Automatic lesion and lymph node segmentation from PET and CT scans of the Head and Neck region: a HECKTOR 2025 Challenge Report

Sébastien Quetin, Shirin Abbasinejad Enger

Keywords: Head and Neck tumor, Deep Learning segmentation, PET and CT scans.

TL;DR: This paper describes our method for the task1 of the HECKTOR 2025 challenge.

Abstract: The HECKTOR 2025 challenge provides a platform to benchmark automatic segmentation methods for Head and Neck (H&N) primary tumors and lymph nodes in FDG-PET and CT scans (Task 1). This study presents a challenge submission based on the nnU-Net framework. PET scans were first resampled to the CT resolution, after which a pseudo brain mask was derived from the PET scan to guide the cropping of both modalities. A custom clipping-based normalization was applied to the PET scan, and the paired PET and CT volumes were then processed by a Residual Encoder U-Net. Training was performed using custom augmentations. The proposed solution, submitted under the user name sebquet, achieved a mean Dice Score of 75.09% for primary tumors and 77.04% for metastatic lymph nodes on the validation set. These results demonstrate competitive performance within the challenge and highlight the effectiveness of combining modality-specific preprocessing with residual encoder architectures for H&N tumor segmentation.

Submission Number: 15