HR-MPF: high-resolution representation network with multi-scale progressive fusion for pulmonary nodule segmentation and classification

Ling Zhu,Yang Yu,Hongqing Zhu,Suyi Yang,Pengyu Wang

doi:10.1186/s13640-021-00574-2

Abstract

Accurate segmentation and classification of pulmonary nodules are of great significance to early detection and diagnosis of lung diseases, which can reduce the risk of developing lung cancer and improve patient survival rate. In this paper, we propose an effective network for pulmonary nodule segmentation and classification at one time based on adversarial training scheme. The segmentation network consists of a High-Resolution network with Multi-scale Progressive Fusion (HR-MPF) and a proposed Progressive Decoding Module (PDM) recovering final pixel-wise prediction results. Specifically, the proposed HR-MPF firstly incorporates boosted module to High-Resolution Network (HRNet) in a progressive feature fusion manner. In this case, feature communication is augmented among all levels in this high-resolution network. Then, downstream classification module would identify benign and malignant pulmonary nodules based on feature map from PDM. In the adversarial training scheme, a discriminator is set to optimize HR-MPF and PDM through back propagation. Meanwhile, a reasonably designed multi-task loss function optimizes performance of segmentation and classification overall. To improve the accuracy of boundary prediction crucial to nodule segmentation, a boundary consistency constraint is designed and incorporated in the segmentation loss function. Experiments on publicly available LUNA16 dataset show that the framework outperforms relevant advanced methods in quantitative evaluation and visual perception.

Highlights

Diagnosis and treatment of lung cancer can reduce mortality rate of patients, and early manifestation of lung cancer is mainly pulmonary nodules [1]
The shape, size, growth position and other characteristics can be observed from precise segmentation results of various types of pulmonary nodules in Computed Tomography (CT) images, which provide a reference for the classification of benign and malignant pulmonary nodules
5 Conclusion In this paper, an effective multi-task framework is designed for pulmonary nodule segmentation and classification, which can contribute to clinical diagnosis of pulmonary nodules

Summary

Introduction

Diagnosis and treatment of lung cancer can reduce mortality rate of patients, and early manifestation of lung cancer is mainly pulmonary nodules [1]. We design a network based on adversarial training scheme for pulmonary nodule segmentation and classification at one time. High-Resolution network with Multi-scale Progressive Fusion (HR-MPF) is proposed based on High-Resolution Network (HRNet) [3]. In HR-MPF, modified boosted module is inserted in the network in multi-scale progressive feature fusion manner to deliver spatial and context information from different resolutions. Corresponding to the feature extraction network HR-MPF, a Progressive Decoding Module (PDM) is proposed to recover the pixel-wise segmentation prediction from the output of HR-MPF. An HR-MPF enhancing feature communication of all scales is proposed in which the boosted module is introduced to HRNet based on multi-scale progressive fusion strategy;. A PDM is proposed to recover final pixel segmentation prediction in progressive fusion manner and refine the output from HR-MPF;. A reasonably designed multi-task loss function jointly optimizes the whole framework

Related works

Results and discussion

Evaluation Evaluation

Methods

Conclusion