Prognosis and Treatment of Non–Small Cell Lung Cancer in the Age of Deep Learning
Menée à partir de données portant sur 17 322 patients atteints d'un cancer du poumon non à petites cellules de stade I à IV diagnostiqué entre 2009 et 2015 (durée médiane de suivi : 24 mois), cette étude évalue la performance d'un algorithme d'apprentissage automatique, basé sur des critères clinico-pathologiques, pour prédire la survie des patients
Lung cancer is both the most common and the most deadly cancer, with more than 2 million cases diagnosed worldwide in 2018 per Global Cancer Observatory estimates and with non–small cell lung cancer (NSCLC) accounting for the great majority of cases. The 8th edition of the American Joint Committee on Cancer TNM stage groupings represents the most well-validated prognostic metric for NSCLC. However, a variety factors unaccounted for by these TNM stage groupings affect outcome, from patient-specific factors such as performance status, age, and socioeconomic status to tumor characteristics such as grade, lymphovascular invasion, programmed cell death 1 ligand 1 expression, and the presence of molecular driver variants. Integrating the various clinical and pathologic characteristics of each case to provide an accurate prognosis is challenging in the absence of easy-to-use and comprehensive predictive models.
JAMA Network Open , Article en libre accès, 2019