Can texture indices derived from PET images differentiate tissue types?
Résumé
Aim: Texture indices (TI) are of growing interest for tumor characterization. Yet, whether FDG-PET images can evidence tissue-specific pattern has received little attention so far. We studied the ability of enhanced TI to determine tissue types. Materials and Methods: Forty-eight patients with non-small cell lung cancer underwent FDG-PET before treatment. Seven enhanced TI were calculated using a new resampling method. Standardized Uptake Value (SUV) and metabolic volume (MV) were also systematically computed. The ability of each index to distinguish between tumor and liver tissue and between two subtypes of cancer was investigated using ROC analyses. Results: All enhanced TI could differentiate tumor from liver tissue with an Area Under the ROC Curve (AUC) higher than 0.692. Homogeneity and Low Gray-Level Emphasis could differentiate the adenocarcinomas (n=28) and squamous cell carcinomas (n=12) with AUC better than that of SUVmax and MV (Delong’s test). Liver tissue had a more homogeneous texture than tumor tissue and adenocarcinomas exhibited a more homogeneous texture than squamous cell carcinomas. Conclusion: Enhanced TI vary as a function of the tissue type and cancer subtype, and might be used as a new tool for tumor characterization.
Domaines
Ingénierie biomédicale
Origine : Fichiers produits par l'(les) auteur(s)