The prognostic role of end-of-treatment FDG-PET/CT in diffuse large B cell lymphoma: a pilot study application of neural networks to predict time-to-event.

摘要: To evaluate the prognostic role of end-of-treatment (EoT) FDG-PET/CT parameters in diffuse large B cell lymphoma (DLBCL), and then to explore a pilot application Neural Networks (NN) predicting time-to-relapse. For conventional survival analysis, as Deauville score (DS) quantitative extension DS (qPET) were correlated adverse events relapse or progression follow-up. build NN multi-regression models (MM) for time-to-event prediction, patients with residual FDG uptake (DS ≥ 2) an event divided into training test group. Models developed on group evaluated Pearson correlation coefficient (R) mean relative error between observed forecasted calculated. data 308 DLBCL analyzed. qPET factors univariate analysis. Positive negative predictive values, respectively, 55% 83% 4–5, 89% 82% positive qPET. Focusing 37 relapsed uptake, R was 0.63 model 0.49 MM. Mean 58% 67% EoT visual is strong outcome predictor monocentric cohort. The semi-quantitative parameter may increase this performance. A applied seems predict