AUTHOR=Yang Geng , Dai Zhenhui , Zhang Yiwen , Zhu Lin , Tan Junwen , Chen Zefeiyun , Zhang Bailin , Cai Chunya , He Qiang , Li Fei , Wang Xuetao , Yang Wei 

TITLE=Multiscale Local Enhancement Deep Convolutional Networks for the Automated 3D Segmentation of Gross Tumor Volumes in Nasopharyngeal Carcinoma: A Multi-Institutional Dataset Study

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.827991

DOI=10.3389/fonc.2022.827991

ISSN=2234-943X

ABSTRACT=Purpose: Accurate segmentation of gross target volume (GTV) from computed tomography (CT) images is a prerequisite in radiotherapy for nasopharyngeal carcinoma (NPC). However, this task is very challenging due to the low contrast at the boundary of the tumor and the great variety of sizes and morphologies of tumors between different stages. Meanwhile, the data source also seriously affect the results of segmentation. In this paper, we propose a novel three-dimensional (3D) automatic segmentation algorithm that adopts cascaded multiscale local enhancement of convolutional neural networks (CNNs) and conduct experiments on multi-institutional datasets to address the above problems. 
Materials and methods: In this study, we retrospectively collected CT images of 257 NPC patients to test the performance of the proposed automatic segmentation model, and conducted experiments on two additional multi-institutional datasets. Our novel segmentation framework consists of three parts. First, the segmentation framework is based on a 3D Res-UNet backbone model that has excellent segmentation performance. Then, we adopt a multiscale dilated convolution block to enhance the receptive field and focus on the target area and boundary for segmentation improvement. Finally, a central localization cascade model for local enhancement is designed to concentrate on the GTV region for fine segmentation to improve the robustness. The DSC, PPV, SEN, ASSD and HD95 are utilized as qualitative evaluation criteria to estimate the performance of our automated segmentation algorithm.
Results: The final DSC, SEN, ASSD and HD95 values can be reached to 76.23±6.45%, 79.14±12.48%, 1.39±5.44mm, 4.72±3.04mm. In addition, the outcomes of multi-institution experiments demonstrate that the segmentation result of other datasets acquired from the single data training model is unsatisfactory, the DSC values were 68.21±5.51% and 61.64±13.55% for the two additional institutions. But it can be solved by transfer learning. After transfer learning the DSC values raised to 76.71±6.97% and 67.03±11.45% respectively.
Conclusions: The proposed algorithm could accurately segment NPC in CT images from multi-institutional datasets and thereby may improve and facilitate clinical applications.