Role of Textural Analysis Parameters Derived from FDG PET/CT in Diagnosing Cardiac Sarcoidosis

Abstract

Introduction Texture and radiomic analysis characterize the lesion's phenotype and evaluate its microenvironment in quantitative terms. The aim of this study was to investigate the role of textural features of 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography–computed tomography (PET/CT) images in differentiating patients with cardiac sarcoidosis (CS) from patients with physiologic myocardial uptake. Methods This is a retrospective, single-center study of 67 patients, 17 diagnosed CS patients, and 50 non-CS patients. These patients underwent FDG PET/CT for the diagnosis of CS. The non-CS group underwent 18F-FDG PET/CT for other oncological indications. The PET/CT images were then processed in a commercially available textural analysis software. Region of interest was drawn over primary tumor with a 40% threshold and was processed further to derive 92 textural and radiomic parameters. These parameters were then compared between the CS group and the non-CS group. Receiver operating characteristics (ROC) curves were used to identify cutoff values for textural features with a p-value < 0.05 for statistical significance. These parameters were then passed through a principle component analysis algorithm. Five different machine learning classifiers were then tested on the derived parameters. Results A retrospective study of 67 patients, 17 diagnosed CS patients, and 50 non-CS patients, was done. Twelve textural analysis parameters were significant in differentiating between the CS group and the non-CS group. Cutoff values were calculated for these parameters according to the ROC curves. The parameters were Discretized_HISTO_Entropy, GLCM_Homogeneity, GLCM_Energy, GLRLM_LRE, GLRLM_LGRE, GLRLM_SRLGE, GLRLM_LRLGE, NGLDM_Coarseness, GLZLM_LZE, GLZLM_LGZE, GLZLM_SZLGE, and GLZLM_LZLGE. The gradient boosting classifier gave best results on these parameters with 85.71% accuracy and an F1 score of 0.86 (max 1.0) on both classes, indicating the classifier is performing well on both classes. Conclusion Textural analysis parameters could successfully differentiate between the CS and non-CS groups noninvasively. Larger multicenter studies are needed for better clinical prognostication of these parameters.