Access the full text.
Sign up today, get DeepDyve free for 14 days.
Tao Sun, Jingjing Wang, Xia Li, Pingxin Lv, Fen Liu, Yanxia Luo, Q. Gao, Huiping Zhu, Xiuhua Guo (2013)
Comparative evaluation of support vector machines for computer aided diagnosis of lung cancer in CT based on a multi-dimensional data setComputer methods and programs in biomedicine, 111 2
K. Kanazawa, Y. Kawata, N. Niki, H. Satoh, H. Ohmatsu, R. Kakinuma (1998)
Computer-aided diagnosis for pulmonary nodules based on helical CT imagesProceedings. Fourteenth International Conference on Pattern Recognition (Cat. No.98EX170), 2
Temesguen Messay, R. Hardie, S. Rogers (2010)
A new computationally efficient CAD system for pulmonary nodule detection in CT imageryMedical image analysis, 14 3
D. Specht (1990)
Probabilistic neural networksNeural Networks, 3
Wookjin Choi, T. Choi (2012)
Genetic programming-based feature transform and classification for the automatic detection of pulmonary nodules on computed tomography imagesInf. Sci., 212
H. Yazid, H. Yazid, M. Harun, S. Mohd, A. Mohamed, M. Rizon, S. Sayuti (2007)
Circular discontinuities detection in welded joints using Circular Hough TransformNdt & E International, 40
J. Erasmus, J. Connolly, H. McAdams, V. Roggli (2000)
Solitary pulmonary nodules: Part I. Morphologic evaluation for differentiation of benign and malignant lesions.Radiographics : a review publication of the Radiological Society of North America, Inc, 20 1
Muzzamil Javaid, Moazzam Javid, M. Rehman, Syed Shah (2016)
A novel approach to CAD system for the detection of lung nodules in CT imagesComputer methods and programs in biomedicine, 135
Zhiqiong Wang, Junchang Xin, Peishun Sun, Zhixiang Lin, Yudong Yao, Xiaosong Gao (2018)
Improved lung nodule diagnosis accuracy using lung CT images with uncertain classComputer methods and programs in biomedicine, 162
Handan Çamdevýren, Nilsun Demýr, A. Kanık, Sýddýk Keskýn (2005)
Use of principal component scores in multiple linear regression models for prediction of Chlorophyll-a in reservoirsEcological Modelling, 181
J. Dehmeshki, Xujiong Ye, Xinyu Lin, M. Casique, H. Amin (2007)
Automated detection of lung nodules in CT images using shape-based genetic algorithmComputerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society, 31 6
E. Dandıl, M. Çakiroglu, Ziya Ekşi, Murat Ozkan, O. Kurt, A. Canan (2014)
Artificial neural network-based classification system for lung nodules on computed tomography scans2014 6th International Conference of Soft Computing and Pattern Recognition (SoCPaR)
M. Giger, K. Bae, H. MacMahon (1994)
Computerized Detection of Pulmonary Nodules in Computed Tomography ImagesInvestigative Radiology, 29
N. Dubrawsky (1989)
Cancer statisticsCA: A Cancer Journal for Clinicians, 39
Hao Han, Lihong Li, Fangfang Han, Bowen Song, W. Moore, Zhengrong Liang (2015)
Fast and Adaptive Detection of Pulmonary Nodules in Thoracic CT Images Using a Hierarchical Vector Quantization SchemeIEEE Journal of Biomedical and Health Informatics, 19
M. Rehman, Muzzamil Javaid, Syed Shah, S. Gilani, Mohsin Jamil, S. Butt (2018)
An appraisal of nodules detection techniques for lung cancer in CT imagesBiomed. Signal Process. Control., 41
C. Hollitt (2013)
A convolution approach to the circle Hough transform for arbitrary radiusMachine Vision and Applications, 24
R. Siegel, D. Naishadham, A. Jemal (2012)
Cancer statistics, 2012CA: A Cancer Journal for Clinicians, 62
T. Kohonen (2001)
Self-Organizing Maps of Massive DatabasesInternational journal of engineering intelligent systems for electrical engineering and communications, 9
L. Dice (1945)
Measures of the Amount of Ecologic Association Between SpeciesEcology, 26
K. Murphy, B. Ginneken, A. Schilham, B. Hoop, H. Gietema, M. Prokop (2009)
A large-scale evaluation of automatic pulmonary nodule detection in chest CT using local image features and k-nearest-neighbour classificationMedical image analysis, 13 5
Dezso Csipo, R. Dayhoff, P. Kuzmak (2001)
Integrating Digital Imaging and Communications In Medicine (DICOM)-structured reporting into the hospital environmentJournal of Digital Imaging, 14
B. Sahiner, Lubomir Hadjiiski, H. Chan, Jiazheng Shi, P. Cascade, E. Kazerooni, Chuan Zhou, Jun Wei, A. Chughtai, C. Poopat, Thomas Song, Jadranka Nojkova, L. Frank, A. Attili (2007)
Effect of CAD on radiologists' detection of lung nodules on thoracic CT scans: observer performance study, 6515
A. Schilham, B. Ginneken, M. Loog (2006)
A computer-aided diagnosis system for detection of lung nodules in chest radiographs with an evaluation on a public databaseMedical image analysis, 10 2
Kenji Suzuki, S. Armato, Feng Li, S. Sone, K. Doi (2003)
Massive training artificial neural network (MTANN) for reduction of false positives in computerized detection of lung nodules in low-dose computed tomography.Medical physics, 30 7
Jinsa Kuruvilla, K. Gunavathi (2014)
Lung cancer classification using neural networks for CT imagesComputer methods and programs in biomedicine, 113 1
D. Chi (2011)
Self-Organizing Map-Based Color Image Segmentation with k-Means Clustering and Saliency MapInternational Scholarly Research Notices, 2011
良治 祐延 (2003)
DICOM Digital Imaging and Communications in Medicine, ダイコム, 1
L. Ries, D. Harkins, Martin Krapcho, A. Mariotto, B. Miller, E. Feuer, L. Clegg, M. Eisner, Marie-Josèphe Horner, N. Howlader, Matthew Hayat, B. Hankey, B. Edwards (2006)
SEER Cancer Statistics Review, 1975-2003
A. Reeves, A. Biancardi, T. Apanasovich, C. Meyer, H. MacMahon, E. Beek, E. Kazerooni, D. Yankelevitz, M. McNitt-Gray, G. McLennan, S. Armato, C. Henschke, D. Aberle, B. Croft, L. Clarke (2007)
The Lung Image Database Consortium (LIDC): a comparison of different size metrics for pulmonary nodule measurements.Academic radiology, 14 12
(2012)
“Review on feature extraction methods in pattern classification,”
R. Duda, P. Hart (1972)
Use of the Hough transformation to detect lines and curves in picturesCommun. ACM, 15
K. Kanazawa, Y. Kawata, N. Niki (1998)
Computer-aided diagnosis for pulmonary nodules based on helical CT imagesComputerized Medical Imaging and Graphics, 22
G. Rubin, J. Lyo, D. Paik, A. Sherbondy, L. Chow, A. Leung, R. Mindelzun, Pamela Schraedley-Desmond, S. Zinck, D. Naidich, S. Napel (2005)
Pulmonary nodules on multi-detector row CT scans: performance comparison of radiologists and computer-aided detection.Radiology, 234 1
Ming Yang, K. Kidiyo, R. Joseph (2008)
A Survey of Shape Feature Extraction Techniques
S. Armato, G. McLennan, M. McNitt-Gray, C. Meyer, D. Yankelevitz, D. Aberle, C. Henschke, E. Hoffman, E. Kazerooni, H. MacMahon, A. Reeves, B. Croft, L. Clarke (2004)
Lung image database consortium: developing a resource for the medical imaging research community.Radiology, 232 3
S. Park, Jun Tan, Xingwei Wang, D. Lederman, J. Leader, Soohyung Kim, B. Zheng (2011)
Computer-aided detection of early interstitial lung diseases using low-dose CT imagesPhysics in Medicine & Biology, 56
Sheeraz Akram, M. Javed, Ayyaz Hussain, F. Riaz, M. Akram (2015)
Intensity-based statistical features for classification of lungs CT scan nodules using artificial intelligence techniquesJournal of Experimental & Theoretical Artificial Intelligence, 27
Kenji Suzuki, Feng Li, S. Sone, K. Doi (2005)
Computer-aided diagnostic scheme for distinction between benign and malignant nodules in thoracic low-dose CT by use of massive training artificial neural networkIEEE Transactions on Medical Imaging, 24
J. Hasegawa, K. Mori, J. Toriwaki, H. Anno, K. Katada (1994)
Automated extraction of lung cancer lesions from multislice chest CT images by using three-dimensional image processingSystems and Computers in Japan, 25
G. Wouwer, P. Scheunders, D. Dyck (1999)
Statistical texture characterization from discrete wavelet representationsIEEE transactions on image processing : a publication of the IEEE Signal Processing Society, 8 4
R. Opfer, R. Wiemker (2007)
Performance analysis for computer-aided lung nodule detection on LIDC data, 6515
Liang-Hwa Chen, Shyang Chang (1995)
An adaptive learning algorithm for principal component analysisIEEE transactions on neural networks, 6 5
C. Jacobs, E. Rikxoort, T. Twellmann, E. Scholten, P. Jong, J. Kuhnigk, M. Oudkerk, H. Koning, M. Prokop, C. Schaefer-Prokop, B. Ginneken (2014)
Automatic detection of subsolid pulmonary nodules in thoracic computed tomography imagesMedical image analysis, 18 2
Xujiong Ye, Xinyu Lin, J. Dehmeshki, G. Slabaugh, Gareth Beddoe (2009)
Shape-Based Computer-Aided Detection of Lung Nodules in Thoracic CT ImagesIEEE Transactions on Biomedical Engineering, 56
David Clausi (2002)
An analysis of co-occurrence texture statistics as a function of grey level quantizationCanadian Journal of Remote Sensing, 28
M. McNitt-Gray, S. Armato, C. Meyer, A. Reeves, G. McLennan, Richard Pais, J. Freymann, Matthew Brown, R. Engelmann, P. Bland, Gary Laderach, C. Piker, Junfeng Guo, Zaid Towfic, David Qing, D. Yankelevitz, D. Aberle, E. Beek, H. MacMahon, E. Kazerooni, B. Croft, L. Clarke (2007)
The Lung Image Database Consortium (LIDC) data collection process for nodule detection and annotation, 6514
(1999)
Neural Networks, A Comprehensive Foundation, Hamilton
Hongtao Xie, Dongbao Yang, Nannan Sun, Zhineng Chen, Yongdong Zhang (2019)
Automated pulmonary nodule detection in CT images using deep convolutional neural networksPattern Recognit., 85
Jorge Suárez-Cuenca, P. Tahoces, M. Souto, M. Lado, M. Rémy-Jardin, J. Remy, J. Vidal (2009)
Application of the iris filter for automatic detection of pulmonary nodules on computed tomography imagesComputers in biology and medicine, 39 10
Wookjin Choi, T. Choi (2013)
Automated Pulmonary Nodule Detection System in Computed Tomography Images: A Hierarchical Block Classification ApproachEntropy, 15
R. Haralick, K. Shanmugam, I. Dinstein (1973)
Textural Features for Image ClassificationIEEE Trans. Syst. Man Cybern., 3
Y. Jeong, C. Yi, K. Lee (2007)
Solitary pulmonary nodules: detection, characterization, and guidance for further diagnostic workup and treatment.AJR. American journal of roentgenology, 188 1
M. Mcnitt-Gray, S. Armato, C. Meyer (2007)
The Lung Image Database Consortium (LIDC) data collection process for nodule detection and annotationAcademic Radiology, 14
(2001)
Engineering Intelligent Systems for Electrical Engineering and Communications
Serhat Ozekes, A. Çamurcu (2006)
Automatic Lung Nodule Detection Using Template Matching
P. Jaccard (1912)
THE DISTRIBUTION OF THE FLORA IN THE ALPINE ZONE.1New Phytologist, 11
K. Mao, K. Tan, W. Ser (2000)
Probabilistic neural-network structure determination for pattern classificationIEEE transactions on neural networks, 11 4
Hindawi Journal of Healthcare Engineering Volume 2018, Article ID 9409267, 12 pages https://doi.org/10.1155/2018/9409267 Research Article A Computer-Aided Pipeline for Automatic Lung Cancer Classification on Computed Tomography Scans Emre Dandıl Department of Computer Engineering, Faculty of Engineering, Bilecik Seyh Edebali University, Gulumbe Campus, 11210 Bilecik, Turkey Correspondence should be addressed to Emre Dandıl; emre.dandil@bilecik.edu.tr Received 28 June 2018; Revised 24 September 2018; Accepted 8 October 2018; Published 1 November 2018 Guest Editor: Cesare Valenti Copyright © 2018 Emre Dandıl. *is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Lung cancer is one of the most common cancer types. For the survival of the patient, early detection of lung cancer with the best treatment method is crucial. In this study, we propose a novel computer-aided pipeline on computed tomography (CT) scans for early diagnosis of lung cancer thanks to the classification of benign and malignant nodules. *e proposed pipeline is composed of four stages. In preprocessing steps, CT images are enhanced, and lung volumes are extracted from the image with the help of a novel method called lung volume extraction method (LUVEM). *e significance of the proposed pipeline is using LUVEM for extracting lung region. In nodule detection stage, candidate nodules are determined according to the circular Hough transform- (CHT-) based method. *en, lung nodules are segmented with self-organizing maps (SOM). In feature computation stage, intensity, shape, texture, energy, and combined features are used for feature extraction, and principal component analysis (PCA) is used for feature reduction step. In the final stage, probabilistic neural network (PNN) classifies benign and malign nodules. According to the experiments performed on our dataset, the proposed pipeline system can classify benign and malign nodules with 95.91% accuracy, 97.42% sensitivity, and 94.24% specificity. Even in cases of small-sized nodules (3–10 mm), the proposed system can determine the nodule type with 94.68% accuracy. Computer-aided detection (CAD) systems have been an 1. Introduction important field in medical image processing. CAD systems Nowadays, lung cancer is one of the ranking first causes of also based on machine learning methods designed to di- mortality worldwide among men and women [1, 2]. Al- agnosis of cancer have become common in recent years. though there are a lot of treatment options like surgery, Radiologists and physicians may use findings of CAD sys- radiotherapy, and chemotherapy, five year survival rate for tems as the second opinion before making their own final patients is quite low [3]. However, survival rate may go up to decisions. *erefore, CAD systems play an important role in 54% in case lung cancer is identified in an early stage [4]. CT scans to help radiologists for detection of lung cancer efficiently. *erefore, early detection of lung cancer is vital to decrease lung cancer mortality. Medical imaging techniques have been important 2. Related Work technology in screening of lung cancer recently. CT scan becomes a standard modality for detecting and assessing Computer-aided detection (CAD) systems have been active lung cancer [5]. Most of the lung nodules are usually benign. research field for the pulmonary nodule detection and However, some nodules such as calcified, swollen, and hard malign/benign nodule classification. Until now, many CAD can also be determined as benign. Similarly, a hard nodule systems have been proposed. For example, Ozekes and generally is cancerous (malignant), but it may be considered Camurcu proposed a method for pulmonary nodule de- as benign case in some cases [6]. Furthermore, medical CT tection method using template matching [7]. Schilham images are needed to be diagnosed by radiologists. et al. presented a CAD system which consists of image 2 Journal of Healthcare Engineering preprocessing, candidate nodule detection, feature ex- (Digital Imaging and Communications in Medicine) files traction, and classification for nodule detection in chest [22]. radiographs [8]. Dehmeshki et al. detected lung nodules *e goal of image enhancement step is to prevent using shape-based genetic algorithm template matching misleading results that may occur in subsequent processes. [9]. Suarez-Cuenca et al. also designed a system which *us, we firstly implemented the median filter to remove discriminates the nodules and non-nodule cases using iris unnecessary noises and enhance the images. Moreover, the filter in CT images [10]. Murphy et al. automatically per- sharpening of nodule contours is an important step for the formed lung nodule detection using k-nearest neighbours detection of nodules. Laplacian filter was used in our study. classifier [11]. Giger et al. realized CAD system to detect So, nodules on lung region were able to be detected more lung nodules on CT images using geometric features [12]. accurately. Furthermore, histogram equalization was also In addition, Hasegawa et al. proposed image processing used in enhancement step in order to minimize contrast methods for identification of lung nodules using CT scans differences which occur due to scanning errors and to [13]. In another study, Kanazawa et al. used a CAD system remove unnecessary grains. to identify pulmonary nodules with fuzzy features [14]. In In lung volume extraction step of image preprocessing 2005, Suzuki et al. proposed a method using ANN for stage, extracting of the lung region from CT image is classification of malignant and benign nodules on CT performed. *ere are many methods for extracting lung images [15]. Sun et al. compared support vector machines volume from a lung CT scan [9, 10, 23–25]. However, these (SVM) with the some classification methods for detection methods are complex, and they require more processing of lung cancer on CT images [16]. Kuruvilla and Gunavathi overhead. In some cases, these methods may lead to losses proposed a system using ANNs for classification of lung of information about lung regions or cause noise. *e cancer [17]. purpose of this step is to extract the lung region completely In a recent study on lung nodule detection, Javaid et al. from the full lung CT image. *erefore, a simple but ef- proposed a computer aided nodule detection method for the fective and novel method has been proposed in this study segmentation and detection of challenging in different type for lung volume extraction named as lung volume ex- traction method (LUVEM). *e pseudocode of LUVEM is nodules [18]. ur Rehman et al. presented a systematic analysis of nodules detection techniques with the current shown in Algorithm 1. trends and future challenges [19]. Wang et al. proposed In LUVEM method, lung lobes are extracted from CT a pulmonary nodule CAD based on semisupervised extreme images with the help of morphological operations. LUVEM learning machine [20]. Xie et al. proposed an automated removes unrelated segments on the sides and edges of the pulmonary nodule detection system with 2D convolutional preprocessed image and obtain the lung region successful. In neural network (CNN) on LUNA16 dataset [21]. the algorithm, input image is firstly converted to double- In this study, we have proposed fully automated formatted image. Afterwards, 1 or 0 values are assigned to computer-aided pipeline for the detection of pulmonary each pixel of double-formatted image according to low and nodules and classification of benign/malign nodules in early high threshold values. *e low and high threshold values are stage. *e contributions of this paper are (1) to review the determined 0.25 and 0.65 in this algorithm, respectively. *e systematic literature review; (2) to present the state of the art method removes the bright areas on the edges of the lung CT detection of pulmonary nodules and classification of lung image since their average values change between low and cancer; (3) to propose the novel preprocessing method high values. After this process, the image is converted to (LUVEM) for the lung volume extraction; (4) to suggest binary format and performed morphological operations a novel candidate lung nodule detection method using CHT; which are eroding, dilating, and filling, respectively. Finally, (5) to design a holistic pipeline for the detection of pul- the image is again converted to gray-scale format. *e monary nodules as well as lung cancer; (6) the detailed segmentation examples of LUVEM can be seen in Figure 2. It comparison of feature extraction methods for lung nodule is clearly seen that LUVEM can successfully extract the lung detection; and (7) to perform the detailed performance volume. In addition, quantitative evaluation of LUVEM will evaluation, high true detection rate, and low false detection be reported below. rate for nodule detection and classification. 3.2. Lung Nodule Detection. *e first step of lung nodule 3. Architecture of the Computer-Aided Pipeline detection is candidate nodule detection. *e nodule can- didates in volume should be detected before nodule seg- Designed pipeline consists of four main stages such as image mentation. *e lung volume includes vessels and nodules. preprocessing (Stage I), lung nodule detection (Stage II), Moreover, the density of nodules, vessels, and lungs is nodule feature computation (Stage III), and nodule classi- different from each other [26]. Since the lung nodules have fication (Stage IV). *e work flow of the pipeline is presented a circular and helical structure, they can be differentiated by in Figure 1. means of circularity determination. Many methods have been suggested for identifying the round objects. Circular Hough transform (CHT), which proposed by Duda et al. 3.1. Lung Image Preprocessing. In the first step of the image [27], is one of the most successful method [28] for detection preprocessing stage, reading of CT images is performed. *e of round objects on the images. In this study, CHT CT scans obtained for the work are stored as DICOM Journal of Healthcare Engineering 3 Image Lung nodule Nodule feature Nodule preprocessing detection computation classification Image DICOM image 10, 20, 20, 32, 18, 33, 48.... Lung Malign Benign Nodule feature Image nodule nodule nodule extraction enhancement segmentation Candiate Lung volume Nodule feature Nodule nodule extraction reduction classification detection Stage I Stage II Stage III Stage IV Figure 1: Work flow of the designed pipeline to detect lung cancer. *e system consists of four stages: Stage I—enhancement of lung CT image and a novel lung volume segmentation method (LUVEM), Stage II—candidate nodule detection using CHTand segmentation of lung nodules using SOM, Stage III—computing of lung nodule features and reduction features using PCA, and Stage IV—classification of malign and benign lung nodule using PNN. (1) procedure LungVolumeExtraction(input image, low, high) //beginning lung vol. extraction algorithm (2) I ⟵ input image //input image (3) O ⟵ Ø, Lo ⟵ low, Hi ⟵ high //output image, low and high threshold (4) DIm ⟵ Converting (I, double) //converting image to double image (5) for (iϵ row length of DIm) //starting loop of extraction step (6) for (jϵ column length of DIm) //starting of inner loop (7) if DIm(i, j) ≥ Lo and DIm(i, j) ≤ Hi then //starting of if (8) DIm(i, j) ⟵ 1; //assignment of binary 1 (9) else (10) DIm(i, j) ⟵ 0; //assignment of binary 0 (11) end if //ending of if (12) end for //ending of inner loop (13) end for //ending loop of extraction step (14) BI ⟵ Converting(DIm, binary) //converting image to binary image (15) FI ⟵ Morphological eroding, dilating, filling (BI) //morphological operations (16) GIm ⟵ Converting(FI, gray-scale) //converting image to gray-scale image (17) O ⟵ GIm //return output image (18) end procedure //ending of LUVEM method ALGORITHM 1: *e pseudocode of lung volume extraction method (LUVEM). operations are used for candidate nodule detection. CHTcan large/complex datasets [31, 32]. Furthermore, it designs detect the round object in the image; moreover, it can also data maps that can be interpreted easily. In addition to detect the noncircular object by means of some operations. these advantages, SOM can easily segment very small *e image dataset is divided into 3 categories according to nodules on the lung CT images [3]. *e examples of seg- the nodule size such as<10 mm, 10–20 mm, and>20 mm. In mented lung nodule images using SOM are shown in Figure 4. order to detect the nodules in different size by CHT, three minimum and maximum radiuses such as 3–12 mm, 10– 20 mm, and 15–45 mm are determined. In Figure 3, it is shown that the examples of determination of candidate 3.3. Nodule Feature Computation. Generally, CAD systems nodules on CT images. segment lung nodules for the determination of nodule *e second step of lung nodule detection is nodule candidates, and then features extract from the candidate segmentation. In this study, SOM [29] is proposed to nodules. *e popular features are geometric feature, gray segment nodules on CT images. SOM is an unsupervised level features, gradient features, and energy level features. neural network learning [30] method. It can perform on *erefore, we extracted 2D significative features from lung 4 Journal of Healthcare Engineering Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7 (a) (b) Figure 2: Extraction examples of lung regions: (a) preprocessed images and (b) lung volume extraction using LUVEM. Figure 3: Determination of candidate lung nodule using CHT. CT images to discriminate benign or malign nodule. First-order statistical features (SSF) of an image are Firstly, we used shape-based features for analyzing nodule calculated from the gray level histogram values of an image geometry. We used first-order statistical features to obtain [33]. In this study, 6 basic features such as standard de- global statistic about nodule region. Moreover, we utilized viation, entropy, means, skewness, kurtosis, and variance gray level co-occurrence matrix (GLCM) texture features for were extracted by SSF using the histogram values of a gray gray level statistic of nodules. Finally, we extracted wavelet level lung CT image. Shape-based features (SBF) allow decomposition transform features to obtain the energy feature extraction from the image by using geometric pa- feature of nodules. All computed features are extracted from rameters [34]. Shape features give some information about the slice of the segmented object. an image such as sharpness, circularity, and convexity. In Journal of Healthcare Engineering 5 Sample 1 Sample 2 Sample 3 Sample 4 (a) (b) Figure 4: Examples of segmented lung nodules: (a) images of extracted lung volume and (b) segmentation of lung nodule using SOM. this study, a total of 16 shape features were extracted to Table 1: *e number of extracted 2D features from lung CT facilitate the determination of nodule type from CT lung images. images. Statistical features of a gray level image (GTF) of Feature extraction method Number of feature Order a texture are first derived with the help of GLCM texture SSF 6 0–6 features proposed by Haralick [35, 36]. GLCM method SBF 16 7–22 shows the relationship between pixels of different gray level GTF 22 ∗ 4 � 88 23–110 and is widely used in applications of medical image pro- TEF 13 111–123 cessing. In this study, a total of 88 features were extracted ° ° ° ° with GLCM from 0 , 45 , 90 , and 135 angle directions in d � 2 distance. Wavelet decomposition transform can denote PCA is used to reduce dimensionality of large dataset distribution of energy features of different regions (TEF). [38, 39]. We can select a number of features only up to one- ROI of the CT image is divided into four subbands with 2D third of the number of data (patterns) in the smaller of the wavelet decomposition. *ree images are created in low two classes. *us, for our work, the smallest class has 104 frequencies, and an image is created in high frequencies with patterns (benign nodules), and since we split the data to half, wavelet decomposition transform from an image [37]. In one-third of 52 is around 17. *erefore, we selected with this study, 13 energy features of an image are extracted with PCA the most appropriate 17 features (components) from wavelet decomposition. *e number of features extracted by 123 features. Figure 6 denotes principal component analysis each feature extraction method used in this study is pre- of extracted features with cumulative variance. As can be sented in Table 1. seen from the chart in Figure 6, it is seen that the variance of On lung CT images, malign nodules are generally more the first 20 components is more selective. complex and irregular, while benign nodules are rounder with certain borders. Most of the benign nodules have small variance values. However, malign nodules show relatively 3.4.NoduleClassification. In the proposed pipeline, we have higher variance values [3]. Figure 5 shows the examples of used a probabilistic neural network (PNN) model to make benign and malign lung nodules on CT images. automated decision about the nodule types (benign or Since 123 features extracted are rather large in size, they malignant). PNN is an effective tool for many classification may negatively affect accuracy during classification step. implementations and can easily make classifications [40, 41]. *us, selecting the most appropriate features instead of Figure 7 presents the architecture of the PNN designed for using all features will be a more efficient method. We used this study. Neuron number in the input layer is selected 17 PCA method for dimension reduction of feature vector. according to the number of inputs. 6 Journal of Healthcare Engineering (a) (b) Figure 5: Examples of benign (a) and malign (b) lung nodules. Korez Hospital. Its acquisition parameters are slice colli- mation 1.0 mm and slice width 1 mm. Scans were acquired in 130 kV and 75 mAs. •e size of the images was 512 × 512 0.8 pixels. •e images were stored as DICOM format Œles. •e database consists of 47 CT scans from 47 di˜erent patients. 35 of volunteer patients are male and remaining of them are 0.6 female. •eir ages are between 30 and 79 (mean 58.7 ± 10.5 years). All patients agreed that they have a legal and moral 0.4 right to accurate and reliable information for the scan. •ese patients should be given clearly the diagnosis and prognosis 0.2 with a simple language. •ere are a total of 9504 CTmodality images in the database, and the number of CTslices per scan 0 varies between 116 and 283. After the CTscan, the physician 0 20 40 60 80 100 120 140 provided the selection of the slice where the nodule is fully Principal components visible. 1128 ROI, which includes a total of 220 nodules (104 Variance benign and 116 malignant), were selected from 9504 CT Cumulative variance images with the help of a lung physician and three expe- rienced radiologists in the lung parenchyma. •is process Figure 6: Principal component analysis of extracted features with cumulative variance. has been conducted by means of an annotation tool. •e nodules were also approved by biopsy. Sizes of nodules change from 3 to 65 mm in diameter. •e size distribution of 4. Experimental Results the nodules is shown in Figure 9. In this work, we realized all experiments using a PC with i7 processor, 16 GB memory, and Windows 10. Moreover, 4.2. Validation of LUVEM with Evaluation Metrics. MATLAB software was used for performance evaluation of Proposed lung volume extraction method (LUVEM) in this the proposed pipeline. In all experiments, leave-one-out study is compared with the standard manual segmentations cross validation technique was run at the level of nodule. So, all of 220 nodules (104 benign and 116 malignant) were using measurement metrics. We evaluate manual segmen- tations of the expert and automated segmentations of used for both trainings and tests. Figure 8 summarizes the processing steps of the proposed pipeline. LUVEM using two popular overlap measures. We used a segmentation software tool developed by us for manual segmentation on the dataset. •e software tool outlines 4.1. CT Lung Dataset. In this study, an image dataset was edges automatically, presenting us to obtain contours of the prepared for the proposed pipeline. CT examinations were nodule boundaries. •e metrics evaluate the overlapping realized by using a helical CT scanner from Sincan NaŒz between the two sets. •e Œrst overlap metric, represents the Proportion of variance Journal of Healthcare Engineering 7 1 Benign nodule ... Output ... layer Malignant nodule Inputs Input layer Summation layer Pattern layer Figure 7: Probabilistic neural network architecture used in the proposed method for nodule classiŒcation. (a) (b) (c) (d) (e) (f) Figure 8: Processing steps of the proposed pipeline: (a) original DICOM image; (b) image preprocessing and enhancement; (c) lung volume extraction from CT scan; (d) detection of candidate nodules; (e) segmentation of nodules; (f) classiŒcation of nodules. Jaccard coe¡cient (union overlap), deŒned as intersection |S1 ∩ S2| Jaccard , over manual and automatic segmentations and measures the |S1 ∪ S2| similarity of the S1 and S2 sets [42]. Our second overlap (1) metric, the Dice coe¡cient (mean overlap), gives double the 2|S1 ∩ S2| weight to agreements between the two sets [43]. Jaccard and Dice . |S1| +|S2| Dice metrics are denoted in the following equations: 8 Journal of Healthcare Engineering Table 2: *e Jaccard and Dice metrics measures for LUVEM. Jaccard overlap Dice overlap Otsu’s method 0.587 ± 0.093 0.786 ± 0.088 LUVEM 0.867 ± 0.051 0.938 ± 0.032 Table 3: Confusion matrixes for feature extraction methods. 20 Classification Classification results without results with PCA FE method PCA TP FP FN TN TP FP FN TN <10 mm 10–20 mm >20 mm SSF 90 22 26 82 90 22 26 82 Nodule diameter SBF 105 17 11 87 109 12 7 92 Benign GTF 101 19 15 85 110 11 6 93 Malignant TEF 109 15 7 89 111 11 5 93 All FE methods Figure 9: Size distribution of benign and malignant nodules in the 111 12 5 92 113 6 3 98 (combined) image dataset. systems. Area under an ROC curve is measured according to We show the overlap metrics (Jaccard and Dice) that sensitivity and specificity values of system. *is area shows result from both LUVEM and Otsu’s method. *ese results how the system is successful. *erefore, we also present ROC are the comparison of automated segmentations of LUVEM curve of our proposed detection system. Figure 10 shows and Otsu methods with manual segmentation on 254 lung ROC curve of the system obtained classification results of CT image in our database. *e results in Table 2 showed each lung nodule group and overall system. As seen in this LUVEM is higher in Jaccard overlap (0.867) and Dice graphic, area under ROC curve and true positive rate of overlap (0.938) than Otsu’s method. small size nodules are lower than big size nodules. Here, as can be seen from this figure, if the nodule size is too large and too small, the success rate decreases. 4.3. Detection Rates. Confusion matrixes of classification Processing time is another performance criterion that we results with PNN according to each feature extraction and have used for the evaluation of the proposed pipeline. PCA method are presented in Table 3. As shown in Table 3, Longest time is needed for nodule detection step due to the the usage of PCA affects the detection performance of the use of SOM method for segmentation. Since SOM is an ANN pipeline positively. Moreover, the usage of combined fea- model, it has a lot of time-consuming mathematical oper- tures extraction methods with PCA gives best success rate. ations. In average, classification of a CT image as benign or Table 4 presents the values of performance criteria ob- malignant takes 2–3 seconds approximately. It can be ac- tained in the classification results of the proposed pipeline cepted as a reasonable time period when it is compared with when feature extraction methods were used separately and the time it needs for a physician to make decisions. together. According to the table, performance values are more successful when all feature extraction methods are used together. Accuracy (Acc) was found to be 92.27% when 5. Conclusions and Discussion 123 features were used in classification without feature se- lection through PCA, and this rate was found to be 95.91% In this study, a fully automated pipeline was proposed to with the use of PCA. Similarly, more successful results were classify benign and malign lung nodules on CT images. By obtained in sensitivity (Sen), specificity (Spc), positive de- means of the designed pipeline, nodule detection as well as cision value (PDV), negative decision value (NDV), and F1 benign/malign distinction was performed with high accu- score criteria as presented in Table 4. racy, sensitivity and specificity rates. Moreover, it was Since our CT image database was divided into 3 groups designed a preprocessing method called LUVEM for according to the size of nodules such as <10, 10–20, and> extracting the lung volume from CT images. SOM method 20 mm, we also realized a performance evaluation according was used to allow successful detection of lung nodules in to size group of the nodules. *ese experiments were realized early stages. According to the detailed experiment per- with all together feature extraction method using PCA. formed on large dataset with combined features, the pro- Table 5 presents the result of detection performance posed pipeline can differentiate benign/malign nodules with depending upon nodule size. As shown in Table 5, the high accuracy rates such as 94.68% (3–10 mm), 96.92% proposed pipeline can classify even small nodules with high (10–20 mm), and 96.25% (>20 mm) using PNN. *e pro- success rates. Overall detection result of proposed pipeline posed pipeline can be used by the physicians as a supple- according to the nodule size is 95.91%. mentary tool for benign and malign nodule classification. Receiver operator characteristic (ROC) curve is another We evaluated the performance of the pipeline on Lung popular performance evaluation criteria used in detection Imaging Database Consortium-Image Database Resource Number of nodules Journal of Healthcare Engineering 9 Table 4: Overall performance results of proposed pipeline. Classification results without PCA Classification results with PCA Performance criteria SSF SBF GTF TEF All SSF SBF GTF TEF All Acc 78.18 87.27 84.55 90.00 92.27 78.18 91.36 92.27 90.00 95.91 Sen 77.57 90.52 87.07 93.97 95.67 77.57 93.97 94.83 95.67 97.42 Spc 78.85 83.65 81.73 85.58 88.46 78.85 88.46 89.42 89.42 94.24 PDV 80.36 86.07 84.17 87.90 90.24 80.36 90.08 90.91 90.98 94.96 NDV 79.93 88.78 85.00 92.71 94.85 79.93 92.93 93.94 94.90 97.03 F1 0.79 0.88 0.85 0.92 0.93 0.79 0.92 0.93 0.94 0.96 Table 5: Assessment of performance measurement criteria according to nodule size. Confusion matrix Performance criteria Nodule size (mm) *e number of nodule TP FP FN TN Acc Sen Spc PDV NDV F1 <10 75 3 4 0 68 94.67 100 94.45 42.86 100 0.60 10–20 65 43 1 1 20 96.92 97.73 95.24 97.73 95.24 0.98 >20 80 67 1 2 10 96.25 97.10 90.91 98.53 83.34 0.98 Overall 220 113 6 3 98 95.91 97.42 94.24 94.96 97.03 0.96 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 – specifity (false positive rate) Nodule size: up to 10 mm Nodule size: bigger than 20 mm Nodule size: between 10 mm and 20 mm Overall Figure 10: ROC curve of classification precision in proposed pipeline in different nodule diameter. Table 6: *e performance evaluation of proposed pipeline on LIDC-IDRI. Confusion matrix Performance criteria *e number of nodules TP FP FN TN Acc Sen Spc PDV NDV F1 38 22 2 4 10 84.21 84.62 83.33 91.67 71.43 0.88 Initiative (LIDC-IDRI) as well [50]. LIDC-IDRI dataset is methods and PCA. In this test, F1 score result was found as the largest publicly available reference database for detection 0.88. *e obtained performance evaluation values of pro- of lung nodules. We choose LIDC-IDRI dataset since it posed pipeline on LIDC-IDRI dataset are denoted in Table 6. contains almost all the related information for lung CT *ere are some advantages of the proposed pipeline including annotations on nodule sizes, locations, diagnosis compared to the state-of-the-art systems. Firstly, the pro- results, and other information. We collected a total of 38 posed pipeline has two diagnosis possibilities. It can perform lung nodules from the dataset, including 26 malignant and nodule detection together with nodule classification. Second 12 benign nodules. According to the evolution results on the advantage of the proposed pipeline is to provide the de- proposed pipeline, accuracy obtained 84.21% using all FE tection of small nodules in the lung with the use of SOM Sensitivity (true positive rate) 10 Journal of Healthcare Engineering Table 7: *e comparison of our pipeline with previously published CADs. CAD system CT image database Number of cases Nodule size(mm) Sensitivity (%) Average FPR Dehmenski et al. [9] *eir own database 70 3–20 90.0 14.6 Suarez-Cuenca et al. [10] *eir own database 22 4–27 80.0 7.7 Opfer and Wiemeker [46] LIDC database [47, 48, 50] 93 ≥4 74.0 4 Rubin et al. [51] *eir own database 20 ≥3 76 3 Sahiner et al. [49] LIDC database [47, 48, 50] 48 3–36.4 79 4.9 Messay et al. [24] LIDC database [47, 48, 50] 84 3–30 82.66 3 Suzuki et al. [52] *eir own database 101 8–20 80.3 16.1 Park et al. [53] *eir own database 38 Indefinite 80 – Choi and Choi [23] LIDC database [47, 48, 50] 32 3–30 94.1 5.45 Choi and Choi [44] LIDC database [47, 48, 50] 58 3–30 95.28 2.27 Proposed method Our database 47 3–35 97.42 4.54 method during segmentation step. *is is remarkable in [4] N. Howlader, A. Noone, M. Krapcho et al., SEER Cancer Statistics Review, 1975-2011, National Cancer Institute, terms of early detection of lung cancer. *ird advantage of Bethesda, MD, USA, 2014. the proposed pipeline is to have relatively high detection [5] H. Han, L. Li, F. Han, B. Song, W. Moore, and Z. Liang, “Fast performance. Accuracy, sensitivity, and specificity of the and adaptive detection of pulmonary nodules in thoracic CT system were calculated as 95.91%, 97.42%, and 94.24%, images using a hierarchical vector quantization scheme,” IEEE respectively. It is fairly difficult to compare formerly re- Journal of Biomedical and Health Informatics, vol. 19, no. 2, ported CAD systems due to different datasets, nodule types, pp. 648–659, 2015. sizes, and validation methods. We picked out some CAD [6] Y. J. Jeong, C. A. Yi, and K. S. Lee, “Solitary pulmonary systems to compare their performances. Some of them nodules: detection, characterization, and guidance for further [23, 24, 44–46] used the LIDC database [47–49], and the diagnostic workup and treatment,” American Journal of other used their own databases. Table 7 denotes the com- Roentgenology, vol. 188, no. 1, pp. 57–68, 2007. [7] S. Ozekes and A. Y. Camurcu, Automatic Lung Nodule De- parison of the proposed pipeline with some CAD systems. tection Using Template Matching, Springer, Berlin, Germany, When the results are analyzed, our pipeline has high sen- sitivity on our CT image dataset. [8] A. M. Schilham, B. Van Ginneken, and M. Loog, “A computer-aided diagnosis system for detection of lung Data Availability nodules in chest radiographs with an evaluation on a public database,” Medical Image Analysis, vol. 10, no. 2, pp. 247–58, *e data used to support the findings of this study are available from the corresponding author upon request. [9] J. Dehmeshki, X. Ye, X. Lin, M. Valdivieso, and H. Amin, “Automated detection of lung nodules in CT images using shape-based genetic algorithm,” Computerized Medical Im- Conflicts of Interest aging and Graphics, vol. 31, no. 6, pp. 408–17, 2007. [10] J. J. Suarez-Cuenca, P. G. Tahoces, M. Souto et al., “Ap- *e author declares that there are no conflicts of interest plication of the iris filter for automatic detection of pul- regarding the publication of this paper. monary nodules on computed tomography images,” Computers in Biology and Medicin, vol. 39, no. 10, pp. 921–33, 2009. Acknowledgments [11] K. Murphy, B. Van Ginneken, A. M. Schilham, B. J. De Hoop, H. A. Gietema, and M. Prokop, “A large-scale evaluation of *e author would like to thank representatives of Sincan automatic pulmonary nodule detection in chest CTusing local Nafiz Korez Hospital for creating the dataset. image features and k-nearest-neighbour classification,” Medical Image Analysis, vol. 13, no. 5, pp. 757–70, 2009. [12] M. L. Giger, K. T. Bae, and H. Macmahon, “Computerized References detection of pulmonary nodules in computed tomography [1] R. Siegel, D. Naishadham, and A. Jemal, “Cancer statistics,” images,” Investigative Radiology, vol. 29, no. 4, pp. 459–65, CA: A Cancer Journal for Clinicians, vol. 62, no. 1, pp. 10–29, 1994. 2012. [13] J. Hasegawa, K. Mori, J. Toriwaki, H. Anno, and K. Katada, [2] C. Jacobs, E. M. Van Rikxoort, T. Twellmann et al., “Auto- “Automated extraction of lung cancer lesions from multi-slice matic detection of subsolid pulmonary nodules in thoracic chest CT images by using three-dimensional image pro- computed tomography images,” Medical Image Analysis, cessing,” Systems and Computers in Japan, vol. 25, no. 11, vol. 18, no. 2, pp. 374–84, 2014. pp. 68–77, 1994. [3] E. Dandil, M. Cakiroglu, Z. Eksi, M. Ozkan, O. K. Kurt, and [14] K. Kanazawa, Y. Kawata, N. Niki et al., “Computer-aided A. Canan, “Artificial neural network-based classification diagnosis for pulmonary nodules based on helical CT images,” system for lung nodules on computed tomography scans,” in Computerized Medical Imaging and Graphics, vol. 22, no. 2, Proceedings of 2014 6th International Conference of Soft pp. 157–67, 1998. Computing and Pattern Recognition (SoCPaR), pp. 382–86, [15] K. Suzuki, F. Li, S. Sone, and K. Doi, “Computer-aided di- Tunis, Tunisia, August 2014. agnostic scheme for distinction between benign and Journal of Healthcare Engineering 11 malignant nodules in thoracic low-dose CT by use of massive [33] U. Akilandeswari, R. Nithya, and B. Santhi, “Review on training artificial neural network,” IEEE Transactions on feature extraction methods in pattern classification,” Euro- pean Journal of Scientific Research, vol. 71, no. 2, pp. 265–272, Medical Imaging, vol. 24, no. 9, pp. 1138–50, 2005. [16] T. Sun, J. Wang, X. Li et al., “Comparative evaluation of 2012. [34] M. Yang, K. Kpalma, and J. Ronsin, “A survey of shape feature support vector machines for computer aided diagnosis of lung extraction techniques,” Pattern Recognition, pp. 43–90, 2008. cancer in CT based on a multi-dimensional data set,” Com- [35] D. A. Clausi, “An analysis of co-occurrence texture statistics as puter Methods and Programs in Biomedicine, vol. 111, no. 2, a function of grey level quantization,” Canadian Journal of pp. 519–24, 2013. Remote Sensing, vol. 28, no. 1, pp. 45–62, 2002. [17] J. Kuruvilla and K. Gunavathi, “Lung cancer classification [36] R. M. Haralick, K. Shanmuga, and I. Dinstein, “Textural using neural networks for CT images,” Computer Methods and features for image classification,” IEEE Transactions on Sys- Programs in Biomedicine, vol. 113, no. 1, pp. 202–9, 2014. tems Man and Cybernetics, vol. SMC-3, no. 6, pp. 610–621, [18] M. Javaid, M. Javid, M. Z. U. Rehman, and S. I. A. Shah, “A novel approach to CAD system for the detection of lung [37] G. Van, P. Wouver, D. Scheunders, and D. Van, “Statistical nodules in CT images,” Computer Methods and Programs in texture characterization from discrete wavelet representa- Biomedicine, vol. 135, pp. 125–139, 2016. tion,” IEEE Transactions on Image Processing, vol. 8, no. 4, [19] M. Z. ur Rehman, M. Javaid, S. I. A. Shah, S. O. Gilani, pp. 592–598, 1998. M. Jamil, and S. I. Butt, “An appraisal of nodules detection [38] H. Camdevyren, N. Demyr, A. Kanik, and S. Keskyn, “Use of techniques for lung cancer in CT images,” Biomedical Signal principal component scores in multiple linear regression Processing and Control, vol. 41, pp. 140–151, 2018. models for prediction of Chlorophyll-a in reservoirs,” Eco- [20] Z. Wang, J. Xin, P. Sun, Z. Lin, Y. Yao, and X. Gao, “Improved logical Modelling, vol. 181, no. 4, pp. 581–589, 2005. lung nodule diagnosis accuracy using lung CT images with [39] L. H. Chen and S. Y. Chang, “An adaptive learning algorithm uncertain class,” Computer Methods and Programs in Bio- for principal component analysis,” IEEE Transactions on medicine, vol. 162, pp. 197–209, 2018. Neural Networks, vol. 6, no. 5, pp. 1255–1263, 1995. [21] H. Xie, D. Yang, N. Sun, Z. Chen, and Y. Zhang, “Automated [40] K. Z. Mao, K. C. Tan, and W. Ser, “Probabilistic neural- pulmonary nodule detection in CT images using deep con- network structure determination for pattern classification,” volutional neural networks,” Pattern Recognition, vol. 85, IEEE Transactions on Neural Networks, vol. 11, no. 4, pp. 109–119, 2019. pp. 1009–1016, 2000. [22] DICOM (Digital Imaging and Communications in Medicine), [41] D. F. Specht, “Probabilistic neural networks,” Neural Net- http://medical.nema.org. works, vol. 3, no. 1, pp. 109–118, 1990. [23] W. J. Choi and T. S. Choi, “Genetic programming-based [42] P. Jaccard, “*e distribution of the flora in the alpine zone,” feature transform and classification for the automatic de- New Phytologist, vol. 11, no. 2, pp. 37–50, 1912. tection of pulmonary nodules on computed tomography [43] L. R. Dice, “Measures of the amount of ecologic association images,” Information Sciences, vol. 212, pp. 57–78, 2012. between species,” Ecology, vol. 26, no. 3, pp. 297–302, 1945. [24] T. Messay, R. C. Hardie, and S. K. Rogers, “A new compu- [44] W.-J. Choi and T.-S. Choi, “Automated pulmonary nodule tationally efficient CAD system for pulmonary nodule de- detection system in computed tomography images: a hierar- tection in CT imagery,” Medical Image Analysis, vol. 14, no. 3, chical block classification approach,” Entropy, vol. 15, no. 2, pp. 390–406, 2010. pp. 507–523, 2013. [25] X. Ye, X. Lin, J. Dehmeshki, G. Slabaugh, and G. Beddoe, [45] J. J. Erasmus, J. E. Connolly, H. P. Mcadams, and V. L. Roggli, “Shape-based computer-aided detection of lung nodules in “Solitary pulmonary nodules: Part I. Morphologic evaluation thoracic CT images,” IEEE Transactions on Biomedical En- for differentiation of benign and malignant lesions 1,” Ra- gineering, vol. 56, no. 7, pp. 1810–1820, 2009. diographics, vol. 20, no. 1, pp. 43–58, 2000. [26] S. Akram, M. Y. Javed, A. Hussain, F. Riaz, and M. Usman [46] R. Opfer and R. Wiemker, “Performance analysis for com- Akram, “Intensity-based statistical features for classification puter aided lung nodule detection on LIDC data,”art. no. of lungs CT scan nodules using artificial intelligence tech- 65151C, Medical Imaging 2007: Image Perception, Observer niques,” Journal of Experimental and >eoretical Artificial Performance, and Technology Assessment, vol. 6515, pp. Intelligence, vol. 27, no. 6, pp. 1–15, 2015. C5151–C5151, 2007. [27] R. O. Duda and P. E. Hart, “Use of Hough transformation to [47] M. F. Mcnitt-Gray, S. G. Armato, C. R. Meyer et al., “*e Lung detect lines and curves in pictures,” Communications of the Image Database Consortium (LIDC) data collection process ACM, vol. 15, no. 1, pp. 11–15, 1972. for nodule detection and annotation,” Academic Radiology, [28] H. Yazid, H. Yazid, M. Harun et al., “Circular discontinuities vol. 14, no. 12, pp. 1464–1474, 2007. detection in welded joints using Circular Hough Transform,” [48] A. P. Reeves, A. M. Biancardi, T. V. Apanasovich et al., “*e NDT and E International, vol. 40, no. 8, pp. 594–601, 2007. Lung Image Database Consortium (LIDC): a comparison of [29] T. Kohonen, “Self-organizing maps of massive databases,” different size metrics for pulmonary nodule measurements,” Engineering Intelligent Systems for Electrical Engineering and Academic Radiology, vol. 14, no. 12, pp. 1475–1485, 2007. Communications, vol. 9, no. 4, pp. 179–185, 2001. [49] B. Sahiner, L. M. Hadjiiski, H. P. Chan et al., “Effect of CAD [30] S. Haykin, Neural Networks, A Comprehensive Foundation, on radiologists’ detection of lung nodules on thoracic CT Hamilton, ON, Canada, 1999. scans: observer performance study,”art. no. 65151D, Medical [31] D. Chi, “Self-organizing map-based color image segmentation Imaging 2007: Image Perception, Observer Performance, and with k-means clustering and saliency map,” ISRN Signal Technology Assessment, vol. 6515, pp. D5151–D5151, 2007. Processing, vol. 2011, Article ID 393891, 18 pages, 2011. [50] S. G. Armato III, G. Mclennan, M. F. Mcnitt-Gray et al., “Lung [32] C. Hollitt, “A convolution approach to the circle Hough image database consortium: developing a resource for the transform for arbitrary radius,” Machine Vision and Appli- medical imaging research community 1,” Radiology, vol. 232, cations, vol. 24, no. 4, pp. 683–694, 2013. no. 3, pp. 739–748, 2004. 12 Journal of Healthcare Engineering [51] G. D. Rubin, J. K. Lyo, D. S. Paik et al., “Pulmonary nodules on multi–detector row CT scans: performance comparison of radiologists and computer-aided detection 1,” Radiology, vol. 234, no. 1, pp. 274–283, 2005. [52] K. Suzuki, S. G. Armato III, F. Li, S. Sone, and K. Doi, “Massive training artificial neural network (MTANN) for reduction of false positives in computerized detection of lung nodules in low-dose computed tomography,” Medical Physics, vol. 30, no. 7, pp. 1602–1617, 2003. [53] S. C. Park, J. Tan, X. Wang et al., “Computer-aided detection of early interstitial lung diseases using low-dose CT images,” Physics in Medicine and Biology, vol. 56, no. 4, pp. 1139–53, 2011. International Journal of Advances in Rotating Machinery Multimedia Journal of The Scientific Journal of Engineering World Journal Sensors Hindawi Hindawi Publishing Corporation Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 http://www www.hindawi.com .hindawi.com V Volume 2018 olume 2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Journal of Control Science and Engineering Advances in Civil Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Submit your manuscripts at www.hindawi.com Journal of Journal of Electrical and Computer Robotics Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 VLSI Design Advances in OptoElectronics International Journal of Modelling & Aerospace International Journal of Simulation Navigation and in Engineering Engineering Observation Hindawi Hindawi Hindawi Hindawi Volume 2018 Volume 2018 Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com www.hindawi.com www.hindawi.com Volume 2018 International Journal of Active and Passive International Journal of Antennas and Advances in Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018
Journal of Healthcare Engineering – Hindawi Publishing Corporation
Published: Nov 1, 2018
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.