TY - JOUR T1 - Correlation of anatomical and functional information from PET-CT images JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 2266 LP - 2266 VL - 53 IS - supplement 1 AU - Ulas Bagci AU - Omer Aras AU - Daniel Mollura Y1 - 2012/05/01 UR - http://jnm.snmjournals.org/content/53/supplement_1/2266.abstract N2 - 2266 Objectives To investigate the degree of correlation between texture features on quantified CT and SUV measurements on PET. Methods With IRB approval, we retrospectively analyzed 20 18F-FDG-PET/CTs of patients with infectious lung diseases (having radiographic findings of ground-glass opacity, linear thickening, nodules, consolidation, pleural effusion, and tree-in-bud) such that 10 of which were follow-up scans, 10 were baseline. For the accurate analysis of uptake regions from PET and corresponding anatomical objects from CT scans, we developed a robust and highly accurate segmentation method to detect and delineate interesting uptake regions (IURs) from PET images, and corresponding anatomical objects from CT images automatically. Standardized uptake value (SUVmax) and several textural features (i.e., gray level co-occurrence matrix (GLCM) based features) were extracted from delineated regions of PET and CT, respectively. Multivariate and Bayesian statistical analysis were used to test possible relations between anatomical and functional objects. Results There was a good correlation (spearman >0.80 for MaxPr feature, p<0.01) between CT and PET images based on the abnormal imaging patterns (anatomical and functional). Fig.1 showed boxplots of the textural features from CT scans compared to SUVmax from PET scans. In addition, Fig.2 shows that Kurtosis, histogram, and skewness measures were quite descriptive in identifying diseased regions in PET images (mean skewness in initial scan=-0.028, in follow-up=0.593, kurtosis in initial scan=-0.842, in follow-up=-0.500, and the differences were significant at a level of p<0.01). Finally, spearman correlations between baseline and followup scans using texture, SUVmax, and both suggests possible predictive utility of combined data (Fig.3). Conclusions Results support the role of longitudinal image analysis in facilitating quantitative analysis of serial of PET/CT images and efficiently assess the infectious lung disease in a comprehensive software platform. Research Support NIH Intramural grants ER -