Locoregional Recurrence Prediction Using a Deep Neural Network of Radiological and Radiotherapy Images.

Kyumin Han,Wonmo Sung,Minhi Han,Joonyoung Francis Joung,Young-Nam Kang

doi:10.3390/jpm12020143

Abstract

Radiation therapy (RT) is an important and potentially curative modality for head and neck squamous cell carcinoma (HNSCC). Locoregional recurrence (LR) of HNSCC after RT is ranging from 15% to 50% depending on the primary site and stage. In addition, the 5-year survival rate of patients with LR is low. To classify high-risk patients who might develop LR, a deep learning model for predicting LR needs to be established. In this work, 157 patients with HNSCC who underwent RT were analyzed. Based on the National Cancer Institute’s multi-institutional TCIA data set containing FDG-PET/CT/dose, a 3D deep learning model was proposed to predict LR without time-consuming segmentation or feature extraction. Our model achieved an averaged area under the curve (AUC) of 0.856. Adding clinical factors into the model improved the AUC to an average of 0.892 with the highest AUC of up to 0.974. The 3D deep learning model could perform individualized risk quantification of LR in patients with HNSCC without time-consuming tumor segmentation.

Highlights

More than 650,000 cases of head and neck cancer including head and neck squamous cell carcinoma (HNSCC) have been reported worldwide annually, making it the seventh most common cancer [1,2]
To predict the prognosis of HNSCC, several methods based on radiomics have been reported [14,15]. Such methods based on radiomics have demonstrated that locoregional recurrence (LR), distant metastasis, and overall survival of HNSCC can be predicted
We proposed a convolutional neural network (CNN) to predict the LR of patients with HNSCC following radiation therapy [1,3,12,27,28,29,30]

Summary

Introduction

More than 650,000 cases of head and neck cancer including head and neck squamous cell carcinoma (HNSCC) have been reported worldwide annually, making it the seventh most common cancer [1,2]. Such methods based on radiomics have demonstrated that LR, distant metastasis, and overall survival of HNSCC can be predicted In these studies, predicting models were evaluated using area under the curve (AUC) of receiver operating characteristic (ROC) curves. Medical images and does distributions were not segmented Their features were not extracted to avoid the issue related to reproducibility [17,18,19,20]. We found that the deep learning model receiving three images and additional clinical factors as input could predict the LR most effectively. This will be discussed below in great detail

Materials and Methods

Conclusions