Development and Validation of a Predictive Model for Metastatic Melanoma Patients Treated with Pembrolizumab Based on Automated Analysis of Whole-Body [(18)F]FDG PET/CT Imaging and Clinical Features

SIMPLE SUMMARY: Blockade of the programmed cell death protein 1 (PD-1) receptor is an established standard-of-care treatment option that significantly improves survival of patients with advanced melanoma. While a smaller proportion of the population can derive a durable remission (even cure), most patients immediately or eventually develop disease progression. Prediction of upfront resistance to therapy as well as durable responders based on biomarkers that correlate with survival is key in selecting an optimal personalised treatment plan. Previously we reported that total metabolic tumour volume (TMTV) determined by whole-body [[Formula: see text] F]FDG PET/CT is a baseline predictive biomarker that deserves further investigation. A fully automated method is proposed for feature extraction from whole-body [[Formula: see text] F]FDG PET/CT. The automatically and manually derived parameters produced similar results in both the feature analysis and survival prediction. This automation can offer a fast, objective and reproducible assessment of TMTV and facilitate further exploration and validation of predictive models on larger datasets. ABSTRACT: Background: Antibodies that inhibit the programmed cell death protein 1 (PD-1) receptor offer a significant survival benefit, potentially cure (i.e., durable disease-free survival following treatment discontinuation), a substantial proportion of patients with advanced melanoma. Most patients however fail to respond to such treatment or acquire resistance. Previously, we reported that baseline total metabolic tumour volume (TMTV) determined by whole-body [[Formula: see text] F]FDG PET/CT was independently correlated with survival and able to predict the futility of treatment. Manual delineation of [[Formula: see text] F]FDG-avid lesions is however labour intensive and not suitable for routine use. A predictive survival model is proposed based on automated analysis of baseline, whole-body [[Formula: see text] F]FDG images. Methods: Lesions were segmented on [[Formula: see text] F]FDG PET/CT using a deep-learning approach and derived features were investigated through Kaplan–Meier survival estimates with univariate logrank test and Cox regression analyses. Selected parameters were evaluated in multivariate Cox survival regressors. Results: In the development set of 69 patients, overall survival prediction based on TMTV, lactate dehydrogenase levels and presence of brain metastases achieved an area under the curve of 0.78 at one year, 0.70 at two years. No statistically significant difference was observed with respect to using manually segmented lesions. Internal validation on 31 patients yielded scores of 0.76 for one year and 0.74 for two years. Conclusions: Automatically extracted TMTV based on whole-body [[Formula: see text] F]FDG PET/CT can aid in building predictive models that can support therapeutic decisions in patients treated with immune-checkpoint blockade.