterogeneity parameters. To achieve this, we analyzed both simulated and phantom PET-CT images using a comprehensive and systematic approach. Based on our evaluation, we identified a subset of heterogeneity parameters that showed promise in enhancing the image analysis beyond the conventional approaches. These parameters will be further optimized using PET-CT images acquired in vivo in mouse tumor xenografts. Our successful evaluation of these parameters provides a valuable foundation for future work in improving the quality and precision of PET-CT imaging in oncological applications.


Acknowledgements (Optional): :

References (Optional): :