Deep learning nodal network.

An important addition to the AJCC head and neck cancer staging was the inclusion of both clinical and pathologic extranodal extension (ENE). Clinical (i.e., radiographic) ENE upstages straight to N3, but far more patients have pathologic ENE than is clinically apparent. In this study, a previously developed deep learning neural network that detects ENE on pre-operative neck CT was validated on 200 individual lymph nodes from an external, multi-institutional cohort as well as the Cancer Genome Atlas. Compared to the expert radiologists, AI had significantly higher accuracy in the external validation (83%, AUC 0.84) and TCGA (87%, AUC 0.9) cohorts. When radiologists had the AI data to consider, their accuracy improved. TBL: AI is highly accurate at predicting the presence of extranodal extension in patients with head and neck cancer. | Kann, J Clin Oncol 2019


Popular Posts