RT Journal Article SR Electronic T1 Establishment and Clinical verification of a mathematical model for diagnosing SPN with 18FDG PET/CT JF Journal of Nuclear Medicine JO J Nucl Med FD Society of Nuclear Medicine SP 1389 OP 1389 VO 56 IS supplement 3 A1 Yumei Chen A1 Gang Huang A1 Jianjun Liu YR 2015 UL http://jnm.snmjournals.org/content/56/supplement_3/1389.abstract AB 1389 Objectives To establish a mathematical model for diagnosing the solitary pulmonary nodules (SPN) with 18FDG PET/CT and to verify the established mathematical model.Methods In the first part, 182 patients with SPN (115 malignance, 67 benign) from Jan 2005 to Feb 2010 were collected in the study. Clinical data including 12 items (age, gender, maximum diameter, site, density, border of tumor and parenchyma, lobulation, spiculation, pleural retraction sign, vascular convergence, FDG uptake) were analyzed by univariate and multi-variate statistical method. The mathematical model was obtained from binary Logistic regression. In the second part, A prospective study included 109 patients with SPN (67 malignance, 42 benign) from Jan 2011 to Jun 2012. Clinical data including 7 items (age, density, border of tumor and parenchyma, lobulation, pleural retraction sign, vascular convergence sign, FDG uptake) were incorporated into the mathematical model. The diagnostic results were compared between the model and 2 senior doctors with rich experience in reading PET/CT imaging. Results The mathematical model established by binary Logistic regression was: p=ex/( 1+ ex)，ｘ= -4.146+0.041×age+2.226×density -1.053×border of tumor and parenchyma +1.211×lobulation +2.579×vascular convergence +1.954×pleural retraction sign +0.286×SUVmax. The clinical value for diagnosing the SPN with mathematical model was verified with sensitivity of 95.52%, specificity of 69.05%, positive predictive value of 83.12%, negative predictive value of 90.63%, accuracy of 85.32%. The diagnostic efficiency from doctors was also obtained with sensitivity of 97.01%, specificity of 52.38%, positive predictive value of 76.47%, negative predictive value of 91.67%, accuracy of 79.82%. The AUCs of mathematical model were 0.887±0.034, while the AUCs from the doctors were 0.747±0.053(p<0.05).Conclusions The mathematical model established by binary Logistic regression has high diagnostic value for estimating the character of SPN and will be used in clinical practice well.Research Support This work was supported by research training foundation of Shanghai Ren Ji hospital (No. RJPY10-006) and Shanghai health bureau youth research project (No. 20114Y182).