Comparing 3D deformations between longitudinal daily CBCT acquisitions using CNN for head and neck radiotherapy toxicity prediction

TL;DR

This study demonstrates that daily CBCT imaging during head and neck radiotherapy can predict severe toxicities using a CNN-based deformable registration and classification pipeline, enabling early intervention.

Contribution

The paper introduces a novel multi-branch 3D CNN and deformable registration method to predict toxicity from longitudinal CBCTs in radiotherapy, showing early predictive accuracy.

Findings

Radionecrosis prediction accuracy: 85.8%.

Hospitalization prediction accuracy: 75.3%.

NG tube risk prediction improves over treatment weeks, reaching 83.1%.

Abstract

Adaptive radiotherapy is a growing field of study in cancer treatment due to it's objective in sparing healthy tissue. The standard of care in several institutions includes longitudinal cone-beam computed tomography (CBCT) acquisitions to monitor changes, but have yet to be used to improve tumor control while managing side-effects. The aim of this study is to demonstrate the clinical value of pre-treatment CBCT acquired daily during radiation therapy treatment for head and neck cancers for the downstream task of predicting severe toxicity occurrence: reactive feeding tube (NG), hospitalization and radionecrosis. For this, we propose a deformable 3D classification pipeline that includes a component analyzing the Jacobian matrix of the deformation between planning CT and longitudinal CBCT, as well as clinical data. The model is based on a multi-branch 3D residual convolutional neural network, while the CT to CBCT registration is based on a pair of VoxelMorph architectures. Accuracies of 85.8% and 75.3% was found for radionecrosis and hospitalization, respectively, with similar performance as early as after the first week of treatment. For NG tube risk, performance improves with increasing the timing of the CBCT fraction, reaching 83.1% after the $5_{th}$ week of treatment.