Improving the U-Net Configuration for Automated Delineation of Head and Neck Cancer on MRI

TL;DR

This study enhances the traditional U-Net architecture for automated head and neck tumor segmentation on MRI by optimizing configuration and training strategies, achieving improved accuracy in a competitive challenge setting.

Contribution

It demonstrates that simple modifications like patch-wise normalization, scheduled data augmentation, and Gaussian weighting can significantly improve U-Net segmentation performance.

Findings

Patch-wise normalization improves training and inference.

Scheduled data augmentation enhances model accuracy.

Gaussian weighting slightly boosts patch prediction fusion.

Abstract

Tumor volume segmentation on MRI is a challenging and time-consuming process that is performed manually in typical clinical settings. This work presents an approach to automated delineation of head and neck tumors on MRI scans, developed in the context of the MICCAI Head and Neck Tumor Segmentation for MR-Guided Applications (HNTS-MRG) 2024 Challenge. Rather than designing a new, task-specific convolutional neural network, the focus of this research was to propose improvements to the configuration commonly used in medical segmentation tasks, relying solely on the traditional U-Net architecture. The empirical results presented in this article suggest the superiority of patch-wise normalization used for both training and sliding window inference. They also indicate that the performance of segmentation models can be enhanced by applying a scheduled data augmentation policy during training. Finally, it is shown that a small improvement in quality can be achieved by using Gaussian weighting to combine predictions for individual patches during sliding window inference. The model with the best configuration obtained an aggregated Dice Similarity Coefficient (DSCagg) of 0.749 in Task 1 and 0.710 in Task 2 on five cross-validation folds. The ensemble of five models (one best model per validation fold) showed consistent results on a private test set of 50 patients with an DSCagg of 0.752 in Task 1 and 0.718 in Task 2 (team name: andrei.iantsen). The source code and model weights are freely available at www.github.com/iantsen/hntsmrg.