%0 Journal Article %A Mohammad R Salmanpour %A Mahdi Hosseinzadeh %A Seyed Masoud Rezaeijo %A Carlos Uribe %A Arman Rahmim %T Robustness and Reproducibility of Radiomics Features from Fusions of PET-CT Images %D 2022 %J Journal of Nuclear Medicine %P 3179-3179 %V 63 %N supplement 2 %X 3179 Introduction: Radiomics is a major frontier in medical image analysis, enabling mining of high-dimensional data from images. Although radiomics features (RF) are increasingly extracted via standardized radiomics software packages towards more reproducible research, employing different feature-generation hyperparameters, fusion techniques, and segmentation methods, may still lead to variable RFs. As such, employing RFs which are robust to processing variations is another important step towards reproducible study. The present work aims, specifically, to identify robust RFs that are less sensitive to different fusion techniques in head and neck (HN) cancer where fused PET-CT imaging hold significant value. To the best of our knowledge, no previous study has investigated the sensitivity of RFs to different fusion models in PET-CT imaging.Methods: 408 patients with HN cancer were extracted from the Cancer Imaging Archive (TCIA). In the pre-processing step, PET images were first registered to CT, enhanced, normalized, and cropped. We employed 15 typical image-level fusion techniques to combine PET and CT information: 1) Laplacian pyramid, 2) Ratio of low-pass pyramid, 3) Discrete wavelet transform, 4) Dual-tree complex wavelet transform (DTCWT), 5) Curvelet transform (CVT), 6) Nonsubsampled contourlet transform (NSCT), 7) Sparse representation (SR), 8) DTCWT+SR, 9) CVT+SR, 10) NSCT+SR, 11) Bilateral cross filter, 12) Wavelet Fusion, 13) Weighted Fusion, 14) Principal Component Analysis, and 15) Hue, Saturation and Intensity Fusion. Subsequently, 211 RFs were extracted from each region of interest in PET-only, CT-only, and 15 fused PET-CT images through the standardized SERA radiomics package. Variabilities of RFs were studied using the Intraclass Correlation Coefficient (ICC) (with carefully selected parameters, including for two-way random effects, absolute agreement and, multiple raters/measurements). ICC>0.90, 0.75<icc<0.90, 0.50=""></icc<0.90,>Results: Best category (vii) included 40 out of 211 RFs, including 24 morphological (Morph), 1 Neighbourhood-grey level dependence matrix-3D (NGLDM), 1 Neighbourhood-grey tone difference matrix-3D (NGTDM), 4 Distance zone matrix-3D (DZM), 2 Size zone matrix-3D (SZM), 1 Run length matrix-3D-merged (RLMM), 1 Run length matrix-3D- averaged (RLMA), 3 Co-occurrence matrix-3D-averaged (CMA), 1 Intensity volume histogram (IVH) and 2 Co-occurrence matrix-3D-merged (CMM) features. Category (vi) consisted of 22 features (4 DZM, 3 NGLDM, 2 IVH, 1 RLMA, 1 RLMM, 1SZM, 3 CMA and 4 CMM). Categories i, ii, iii, iv, and v, contained 58, 32, 3, 49, and 7 feartures, respectively. The features cover Morph, NGLDM, NGTDM, DZM, SZM, RLMM, RLMA, CMA, CMM, IVH, intensity histogram, local intensity, and statistics.Conclusions: We assessed the reproducibility of radiomics features to different fusion techniques, aiming to inform the pre-selection of robust radiomics features for further model development for tasks based on fused PET-CT images in HN cancer. %U