Deep learning for automatic tumour segmentation in PET/CT images of patients with head and neck cancers

Yngve Mardal Moe, Aurora Rosvoll Groendahl, Martine Mulstad, Oliver Tomic, Ulf Indahl, Einar Dale, Eirik Malinen, Cecilia Marie Futsaether

Keywords: Deep learning, segmentation, delineation, head and neck cancer, HNC, PET/CT, GTV delineation

TL;DR: This work demonstrates that neural networks can achieve close-to-oncologist level delineations in PET/CT images of patients with head and neck cancers.

Abstract: An automatic segmentation algorithm for delineation of the gross tumour volume and pathologic lymph nodes of head and neck cancers in PET/CT images is described. The proposed algorithm is based on a convolutional neural network using the U-Net architecture. Several model hyperparameters were explored and the model performance in terms of the Dice similarity coefficient was validated on images from 15 patients. A separate test set consisting of images from 40 patients was used to assess the generalisability of the algorithm. The performance on the test set showed close-to-oncologist level delineations as measured by the Dice coefficient (CT: 0.65 ± 0.17, PET: 0.71 ± 0.12, PET/CT: 0.75 ± 0.12).

Code Of Conduct: I have read and accept the code of conduct.