Pulmonary nodule detection algorithm based on attention feature pyramid networks

doi:10.11772/j.issn.1001-9081.2022060924

Abstract

Abstract:

Aiming at the problems of low sensitivity and high false positive rate caused by various shapes and difficulty in detecting pulmonary nodules by the Computer-Aided Detection （CAD） system of pulmonary nodules， a pulmonary nodule detection algorithm based on attention feature pyramid networks was proposed. In the first stage， a more compact Dual Path Network （DPN） was used as the backbone network， and a Feature Pyramid Network （FPN） was combined for multi-scale prediction to obtain feature information at different levels. At the same time， the Global Attention Mechanism （GAM） was embedded to refine the semantic features to be emphasized in learning and improve the sensitivity of the algorithm. In the second stage， a false positive reduction network was proposed to obtain the final classification prediction results. In the training stage， the focal loss function and various data augmentation techniques were used to deal with the data imbalance problem. Experimental results on the public dataset LUNA16 （LUng Nodule Analysis 2016） show that the Competitive Performance Metric （CPM） of the algorithm only with the first stage reaches 0.908， and after adding the false positive reduction network， the CPM of the algorithm reaches 0.933， which is 1.1 percentage points higher than that of the classic algorithm — Convolutional Neural Network （CNN） based on Maximum Intensity Projection （MIP）. And ablation experimental results show that the dual path network， FPN， and GAM are effective in improving the detection sensitivity. The above proves that the proposed two-stage detection algorithm can obtain multi-scale nodule information， improve the sensitivity of pulmonary nodule detection， and reduce the false positive rate.

Key words: pulmonary nodule detection, attention mechanism, Feature Pyramid Network (FPN), false positive reduction, Convolutional Neural Network (CNN)

摘要：

针对肺结节计算机辅助检测（CAD）系统中肺结节形态各异难以检测带来的敏感度低、假阳性率高的问题，提出一种基于注意力特征金字塔网络的肺结节检测算法。在第一阶段，以更加紧凑的双路径网络（DPN）为骨干网络，并结合特征金字塔网络（FPN）进行多尺度预测，以获取不同层次的特征信息，同时嵌入全局注意力机制（GAM）来细化学习要强调的语义特征，并提高算法的敏感度；在第二阶段，提出一种假阳性抑制网络，以获得最终分类预测结果；在训练阶段，采用焦点损失函数和多种数据增强技术来处理数据不平衡问题。在公开数据集LUNA16 （LUng Nodule Analysis 2016）上的实验结果显示：仅有第一阶段的算法的竞争性能指标（CPM）达到了0.908，而加入假阳性抑制网络后算法的CPM达到了0.933，这与经典算法基于最大强度投影（MIP）的卷积神经网络（CNN）算法相比提升了1.1个百分点；而消融实验的结果表明DPN、FPN、GAM对于提升检测敏感度是有作用的。以上证明了所提出的两阶段检测算法可以获取多尺度结节信息，提高肺结节检测的敏感度，并且降低假阳性率。

关键词: 肺结节检测, 注意力机制, 特征金字塔网络, 假阳性抑制, 卷积神经网络

CLC Number:

TP391.4

Yuanyuan QIN, Hong ZHANG. Pulmonary nodule detection algorithm based on attention feature pyramid networks[J]. Journal of Computer Applications, 2023, 43(7): 2311-2318.

秦源源, 张鸿. 基于注意力特征金字塔网络的肺结节检测算法[J]. 《计算机应用》唯一官方网站, 2023, 43(7): 2311-2318.

Figures/Tables 14

Fig. 1 Improved pulmonary nodule detection flow

Fig. 2 Network framework for candidate nodule generation based on DPAFPN

Fig. 3 Residual block and dual connection module

Fig. 4 Network structure of GAM

Fig. 5 Channel attention module

Fig. 6 Spatial attention module

Fig. 7 False positive reduction network framework

Fig. 8 Pulmonary nodules with various shapes

Tab. 1 Comparison of pulmonary nodules detection performance of different algorithms

算法	不同假阳性数下的敏感度							CPM
算法	0.125	0.25	0.5	1	2	4	8	CPM
文献［18］算法	0.594	0.727	0.781	0.844	0.875	0.891	0.898	0.801
文献［19］算法	0.677	0.737	0.815	0.848	0.879	0.907	0.922	0.827
文献［9］算法	0.692	0.769	0.824	0.865	0.893	0.917	0.933	0.842
文献［15］算法	0.678	0.772	0.836	0.884	0.918	0.940	0.951	0.854
文献［11］算法	0.739	0.803	0.858	0.888	0.907	0.916	0.920	0.862
文献［14］算法	0.712	0.809	0.854	0.889	0.915	0.930	0.942	0.864
文献［20］算法	0.712	0.802	0.865	0.901	0.937	0.946	0.955	0.874
文献［21］算法	0.676	0.776	0.879	0.949	0.958	0.958	0.958	0.878
文献［22］算法	0.748	0.853	0.887	0.922	0.938	0.944	0.946	0.891
文献［13］算法	0.788	0.876	0.916	0.921	0.928	0.936	0.944	0.901
文献［8］算法	0.876	0.899	0.912	0.927	0.942	0.948	0.953	0.922
DPAFPN（本文算法）	0.827	0.878	0.903	0.914	0.936	0.947	0.951	0.908
DPAFPN+FPR（本文算法）	0.846	0.911	0.924	0.952	0.963	0.968	0.968	0.933

Tab. 2 Detection results of nodules with different sizes

算法	结节检出率/%			敏感度
算法	3~5 mm	5~10 mm	>10 mm	敏感度
DPAFPN	95.6	96.4	95.7	0.960
DPAFPN+FPR	97.0	97.1	98.2	0.974

Fig. 9 Visualization results of nodule detection of different sizes

Tab. 3 Comparison of parameters of loss functions

$γ$	$α$	敏感度
1	0.20	0.961
	0.25	0.972
	0.30	0.960
2	0.20	0.970
	0.25	0.974
	0.30	0.968

Tab. 3 Comparison of parameters of loss functions

$γ$	$α$	敏感度
1	0.20	0.961
	0.25	0.972
	0.30	0.960
2	0.20	0.970
	0.25	0.974
	0.30	0.968

Fig. 10 FROC curves of different algorithms

Tab. 4 GAM submodule verification

算法	平均扫描假阳性数（最佳性能）	最高敏感度
D_FPN	27	0.932
D_FPN+SA	29	0.954
D_FPN+CA	35	0.961
DPAFPN	33	0.967
DPAFPN+FPR	28	0.974

References 22

1	董林佳，强彦，赵涓涓，等.基于三维形状指数的肺结节自动检测方法［J］.计算机应用， 2017， 37（11）： 3182-3187.
	DONG L J， QIANG Y， ZHAO J J， et al. Automatic detection of pulmonary nodules based on 3D shape index［J］. Journal of Computer Applications， 2017， 37（11）： 3182-3187.
2	OKUMURA T， MIWA T， KAKO J I， et al. Automatic detection of lung cancers in chest CT images by variable N-Quoit filter ［C］// Proceedings of the 14th International Conference on Pattern Recognition — Volume 2. Piscataway： IEEE， 1998： 1671-1673.
3	WU S Y， WANG J F. Pulmonary nodules 3D detection on serial CT scans ［C］// Proceedings of the 3rd Global Congress on Intelligent Systems. Piscataway： IEEE， 2012： 257-260. 10.1109/gcis.2012.46
4	LASSEN B C， JACOBS C， KUHNIGK J M， et al. Robust semi-automatic segmentation of pulmonary subsolid nodules in chest computed tomography scans［J］. Physics in Medicine and Biology， 2015， 60（3）： No.1307. 10.1088/0031-9155/60/3/1307
5	高智勇，黄金镇，杜程刚.基于特征金字塔网络的肺结节检测［J］.计算机应用， 2020， 40（9）： 2571-2576.
	GAO Z Y， HUANG J Z， DU C G. Pulmonary nodule detection based on feature pyramid networks［J］. Journal of Computer Applications， 2020， 40（9）： 2571-2576.
6	XIE H T， YANG D B， SUN N N， et al. Automated pulmonary nodule detection in CT images using deep convolutional neural networks［J］. Pattern Recognition， 2019， 85： 109-119. 10.1016/j.patcog.2018.07.031
7	戚永军，顾军华，张亚娟，等.基于深度混合卷积模型的肺结节检测方法［J］.计算机应用， 2020， 40（10）： 2904-2909.
	QI Y J， GU J H， ZHANG Y J， et al. Deep mixed convolution model for pulmonary nodule detection［J］. Journal of Computer Applications， 2020， 40（10）： 2904-2909.
8	ZHENG S Y， GUO J P， CUI X N， et al. Automatic pulmonary nodule detection in CT scans using convolutional neural networks based on maximum intensity projection［J］. IEEE Transactions on Medical Imaging， 2020， 39（3）： 797-805. 10.1109/tmi.2019.2935553
9	ZHU W T， LIU C C， FAN W， et al. DeepLung： deep 3D dual path nets for automated pulmonary nodule detection and classification ［C］// Proceedings of the 2018 IEEE Winter Conference on Applications of Computer Vision. Piscataway： IEEE， 2018： 673-681. 10.1109/wacv.2018.00079
10	LI J Q， WANG K， YANG D， et al. Deepnodule： multi-task learning of segmentation bootstrap for pulmonary nodule detection ［C］// Proceedings of the 2021 IEEE International Conference on Acoustics， Speech and Signal Processing. Piscataway： IEEE， 2021： 1215-1219. 10.1109/icassp39728.2021.9413825
11	LI Y M， FAN Y. DeepSEED： 3D squeeze-and-excitation encoder-decoder convolutional neural networks for pulmonary nodule detection ［C］// Proceedings of the IEEE 17th International Symposium on Biomedical Imaging. Piscataway： IEEE， 2020： 1866-1869. 10.1109/isbi45749.2020.9098317
12	LIU J N， LI J， XUE F Y， et al. Dense attention module for accurate pulmonary nodule detection ［C］// Proceedings of the 2021 IEEE International Conference on Acoustics， Speech and Signal Processing. Piscataway： IEEE， 2021： 1220-1224. 10.1109/icassp39728.2021.9413936
13	袁金丽，赵琳琳，郭志涛，等.改进U型残差网络用于肺结节检测［J］.计算机工程与应用， 2022， 58（13）： 195-203. 10.3778/j.issn.1002-8331.2011-0211
	YUAN J L， ZHAO L L， GUO Z T， et al. Improved U-shaped residual network for lung nodule detection［J］. Computer Engineering and Applications， 2022， 58（13）： 195-203. 10.3778/j.issn.1002-8331.2011-0211
14	许正玺，张少敏，支力佳，等.三维多尺度嵌套U结构CT影像肺结节检测［J］.中国图象图形学报， 2022， 27（3）： 797-811. 10.11834/jig.210422
	XU Z X， ZHANG S M， ZHI L J， et al. Detection of pulmonary nodules in three-dimensional multiscal nested U-structure computed tomography images［J］. Journal of Image and Graphics， 2022， 27（3）： 797-811. 10.11834/jig.210422
15	张福玲，张少敏，支力佳，等.融合注意力机制和特征金字塔网络的CT图像肺结节检测［J］.中国图象图形学报， 2021， 26（9）： 2156-2170. 10.11834/jig.210160
	ZHANG F L， ZHANG S M， ZHI L J， et al. Detection of pulmonary nodules in CT images by combining an attention mechanism and a feature pyramid network［J］. Journal of Image and Graphics， 2021， 26（9）： 2156-2170. 10.11834/jig.210160
16	LIN T Y， DOLLÁR P， GIRSHICK R， et al. Feature pyramid networks for object detection ［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 936-944. 10.1109/cvpr.2017.106
17	CHEN Y P， LI J N， XIAO H X， et al. Dual path networks ［C］// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook， NY： Curran Associates Inc.， 2017： 4470-4478.
18	LIAO F Z， LIANG M， LI Z， et al. Evaluate the malignancy of pulmonary nodules using the 3-D deep leaky noisy-OR network［J］. IEEE Transactions on Neural Networks and Learning Systems， 2019， 30（11）： 3484-3495. 10.1109/tnnls.2019.2892409
19	DOU Q， CHEN H， JIN Y M， et al. Automated pulmonary nodule detection via 3D ConvNets with online sample filtering and hybrid-loss residual learning ［C］// Proceedings of the 2017 International Conference on Medical Image Computing and Computer-Assisted Intervention， LNCS 10435. Cham： Springer， 2017： 630-638.
20	MEI J， CHENG M M， XU G， et al. SANet： a slice-aware network for pulmonary nodule detection［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2022， 44（8）： 4374-4387.
21	WANG B， QI G J， TANG S， et al. Automated pulmonary nodule detection： high sensitivity with few candidates ［C］// Proceedings of the 2018 International Conference on Medical Image Computing and Computer-Assisted Intervention， LNCS 11071. Cham： Springer， 2018： 759-767.
22	DING J， LI A X， HU Z Q， et al. Accurate pulmonary nodule detection in computed tomography images using deep convolutional neural networks ［C］// Proceedings of the 2017 International Conference on Medical Image Computing and Computer-Assisted Intervention， LNCS 10435. Cham： Springer， 2017： 559-567.

[1]	Jing QIN, Zhiguang QIN, Fali LI, Yueheng PENG. Diagnosis of major depressive disorder based on probabilistic sparse self-attention neural network [J]. Journal of Computer Applications, 2024, 44(9): 2970-2974.
[2]	Liting LI, Bei HUA, Ruozhou HE, Kuang XU. Multivariate time series prediction model based on decoupled attention mechanism [J]. Journal of Computer Applications, 2024, 44(9): 2732-2738.
[3]	Yun LI, Fuyou WANG, Peiguang JING, Su WANG, Ao XIAO. Uncertainty-based frame associated short video event detection method [J]. Journal of Computer Applications, 2024, 44(9): 2903-2910.
[4]	Zhiqiang ZHAO, Peihong MA, Xinhong HEI. Crowd counting method based on dual attention mechanism [J]. Journal of Computer Applications, 2024, 44(9): 2886-2892.
[5]	Hong CHEN, Bing QI, Haibo JIN, Cong WU, Li’ang ZHANG. Class-imbalanced traffic abnormal detection based on 1D-CNN and BiGRU [J]. Journal of Computer Applications, 2024, 44(8): 2493-2499.
[6]	Kaipeng XUE, Tao XU, Chunjie LIAO. Multimodal sentiment analysis network with self-supervision and multi-layer cross attention [J]. Journal of Computer Applications, 2024, 44(8): 2387-2392.
[7]	Pengqi GAO, Heming HUANG, Yonghong FAN. Fusion of coordinate and multi-head attention mechanisms for interactive speech emotion recognition [J]. Journal of Computer Applications, 2024, 44(8): 2400-2406.
[8]	Zhonghua LI, Yunqi BAI, Xuejin WANG, Leilei HUANG, Chujun LIN, Shiyu LIAO. Low illumination face detection based on image enhancement [J]. Journal of Computer Applications, 2024, 44(8): 2588-2594.
[9]	Shangbin MO, Wenjun WANG, Ling DONG, Shengxiang GAO, Zhengtao YU. Single-channel speech enhancement based on multi-channel information aggregation and collaborative decoding [J]. Journal of Computer Applications, 2024, 44(8): 2611-2617.
[10]	Yangyi GAO, Tao LEI, Xiaogang DU, Suiyong LI, Yingbo WANG, Chongdan MIN. Crowd counting and locating method based on pixel distance map and four-dimensional dynamic convolutional network [J]. Journal of Computer Applications, 2024, 44(7): 2233-2242.
[11]	Li LIU, Haijin HOU, Anhong WANG, Tao ZHANG. Generative data hiding algorithm based on multi-scale attention [J]. Journal of Computer Applications, 2024, 44(7): 2102-2109.
[12]	Song XU, Wenbo ZHANG, Yifan WANG. Lightweight video salient object detection network based on spatiotemporal information [J]. Journal of Computer Applications, 2024, 44(7): 2192-2199.
[13]	Dahai LI, Zhonghua WANG, Zhendong WANG. Dual-branch low-light image enhancement network combining spatial and frequency domain information [J]. Journal of Computer Applications, 2024, 44(7): 2175-2182.
[14]	Wenliang WEI, Yangping WANG, Biao YUE, Anzheng WANG, Zhe ZHANG. Deep learning model for infrared and visible image fusion based on illumination weight allocation and attention [J]. Journal of Computer Applications, 2024, 44(7): 2183-2191.
[15]	Wu XIONG, Congjun CAO, Xuefang SONG, Yunlong SHAO, Xusheng WANG. Handwriting identification method based on multi-scale mixed domain attention mechanism [J]. Journal of Computer Applications, 2024, 44(7): 2225-2232.