TY - JOUR T1 - Image-based biomarkers for low-risk NSCLC using texture analysis of F-18-FDG-PET/CT: determination of optimal parameters and their prognostic threshold JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 88 LP - 88 VL - 60 IS - supplement 1 AU - Katharina Kneer AU - Elham Yousefzadeh-Nowshahr AU - Joel Raacke AU - Stefan Rüdiger AU - Cornelia Kropf-Sanchen AU - Meinrad Beer AU - Vikas Prasad AU - Gerhard Glatting AU - Ambros Beer Y1 - 2019/05/01 UR - http://jnm.snmjournals.org/content/60/supplement_1/88.abstract N2 - 88Objectives: There is still need for better prognostic parameters in the primary workup in patients with newly diagnosed, low-risk NSCLC. In this study we aimed to a) find prognostic parameters for NSCLC using texture analyses (TA) of F-18 FDG PET/CT (PET/CT) and b) determination of their optimal prognostic threshold. Methods: Pre-therapy PET/CT data of 81 patients (age 66 a, 42-82 a, f 21, m 60; Grading (G) 1: 10; G2: 39; G3: 30; Gx: 2) with histologically confirmed NSCLC (squamous epithelium n= 38, adeno- n= 42, large cell n=3) with a median follow-up of 4.6 years (range 1.1-14.7) were retrospectively analysed. A volume of interest defined for the primary tumor on CT was analysed for PET and CT using Medical Image Processing, Analysis and Visualization (MIPAV; National Institutes of Health, Version 7.4.0). We focused on the following imaging parameters because of promising results in a first univariate analysis: volume for CT and solidity, circularity, kurtosis, excentricity and SUVmax for PET. While the Cox proportional hazard model is often used, our parameters showed no proportional hazard. Therefore we developed an alternative method for evaluation of the effects on survival time. The optimal thresholds were determined as follows: first, patient data were sorted in ascending order of the investigated parameter. Then different thresholds were defined using Matlab. Both groups as divided by the threshold were tested for significance using the log-rank test (p=0.05). Second, we defined the best threshold as the one with the lowest p-value and examined the clinical data of generated subgroups for differences. Third, we compared four groups, combining two parameters, e.g. SUVmax and CT-volume (Vol) (SUV-/Vol-, SUV+/Vol-, SUV+/Vol+, SUV-/Vol+). P-value was calculated with the log-rank test. Results: Poor negative prognostic value was observed in patients for CT-volume > 1434 mm2 (p=0.018; HR = 6.98, CI 2.75 - 17.71). PET-parameters showed significant negative prognostic value for SUVmax > 12.2 (p=0.003; HR=2.98, CI 1.33-6.66), solidity > 0.9810 (p=0.004; HR = 3.02, CI 1.03-8.89) or circularity > 0.1966 (p=0.009; HR = undefined because of missing death in one group). Skewness, kurtosis and excentricity in PET-TA showed no significant prognostic value. For small tumor volumes, prognosis was excellent independent of other PET / TA parameters. However, in larger tumors different subgroups could be defined by combining CT-volume and SUVmax or TA-parameters of PET. Tumors with larger volume and higher SUVmax showed significantly worse prognosis compared to tumors with smaller volume and lower SUVmax (p=0.018; HR=2.48, CI 1.12 - 5.49). Similar, tumors with larger CT-volume and higher PET-Solidity showed worse prognosis than those with higher volume and lower PET-Solidity (p=0.014; HR=2.43, CI 0.89 - 6.64). Furthermore, poor prognosis was seen in patients with higher SUVmax and higher PET-Solidity compared to lower SUVmax and lower PET-Solidity (p=0.016, HR=4.03, CI 0.85 - 19.07). Conclusions: We successfully applied a novel algorithm for determination of optimal parameters and their prognostic threshold for parameters not showing proportional hazard as assumed in the Cox proportional hazard model. Even in our homogenous low-risk NSCLC population with resectable tumors, powerful prognostic parameters could be defined for CT and PET using TA of FDG PET/CT, with SUVmax, PET-solidity and CT tumor volume being the best predictive parameters. Moreover our results suggest synergistic information by combining SUVmax and PET-solidity as well as combining PET parameters SUVmax / solidity and CT volume, which could help in defining a subgroup of patients with a significantly worse prognosis, who might profit from more aggressive neoadjuvant / adjuvant treatment. ER -