Lung Nodules Detection with CNN for Deep Learning

Abstract:

Lung cancer is one of the most serious cancers in the world, with the smallest survival rate after the diagnosis. In CT scans, lung nodules appear as dense masses of various shapes and sizes. They may be isolated from or attached to other structures such as blood vessels or the pleura. In this paper a detection of Candidate Nodules (solitary or juxtapleural in a 2D CT slice is achieved using two schemes of segmentation and enhancement algorithms. Convolutional Neural Network (CNN) for deep learning classification is used as a revolutionary image recognition method to distinguish between two types of nodules according to its location (juxtapleural and solitary lung nodules). Our CAD system achieves accuracies of first scheme of segmentation for detecting solitary nodules and juxtapleural nodules by using CNN are ……and ……. respectively. Also achieves accuracies of second scheme of segmentation for detecting solitary nodules and juxtapleural nodules by using CNN are ……and ……. respectively.

Keywords: CT scans, Convolutional Neural Network (CNN), Deep Learning, Computer Aided Detection (CAD).

Save your time!
We can take care of your essay

• Proper editing and formatting
• Free revision, title page, and bibliography
• Flexible prices and money-back guarantee

Place an order

I. Introduction

Lung cancer is a disease that consists of uncontrollable growth of cell and tissues of the lung which may lead to metastasis that is the infestation of adjacent and nearby tissue and infiltration beyond the lungs. From epithelial cells Carcinomas are derived which are the vast majority of primary lung cancers. Lung cancer, the most usual cause of cancer-imputed death in men and women. An estimated new lung cancer cases 14% for males and 12% females in US in 2017 [1].

The early detection of lung cancer can increase overall 5-year survival rates by extracting the lung nodules. Hence, this diagnosis can improve the effectiveness of treatment. Traditional x-ray and computed tomography (CT scan) are attempted to diagnose lung nodules. Treatment of lung nodules depends on the histological

type of cancer, the stage, and the patient's status, but overall only 14% of people diagnosed with lung cancer survive five years after the diagnosis[2].

Because of small size of nodule in the lung, it is difficult to distinguish between it and another mass in a 2D slice.

Actually, the search for micro-nodules does not always make sense on single slices: the nodule shape, size and gray tone are very similar to vessels sections. therefore, a segmentation step is very important to distinguish between the small nodules and blood vessels. Hence, the type of nodules according to its location (solitary or juxtapleural) will be easily classified.

In this paper, we propose a two schemes of segmentation and enhancement for nodule emphasis which extracts a 2D candidate lung nodules. A convolutional neural network is used as a deep learning tool for the classification of juxtapleural and solitary lung nodules.

II. Related Work:

To date, many types of research about nodule detection by using CAD system have been developed. It begins with preprocessing and segmentation followed by the classification step.

For instance, Diego et al [3] used a method composed of four processes for lung nodule detection. The first step employed image acquisition and pre-processing. The second stage involved a 2D algorithm to affect every layer of a scan eliminating non-informative structures inside the lungs, and a 3D blob algorithm associated with a connectivity algorithm to select possible nodule shape candidates. The final step utilized a support vector machine for classifying the possible candidates into nodules and non-nodules depending on their features. QingZeng et al [4] proposed to employ, respectively, the convolution neural network (CNN), deep neural network (DNN), and stacked autoencoder (SAE). The experimental results show that the CNN network archived the best performance with an accuracy of 84.15%, sensitivity of 83.96%, and specificity of 84.32%, which has the best result among the three networks. Yang et al [5] described a 3D pulmonary nodule detection scheme utilizing MSCT images. This method segmented the initial nodule candidates first and extracted voxels features based on analysis eigenvalues of Hessian matrix. Then support vector machine (SVM) and decision rule are applied to categorize them into two sorts to remove FPs. Sarah et al [6] used Gaussian smoothing kernel for filtration which helps to reduce noise effects. Next, features such as sphericity, mean and variance of the gray level, elongation and border variation of potential nodules are extracted to classify detected nodules to malignant and benign groups. Fuzzy KNN is employed to classify potential nodules as non-nodule or nodule with different degree of malignancy.

Serhat et al. [7] used Genetic Cellular Neural Networks (G-CNN) for segmentation. Then, for each lung region, ROIs were specified with using the 8 directional search; +1 or -1 values were assigned to each voxel. A 3D template was created to find the nodule-like structures on the 3D ROI image. Convolution of the 3D ROI image with the proposed template strengthens the shapes that are similar to those of the template and it weakens the other ones.

Finally, fuzzy rule based thresholding was applied and the ROI’s were found. The computer aided diagnosis (CAD) system achieved 100% sensitivity with 13.375 FPs per case when the nodule thickness was greater than or equal to 5.625 mm.

Jin et al [8] proposed convolution neural network as a classifier to detect the lung nodules. The system achieved 84.6% of accuracy, 86.7% of specificity and 82.5% of sensitivity.

Thomas et al [9] developed CAD system to detect and localize 60.1% of all the nodules with an average number of 2.1 (1.5%) false positives per slice. In addition, three different types of neural network structures for this CAD system are tested and compared. With 95% confidence we can conclude that deeper neural networks decrease the false positives significantly

Wu et al [10] combined a several common image processing techniques with complex segmentation step, such as connectivity labeling, binarization, mathematical operation and hole-filling.

Ming et al [11] applied the watershed algorithm to estimate the segmentation and then used a region growing method. Xujiong et al [12] proposed a two-step segmentation method for lung extraction. Firstly, a 3-D adaptive fuzzy thresholding technique and secondly applied a 2-D-based post refinement process on the lung contour chain code to obtain a complete lung mask. Michela et al [13] used a dynamic threshold for identification of three different groups corresponding, respectively, to the upper, middle and lower parts of the lung volume. The slices of the lung middle part and a threshold determined empirically for all other slices. In their tests, they applied thresholding with fixed threshold to the first 30 and the last 30 slices of a CT scan. Sasidhar et al [14] applied two steps of segmentation: firstly, extracted lung parenchyma by using a threshold of -420 Hounsfield Unit (HU) as:

Gray Level Value = 1024 + HU

Secondly, extract lung nodules using the threshold of -150 HU.

Qingxiang et al [15] proposed a method implements an active evolution and structure enhancement which can segment blood vessels and detect pulmonary nodules at a high accuracy.

Firstly, he introduced a vessel energy function (VEF) during active evolution to help distinguishing the nodules from vessels. VEF consists of three energy terms, which are gradient term, intensity term, and structure term.

= Fgradient ・ Fintensity・ Fstructure

Secondly, he utilized a radius-variable sphere model to refine the extracted contours. Candidate blood vessel centerline points, denoted as

V = {,,…..,}, are first selected.

Serhat et al [16] segmented the lung regions of the CTs by using Genetic Cellular Neural Networks

(G-CNN). Then, for each lung region, ROIs were specified with using the 8 directional search; +1 or -1 values were assigned to each voxel. The 3D ROI image was obtained by combining all the 2-Dimensional (2D) ROI images. A 3D template was created to find the nodule-like structures on the 3D ROI image.

Qing et al [17] transformed a number of correlated variables into a smaller number of uncorrelated variables, which are called principal components depended on their cumulative variance proportion that is called principal component analysis (PCA).

III. Materials And Methods:

A. Dataset:

A 14 digital CT consisting of 2991 2D slices which contains 172 nodules (100 solitary nodule and 72 juxtapleural nodule) are collected with approval from Cornell University [18] each abnormal image contains a tumor with equivalent diameters of lung nodules ranging from 7.78 mm to 22.48 mm. The in-slice (x, y) resolution is 0.703×0.703 mm and the CT slice thickness is 1.25 mm in DICOM format and has 512×512 pixels.

B. Implementation:

• Software:
  • Windows 10 Pro 64-bit
  • Matlab R2018b toolboxes
• Hardware:
  • Processor: Intel(R) Core™ i7, 2.70Ghz
  • RAM: 8 GB
  • Display Adapter: NVIDIA NVS 4200M

C. Segmentation and Enhancement for Nodule emphasis:

Lung segmentation is a prerequisite step for developing an automated computer-aided diagnosis system for CT scans of thorax that can lead to the early diagnosis of lung cancer as it is a separating task of the lung region from other anatomical portion of the chest CT image.

Inhomogeneity in the lung region is a very challenging problem as there is similar densities such as veins, arteries, bronchioles. A wealth of known publications has addressed the segmentation of lung regions from CT images and chest radiographs.

Here this phase is meant to remove the unwanted parts and to enhance the visibility of extracted pulmonary nodule. The preprocessing component reduces noise and artifacts in the lung image slices. refers to tasks necessary for enhancing the quality of displayed image by rectifying distortions due to media decay or motion artifacts.

In this paper, a two schemes of segmentation algorithm are proposed as follows:

a. Scheme I (Thresholding + Morphology):

It consists of four main steps as shown in fig.1

Figure 1: Scheme I of segmentation

(Thresholding + Morphology)

· Thresholding:

Selects thresholds for the input DICOM image, by calculating the optimum threshold separating two classes (foreground as 1 and background as 0). It reduces grey level images into binary images [19] [20] [21], If g (x, y) could be a threshold version of f (x, y) at some global threshold T, it is often outlined as g (x, y) = 1 if f (x, y) ≥ T

= 0 otherwise

The thresholding equation is outlined as:

T = M [x, y, p (x, y), f (x, y)]

during this equation, T refers to the threshold; f (x, y) is the gray value of point (x, y) and p (x, y) denotes some native property of {the point | the purpose} such as the average gray value of the neighborhood centered on point (x, y) [22].

the resulted image of the Thresholding step shown in fig.1 (B)

· Clearing lung border:

“Imclearborder” function Suppresses structures that are lighter than their surroundings and that are connected to the image border [23] [14], as shown in fig.1 (C).

· Closure operation [21]:

Morphological operation (Erosion and Dilation) was applied for smoothing the outline of the segmented lung. The morphological close operation is a dilation followed by an erosion, using the same structuring element (5×5 kernel) for both operations. This operation enhances the lung borders and fills the gaps which can cause deficiencies in the phase of border detection.

Fig.1 (D) showed the resulted image of the closing operation step.

· Superimposing [14] [21] and 2D candidate

Representing the lung masses which are superimposed on the original image as shown in fig.1 (E).

(A)

(B)

(E)

(D)

Figure 2: Results of scheme I of segmentation (Thresholding + Morphology):

A. original CT image, B. Thresholding, C. Clearing lung border, D. Closure operation, E. Superimposing and 2D candidate nodule

b. Scheme II (Bounding box + Maximum intensity projection):

It consists of the following image processing methods:

· Bilateral filter [24] [25]:

Is a non-linear and edge preserving filter. It replaces the pixel values of the image with a weighted average of similar and nearby pixel values. It filters the image using range and domain filter. Bilateral filter is defined as

( (1)

(C) where the normalization term

(2) ensures that the filter preserves image energy and is the original input image to be filtered;

{displaystyle x} is the original input image to be filtered; are the coordinates of the current pixel to be filtered;

{displaystyle Omega }Ω is the window centered in ; {displaystyle x}

{displaystyle f_{r}} is the range kernel for smoothing differences in intensities;

{displaystyle g_{s}} is the spatial kernel for smoothing differences in coordinates (this function can be a Gaussian function). The resulted image of bilateral filter shown in fig. 3(A)

· Thresholding gray-level transformation function [26]:

Applied by using bit plane slicing method, it maps all gray levels of the image from 0 to 255, by converting the value of pixels that ranging from 0 to 127 to one level (e.g., 0) and maps all levels above 127 to another level (e.g., 1) that resulting in transformation of the gray image into binary image as shown in fig. 3 (B).

· Bounding box:

Measuring the image properties by initializing small rectangle (bounding box) which containing all the regions, specified as a 1 [27] [28]. Thresholding to isolate the two lungs easily is used by converting all zeros pixels to ones and the opposite. remove border artifact by clearing lung border, erosion and dilation are applied also [29]. Superimposing is done by multiplication of original image by the last image resulted from the dilation step.

· Maximum intensity projection (MIP) [30]:

evaluates the projecting of the voxel with the highest attenuation value on every slice throughout the volume onto each XY coordinate or a 2D image, only the pixel with the highest Hounsfield number along the Z-axis is projected so that in a single bidimensional image all dense structures (nodules) in a given volume as illustrated in next figure [31].

Figure 3: MIP principle

Also a morphological close operation is done after MIP using a structuring element (size 2 pixels), so that the detection of lung nodule become easily. Fig. 4 shows the segmentation results of scheme II.

D. Convolutional neural network for deep learning classification:

Convolutional neural networks (CNN) are a successful tool for deep learning classification [30] [32], and developed to suit image recognition as it is a multilayer neural network which consists of one or more convolution layers followed by one or more fully connected layers.

Our convolutional neural network architecture is consisting of convolution layer, max pooling layer, fully connected layer and softmax layer as shown in fig.4

In the input image layer, we specify the input image size as 512×512 and the channel size is 1. the filter size in the convolution layer is [5,5] and number of filters is 20. The max pooling layer returns the maximum values of rectangular regions of inputs, in our work the size of the rectangular region is [2,2]. The fully connected layer combines all of the features (local information) that learned by all previous layers across the image to classify these images. As the output parameter in this layer is equal to the number of classes in the target data, the output size will be 2 classes (solitary and juxtapleural nodules).

Solitary nodule

Juxta pleural nodule

Figure 5: The convolutional neural network CNN architecture.

(A)

(B)

(F)

(E)

(I)

Figure 3: Results of scheme II of segmentation (MIP + Bounding box):

A. original CT image, B. bilateral filtering, C. Bit plane slicing, D. bounding box, E. thresholding, F. clearing lung border, G. Filling holes, H. Superimposing and I. candidate nodule with MIP

(D)

(C)

(H)

(G)

IV. Results And Discussion:

In this paper we propose a Computer Aided Detection (CADe) system to detect candidate nodules either solitary or juxtapleural nodules in a 2D CT slice regarding to its location. Two Segmentation and enhancement schemes (Thresholding + Morphology and bounding box + MIP) are achieved and their results is shown in figure 2 and figure 3 respectively. Convolutional Neural Network (CNN) is achieved for deep learning classification as a revolutionary image recognition method to classify between juxtapleural and solitary lung nodules. The results of sensitivity, specificity and accuracy of CNN for deep learning classification distinguishing between solitary and juxtapleural nodules for the two schemes of segmentation is shown in table 1 and table 2. Our CAD system achieves ….. , ….. and ….. as a sensitivity, specificity and accuracy of CNN respectively to classify the solitary nodules and ……, ……. And …… as sensitivity, specificity and accuracy of CNN respectively to classify the juxtapleural nodules when using scheme 1 of segmentation. It achieves also ….. , ….. and ….. as sensitivity, specificity and accuracy of CNN respectively to classify the solitary nodules and ……, ……. And …… as sensitivity, specificity and accuracy of CNN respectively to classify the juxtapleural nodules when using scheme 2 of segmentation. Results provided that the CNN accuracy of solitary nodules detection

Table1: Results of CNN for deep learning classification with scheme 1 of segmentation (Thresholding + Morphology) distinguishing between Solitary and juxtapleural nodules

Nodule type

Sensitivity%

Specificity%

Accuracy%

Solitary

97

Juxtapleural

96

Table2: Results of CNN for deep learning classification with (Bounding box + MIP) distinguishing between Solitary and juxtapleural nodules

Nodule type

Sensitivity%

Specificity%

Accuracy%

Solitary

95

Juxtapleural

93.3

Based on literature research shown in table 3 a sensitivity, accuracy and database are observed. However, systems showed promising results, for example, regarding to the parameters of sensitivity, accuracy and database, stood out the systems of Qaisar et al. [37], Mizuho et al. [38] and Patrice et al. [39]. The first tested his method with 250 different nodules and had an accuracy of 88%. The second validated his system with 665 nodules and obtained accuracy of 68%. The third tested his method with 2635 nodules achieved an accuracy of 88.28% and Sensitivity 83.82%. However, validation of the systems was not tested with a

broad range of nodules types; it shows promising results. Comparing to our system, Patrice et al. [39] achieved 88.28% and Sensitivity 83.82% when using Deep Convolutional Neural Network (DCNN) for nodule detection knowing that the nodule type not informed and our system showed best result when using scheme 1 of segmentation achieving 97% accuracy for detection of solitary nodules when using CNN for deep learning classification with sensitivity …………… and 97% for detection of juxtapleural nodules when using CNN for deep learning classification with sensitivity ……………

Study

Year

Database

Nodule type

Accuracy%

Qaisar et al. [37]

2017

250

Ground glass opacity

88%

Mizuho et al. [38]

2018

665

NI

68%

Patrice et al. [39]

2018

2635

NI

Accuracy 88.28%

Sensitivity 83.82%

Table3: Comparison with related work of CNN for deep learning classification

(NI: Not Informed)

V. Conclusion

A Computer-Aided Detection (CADe) system to detect candidate nodules either solitary or juxtapleural nodules in a 2D CT slice regarding to its location is proposed by applying Two segmentation and enhancement schemes (Thresholding + Morphology and bounding box + MIP) on a database consisting of 14 digital CT consisting of 2991 2D slices which contain 172 nodules (100 solitary nodule and 72 juxtapleural nodule). Convolutional Neural Network (CNN) for deep learning classification is applied to classify between juxtapleural and solitary lung nodules with equivalent diameters of lung nodules (solitary and juxtapleural) ranging from 7.78 mm to 22.48 mm.

References:

1. American Cancer Society. Cancer Facts and Figures, 2017.
2. Minna. JD and Schiller JH,' Harrison's Principles of Internal Medicine (17th ed.)', McGraw-Hill, pp. 551–562, 2008.
3. Diego M. Peña, Shouhua Luo, and Abdeldime M. S. Abdelgader,' Auto Diagnostics of Lung Nodules Using Minimal Characteristics Extraction Technique”, Diagnostics, vol. 6, 13, 2016.
4. QingZeng Song, Lei Zhao, XingKe Luo, and XueChen Dou, 'Using Deep Learning for Classification of Lung Nodules on Computed Tomography Images', Journal of Healthcare Engineering, Vol., 2017.
5. Yang Liu, Jinzhu Yang, Dazhe Zhao and Jiren Liu 'Computer-Aided Detection of Lung Nodules Based on Voxel Analysis utilizing Support Vector Machines', International Conference on Future BioMedical Information Engineering, 2009.
6. Sarah Taghavi Namin, Hamid Abrishami and Mohammad Esmaeil-Zadeh,' Automated Detection and Classification of Pulmonary Nodules in 3D Thoracic CT Images', IEEE, pp. 4244-6588, 2010.
7. Serhat Ozekes,, Onur Osman and Osman N. Ucan, 'Nodule Detection in a Lung Region that’s Segmented with Using Genetic Cellular Neural Networks and 3D Template Matching with Fuzzy Rule-Based Thresholding', Korean J Radiol, vol. 9, pp. 1-9, 2008.
8. Jin, X., Zhang, Y., and Jin, Q., “Pulmonary Nodule Detection Based on CT Images Using Convolution Neural Network”, 9Th International Symposium On Computational Intelligence and Design, 2016.
9. Thomas HEENEMAN and Mark HOOGENDOORN, 'Lung nodule detection by using Deep Learning', VRIJE UNIVERSITEIT AMSTERDAM, research paper, Jan 2018.
10. Wu Suiyuan and Wang Junfeng,” Pulmonary Nodules 3D Detection on Serial CT Scans”, Third Global Congress on Intelligent Systems, 2012
11. Ming Yang, Senthil Periaswam and Ying Wu, “False positive reduction in lung GGO nodule detection with 3D volume shape descriptor”, IEEE International Conference on, Vol. 1, 2007.
12. Xujiong Ye∗, Xinyu Lin, Jamshid Dehmeshki and Gareth Beddoe,” Shape-Based Computer-Aided Detection of Lung Nodules in Thoracic CT Images”, IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 56, NO. 7, JULY 2009.
13. Michela Antonelli, Beatrice Lazzerini, Francesco Marcelloni,” Segmentation and reconstruction of the lung volume in CT images”, ACM Symposium on Applied Computing, 2005.
14. Sasidhar B, Ramesh Babu D R, Bhaskarao N, Basheer Jan,” Automated Segmentation of Lung Regions and Detection of Lung Cancer in CT scan”, International Journal of Engineering and Advanced Technology (IJEAT), Vol. 2, Issue-4, April 2013.
15. Qingxiang Zhu, Hongkai Xiong, and Xiaoqian Jiang,” Pulmonary Blood Vessels and Nodules Segmentation via Vessel Energy Function and Radius-Variable Sphere Model”, IEEE Second Conference on Healthcare Informatics, Imaging and Systems Biology, 2012.
16. Serhat Ozekes, Onur Osman, Osman N. Ucan, “Nodule Detection in a Lung Region that’s Segmented with Using Genetic Cellular Neural Networks and 3D Template Matching with Fuzzy Rule-Based Thresholding”, Korean J Radiol, 2008.
17. Qing-zhu Wang, Ke Wang, Yang Guo and Xin-zhu Wang, “Automatic Detection of Pulmonary Nodules in Multi-slice CT Based on 3D Neural Networks With adaptive Initial Weights”, International Conference on Intelligent Computation Technology and Automation, 2010.
18. Lung Cancer Database, Available at Available at:https://eddie.via.cornell.edu/cgi-in/datac/ signon.cgi. Accessed April 2018.
19. Wu Suiyuan and Wang Junfeng,” Pulmonary Nodules 3D Detection on Serial CT Scans”, Third Global Congress on Intelligent Systems, 978-0-7695-4860-9/12, IEEE, 2012.
20. Jibi Johna and M G Mini, 'Multilevel Thresholding Based Segmentation and Feature', International Conference on Emerging Trends in Engineering, Science and Technology (ICETEST), 957 – 963, 2016.
21. Michela Antonelli, Beatrice Lazzerini, Francesco Marcelloni, ”Segmentation and reconstruction of the lung volume in CT images”, ACM Symposium on Applied Computing, 2005.
22. 'Lung Nodule Detection UsingImage Segmentation Methods', International Journal of Advanced Research in Electronics and Communication Engineering (IJARECE), Volume 6, Issue 7, July 2017.
23. Mai Mabrouk, Ayat Karrar and Amr Sharawy, “Computer-Aided Detection of Large Lung Nodules using Chest Computer Tomography Images”, International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868, Vol.3– No.9, August 2012.
24. Francesco Banterle, Massimiliano Corsini, Paolo Cignoni and Roberto Scopigno,'A Low-Memory, Straightforward and Fast Bilateral Filter Through Subsampling in Spatial Domain', COMPUTER GRAPHICS forum, Volume 31, number 1, pp. 19–32, 2012.
25. Sweety Deswal, Shailender Gupta and Bharat Bhushan,'A Survey of Various Bilateral Filtering Techniques', International Journal of Signal Processing, Image Processing and Pattern Recognition, Vol.8, No.3, pp.105-120, 2015.
26. Mai Mabrouk1, Ayat Karrar and Amr Sharawy, 'Support Vector Machine Based Computer-Aided Diagnosis System for Large Lung Nodules Classification', Journal of Medical Imaging and Health Informatics, Vol. 3, 214–220, 2013.
27. Leo Grady and Marie Pierre Jolly, 'Segmentation from a Box', IEEE International Conference on Computer Vision, ICCV, November 6-13, 2011.
28. Kumar Abhinav, Jaideep Singh Chauhan and Debasis Sarkar,' Image Segmentation of Multi-Shaped Overlapping Objects', International Conference on Computer Vision Theory and Applications, 2018
29. Ritesh Kumar Pandey, S. S. Mathurkar, ' A Review on Morphological Filter and its Implementation', International Journal of Science and Research (IJSR), 2015.
30. Antropova N, Abe H and Giger ML,'Use of clinical MRI maximum intensity projections for improved breast lesion classification with deep convolutional neural networks', J Med Imaging (Bellingham), Jan;5(1):014503, 2018.
31. Qiu shi, Wen desheng, Cui ying and Feng jun, 'Lung Nodules Detection in CT Images Using Gestalt-Based Algorithm', Chinese Journal of Electronics, Vol.25, No.4, July 2016.
32. Natalia M. Jenuwine, Sunny N. Mahesh, Jacob D. Furst and Daniela S. Raicu,' Lung nodule detection from CT scans using 3D convolutional neural networks without candidate selection', Medical Imaging: Computer-Aided Diagnosis; Vol. 10575, 2018.
33. Sarah Soltaninejad, Mohsen Keshani, Farshad Tajeripour,” Lung Nodule Detection by KNN Classifier and Active Contour Modelling and 3D Visualization”, The 16th CSI International Symposium on Artificial Intelligence and Signal Processing (AISP), IEEE, 2012.
34. Mohd Shafry Mohd Rahim, Alireza Norouzi, Amjad Rehman and Tanzila Saba, '3D bones segmentation based on CT images visualization', Biomedical Research, Vol. 28, Issue 8,2017.
35. Gustavo Ch´avez, Alyn Rockwood,” Marching Surfaces: Isosurface Approximation using G1 Multi-Sided Surfaces”, arXiv, Volume xx (200y), Number z, pp. 1–8, 2015
36. William Lorensen, Harvey_E_Cline, “Marching Cubes: A High-Resolution 3D Surface Construction Algorithm”, ACM SIGGRAPH Computer Graphics, (Impact Factor: 1.13). 08/1987; 21(4):163—169, 2014.
37. Qaisar Abbas,' Lung-Deep: A Computerized Tool for Detection of Lung Nodule Patterns using Deep Learning Algorithms, International Journal of Advanced Computer Science and Applications, Vol. 8, No. 10, 2017.
38. Patrice Monkam, Shouliang Qi, Mingjie Xu, Fangfang Han, Xinzhuo Zhao and Wei Qian,' CNN models discriminating between pulmonary micronodules and nonnodules from CT images', BioMed Eng OnLine, 17:96, 2018.