ICARE: winning solution for the MICCAI 2022 challenge: HEad and neCK TumOR (HECKTOR) for the outcome prediction task from PET/CT.

Created as a part of the HEad and neCK TumOR segmentation and outcome prediction in PET/CT images challenge, 3rd edition (HECKTOR 2022), the Individual Coefficient Approximation for Risk Estimation (ICARE) was ranked as the winning solution. Developped in collaboration with LITO research team, this survival model uses a minimal learning strategy to reduce to risk of overfitting on the often noisy and censored survival data. For ensembling strategies, it also includes a dedicated bagging aggregator. This easy-to-use package was presented by Louis Rebaud and Thibault Escobar at the 25th international conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) congress in Singapore.

 “Head and Neck Tumor and Lymph Node Segmentation and Outcome Prediction from 18F-FDG PET/CT Images: Simplicity is All You Need” preprint paper is available here.