Search

Purpose : To evaluate the correlation of radiomic features in pelvic [2-deoxy-2-18F]fluoro-d-glucose positron emission tomography/magnetic resonance imaging and computed tomography ([18F]FDG PET/MRI and [18F]FDG PET/CT) in patients with primary cervical cancer (CCa). Procedures: Nineteen patients with histologically confirmed primary squamous cell carcinoma of the cervix underwent same-day [18F]FDG PET/MRI and PET/CT. Two nuclear medicine physicians performed a consensus reading in random order. Free-hand regions of interest covering the primary cervical tumors were drawn on PET, contrast-enhanced pelvic CT, and pelvic MR (T2 weighted and ADC) images. Several basic imaging features, standard uptake values (SUVmean, SUVmax, and SUVpeak), total lesion glycolysis (TLG), metabolic tumor volume (MTV), and more advanced texture analysis features were calculated. Pearson's correlation test was used to assess the correlation between each pair of features. Features were compared between local and metastatic tumors, and their role in predicting metastasis was evaluated by receiver operating characteristic curves. Results: For a total of 101 extracted features, 1104/5050 pairs of features showed a significant correlation (ρ ≥ 0.70, p < 0.05). There was a strong correlation between 190/484 PET pairs of features from PET/MRI and PET/CT, 91/418 pairs of CT and PET from PET/CT, 79/418 pairs of T2 and PET from PET/MRI, and 50/418 pairs of ADC and PET from PET/MRI. Significant difference was seen between eight features in local and metastatic tumors including MTV, TLG, and entropy on PET from PET/CT; MTV and TLG on PET from PET/MRI; compactness and entropy on T2; and entropy on ADC images. Conclusions: We demonstrated strong correlation of many extracted radiomic features between PET/MRI and PET/CT. Eight radiomic features calculated on PET/CT and PET/MRI were significantly different between local and metastatic CCa. This study paves the way for future studies to evaluate the diagnostic and predictive potential of radiomics that could guide clinicians toward personalized patients care. [ABSTRACT FROM AUTHOR]