Expert Systems With Applications 86 (2017) 135–144
Contents lists available at ScienceDirect
Expert Systems With Applications
journal homepage: www.elsevier.com/locate/eswa
Microcalcification diagnosis in digital mammography using extreme
learning machine based on hidden Markov tree model of dual-tree
complex wavelet transform
Kai Hu
a , b
, Wei Yang
a , b
, Xieping Gao
a , b , ∗
a
The MOE Key Laboratory of Intelligent Computing & Information Processing, Xiangtan University, Xiangtan 41110 5 , China
b
College of Information Engineering, Xiangtan University, Xiangtan 411105 , China
a r t i c l e i n f o
Article history:
Received 1 December 2016
Revised 5 May 2017
Accepted 26 May 2017
Available online 26 May 2017
Keywords:
Microcalcification diagnosis
Digital mammography
Dual-tree complex wavelet transform
Hidden Markov tree model
Extreme learning machine
Feature extraction
a b s t r a c t
Diagnosis of benign and malignant microcalcifications in digital mammography using Computer-aided Di-
agnosis (CAD) system is critical for the early diagnosis of breast cancer. Wavelet transform based diagno-
sis methods are effective to accomplish this task, but limited by representing the correlation within each
wavelet scale, these methods neglect the correlation between wavelet scales. In this paper, we apply the
hidden Markov tree model of dual-tree complex wavelet transform (DTCWT-HMT) for microcalcification
diagnosis in digital mammography. DTCWT-HMT can effectively capture the correlation between different
wavelet coefficients and model the statistical dependencies and non-Gaussian statistics of real signals, is
used to characterize microcalcifications for the diagnosis of benign and malignant cases. The combined
features which consist of the DTCWT-HMT features and the DTCWT features are optimized by genetic
algorithm (GA). Extreme learning machine (ELM), an efficient learning theory is employed as the clas-
sifier to diagnose the benign and malignant microcalcifications. The validity of the proposed method is
evaluated on the Nijmegen, MIAS and DDSM datasets using area under curve (AUC) of receiver operating
characteristic (ROC). The AUC values of 0.9856, 0.9941 and 0.9168 of the proposed method are achieved
on Nijmegen, MIAS and DDSM, respectively. We compare the proposed method with state-of-the-art di-
agnosis methods, and the experimental results show the effectiveness of the proposed method for the
diagnosis of the benign and malignant microcalcifications in mammograms in terms of the accuracy and
stability.
©2017 Elsevier Ltd. All rights reserved.
1.
Introduction
Reported by the American Cancer Society (2008) , breast cancer
is the second most common cancer in women all over the world.
Early detection and diagnosis is essential for decreasing the death
rate caused by breast cancer. Recently, digital mammography be-
comes one of the most effective techniques for early detection and
diagnosis of breast cancer ( Alayhoglu, & Aghdasi, 1999; Ganesan
et al., 2013; Lee, & Chen, 2015 ). Microcalcification is as one of the
important signs of breast cancer. The earlier the microcalcifications
are detected, the greater the chance of cure can be provided. Fig. 1
shows the examples of benign and malignant cases of microcalci-
fications in mammograms. In clinical practice, correct detection of
microcalcifications is very difficult ( Ciecholewski, 2017 ). Especially
∗
Corresponding author at: College of Information Engineering, Xiangtan Univer-
sity, Xiangtan 411105 , China.
E-mail addresses: kaihu@xtu.edu.cn (K. Hu), 16043395@qq.com (W. Yang),
xpgao@xtu.edu.cn , xpgao448@163.com (X. Gao).
it is difficult and time consuming for breast radiologists to distin-
guish malignant from benign microcalcifications ( Chen et al., 2015 ).
In order to improve the diagnostic accuracy and make the diagnos-
tic process easier for breast radiologists, computer-aided diagnosis
(CADx) is developed as the “second reader” which can provide use-
ful information for the diagnosis of microcalcifications in mammo-
grams.
In the past decades, there are a series of breast cancer CADx
algorithms have been developed in the literature ( Beura, Majhi, &
Dash, 2015; Buciu, & Gacsadi, 2011; Chen et al., 2012, 2015; Choi,
Kim, Plataniotis, & Ro, 2016; Ciecholewski, 2017; Crouse, Nowak, &
Baraniuk, 1998; Dhawan et al., 1996; Diaz-Huerta, Felipe-Riveron,
& Montaño-Zetina L., 2014; Görgel, Sertba ¸s , Kilic, Ucan, & Osman,
2009; Görgel, Sertbas, & Ucan, 2013; Hamid, Farshid, & Siamak,
2004; Hu, Gao, & Li, 2011; Jen, & Yu, 2015; Jiang, Zhang, & Li, 2015;
Krishnan et al., 2010; Mousa, Munib, & Moussa, 2005; Nascimento
et al., 2013; Orchard, & Ramchandran, 1994; Rao, & Subramanyam,
2008; Strange, et al., 2014; Wajid, & Hussain, 2015; Zhang, & Gao,
2007 ). Overall reviewing the literature, wavelet transform has been
http://dx.doi.org/10.1016/j.eswa.2017.05.062
0957-4174/© 2017 Elsevier Ltd. All rights reserved.