Lung Nodules Detection Based on Modified Extreme Learning
Machine with Deep Convolutional Features
Guodong Zhang
a,c
, Yuxuan Sun
a
, Lingchuang Kong
a
, Jing Bi
a
, Zhaoxuan Gong
a
, Yoohwan Kim
c
,
Wei Guo
a,b,*
a
School of Computer Science, Shenyang Aerospace University, Shenyang, Liaoning China 110136;
b
Shenyang Institute of Computing Technology, Chinese Academy of Sciences, Shenyang, Liaoning
China 110168;
c
Department of Computer Science, University of Nevada Las Vegas, Las Vegas, NV
USA 89154
ABSTRACT
This work achieves a method based on modified extreme learning machine (ELM) with deep convolutional features to
detect lung nodules automatically. Convolutional neural networks (CNNs) are employed to extract the features of lung
nodules for classification. And then ELM is used to detect the lung nodules by combining the normalization and vote
selection. In comparison with the traditional methods, it is shown that our method achieves a higher performance and it
can be used as an effective tool for lung nodules computer aided diagnosis.
Keywords: Extreme learning machine, convolutional neural network, detection of lung nodules, computer aided
diagnosis
1. INTRODUCTION
Lung cancer is a major disease that endangers the health of human body. Most of the early stages of lung cancer are
manifested in the form of nodules. As such, the detection of lung nodules has become one of the most important methods
for diagnosing lung cancers. However, the lung nodules are not easy to be found. A large number of CT images bring a
big challenge to doctors in manual analysis. To ease the situation, computer aided diagnosis (CAD) can be used for
detection of lung nodules. It can automatically or semi-automatically detect the lung nodules and provide reference for
doctors to reduce the workload of doctors and human errors.
At present, the classification method of medical image including support vector machine (SVM), AdaBoost, Naive
Bayes (NB) and other methods. Zhang et al.
1
used the approach based on SVM to distinguish whether the region of
interest contains nodules. Sampaio et al.
2
used the combination of convolutional neural networks (CNNs) and SVM to
get the bossing. Dolejsi et al.
3
used AdaBoost and SVM respectively in the classification of lung nodules, where the
experimental results show that the classification accuracy of the two methods has little difference. Krishnaiah et al.
4
used
the Naive Bayes classification technique to diagnose lung cancer, and achieved good classification results. Kumar et al.
5
achieved lung nodule classification using deep features in Computed Tomography (CT) images. However, these methods
have the problems of low classification accuracy, long learning time and complex parameter setting.
Compared with the traditional image classification methods, Extreme Learning Machine (ELM)
6
is a new fast learning
algorithm. Daliri
7
detected breast cancer by combining ELM with SVM, and achieved the good classification results.
Pangaribuan
8
used ELM for computer aided diagnosis of diabetes. The experiment shows that ELM has faster learning
speed and higher recognition accuracy. However, the input weights and bias of ELM are chosen randomly, so the
performance of the network is not stable. We develop a kind of lung nodule detection method based on modified ELM to
stabilize the performance of the network. The experimental results show that our method achieves a higher performance
than traditional lung nodules detection methods.
2. METHODS
At present, the popular detection algorithm for lung nodule is mainly used to manually extract the features and then use
the classifier to classify. We use CNNs to extract the features of lung nodules. It avoids the complex feature extraction
*zhanggd@sau.edu.cn and guowei@sau.edu.cn; phone (86)24-8972-3471