Pulmonary nodules detection based on multi-scale attention networks

Hui Zhang,Yanjun Peng,Yanfei Guo

doi:10.1038/s41598-022-05372-y

Hui Zhang, Yanjun Peng + Show 1 more

Open Access

https://doi.org/10.1038/s41598-022-05372-y

Copy DOI

Abstract

Pulmonary nodules are the main manifestation of early lung cancer. Therefore, accurate detection of nodules in CT images is vital for lung cancer diagnosis. A 3D automatic detection system of pulmonary nodules based on multi-scale attention networks is proposed in this paper to use multi-scale features of nodules and avoid network over-fitting problems. The system consists of two parts, nodule candidate detection (determining the locations of candidate nodules), false positive reduction (minimizing the number of false positive nodules). Specifically, with Res2Net structure, using pre-activation operation and convolutional quadruplet attention module, the 3D multi-scale attention block is designed. It makes full use of multi-scale information of pulmonary nodules by extracting multi-scale features at a granular level and alleviates over-fitting by pre-activation. The U-Net-like encoder-decoder structure is combined with multi-scale attention blocks as the backbone network of Faster R-CNN for detection of candidate nodules. Then a 3D deep convolutional neural network based on multi-scale attention blocks is designed for false positive reduction. The extensive experiments on LUNA16 and TianChi competition datasets demonstrate that the proposed approach can effectively improve the detection sensitivity and control the number of false positive nodules, which has clinical application value.

Highlights

Pulmonary nodules are the main manifestation of early lung cancer
It effectively alleviates network over-fitting by pre-activation and further improves the detection sensitivity by convolutional quadruplet attention module (CQAM). (b) To fully exploit the 3D Computed Tomography (CT) images, this paper proposes a 3D Faster R-convolutional neural network (CNN) based on 3D MSA blocks and a U-net-like encoder-decoder structure to automatically detect pulmonary nodules, and a 3D deep multi-scale attention networks to reduce false positive numbers. (c) The proposed system achieves a competition performance metric (CPM) score of 0.927 on the LUNA16 dataset, which indicates that the model has excellent performance for accurate nodule detection
The model based on 3D CNN is a common method for pulmonary nodule detection, which can fully extract the 3D spatial information of the nodules and has a significant effect on the detection of pulmonary nodules with a specific size

Summary

Introduction

Pulmonary nodules are the main manifestation of early lung cancer. accurate detection of nodules in CT images is vital for lung cancer diagnosis. (b) To fully exploit the 3D CT images, this paper proposes a 3D Faster R-CNN based on 3D MSA blocks and a U-net-like encoder-decoder structure to automatically detect pulmonary nodules, and a 3D deep multi-scale attention networks to reduce false positive numbers.

Results

Conclusion