Professional Documents
Culture Documents
Abstract - Lung cancer is life-threatening diseases This paper proposes a DCNN based architecture for
and it now affects all people regardless of gender. Precise lung classifying the lung tumor. If the tumor is benign then chances
tumor classification helps to diagnosis lung cancer early, which of cancer could be low and cancer can be initial stage and also
decreases the rate of death of lung cancer, but it is hard to curable. When the tumor becomes malignant then chances of
detect early. Harmless or malignant is known as a lung tumor. cancer to be high. Deep neural networks consisting of multiple
hidden layers such as convolutional layer, pooling layer, and
It is innocuous when the tumor cells were healthy, when the
fully connected layer. Convolutional layer is characterized by
cells are abnormal and can grow hysterically, they are
local connectivity and weight sharing. This layer's nodes are
cancerous cells, and the tumor is in curable. A classifier based grouped into feature maps that share the same weights, weight
on Deep Convolutional Neural Networks (DCNN), which sharing radically reduces the number of network parameters and
classifies the lung tumor composed of different fully connected thus increases efficiency and prevents over fitting. Through
pooling and Convolutional layers. Three architectures were aggregating small sub-sets of values, the pooling layer is used to
defined for DCNN classifier each one is trained with different subsample the previous layer. Max or average pooling
patch size. DCNN is applied to the CT image for classification substitutes input values by maximum or average values to
of benign and malignant lung tumor. The proposed reduce output sensitivity to minor input changes. The
architectures were examined on the LIDC database and cross classification results are generated by a fully connected layer.
checked with other classifiers result such as Artificial Neural II. RELATED WORK
Network Simulation result presents DCNN classifier achieves
better performance. An image segmentation using ANN for detecting the
lung cancer using X-Ray images. Artificial neuron consists of
Keywords: Deep Learning, Convolutional Neural three basic sets of rules, multiplication, summation, and
Network, Tumor, CT images, Benign and Malignant. activation. Initially, Seed fill operation is applied; it performs the
process on binary image background pixels starting from the
I. INTRODUCTION points. This reveals the images that struggle with 60%, 70% and
Recent statistics from the world health institutions 80% of lung cancer. Lung Cancer Prediction [5] is performed
demonstrate that approximately 7.6 million people die every using the genomic features of the Oriented Gradient Histogram
year in the world due to cancer. In regards, it is expected (HOG) all together with Artificial Neural Network. Different
which cancer mortality will continue to rise to around 18 phases were included for implementing the prediction model.
million worldwide in 2030. The main causes of death are Such as pre-processing based on data acquisition and
Lung cancer mortality rate is counting for about 27% of all classification based on extracted features. After feature
deaths from cancer [2]. The death of the lung cancer patients’ extraction, the HOG descriptor is investigated to remove agent
numbers is more-than that of breast cancer. Due to lack of highlights from the images of the genomic arrangements from
literacy towards lung cancer, the possibility of detecting the Voss encoder. To choose the hyperbolic digression sigmoid
disease is harder at the early stage. In other words, at the capacity for the yield layer neurons because it has the capability
time of detection, already the disease reach to its advance to handle either ceaseless qualities or -class binary
stage [3]. The stages of lung cancer can be classified as first
arrangement issues.
and second stage. The First stage refers to tumor presents
only in limited area and the second stage is tumors that To upgrade the structures of the fragmented images,
expand over to various body parts. Deep learning algorithms, decrease the bogus identification and abstain from missing
especially DCNN, have rapidly become a popular knobs joined on the lung walls. After this, intensity based
methodology for processing medical images as well, highlights are determined from every knob. The authors of [7]
including detection of medical problems, Feature Extraction, segment the complete lung and classified from the CT scans
segmentation, and classification. picture using neural networks. The arguments are determined
from the sectioned image using the standard deviation, mean
978-1-7281-3778-0/19/$31.00 ©2019 IEEE
800
2019 International Conference on Computational Intelligence and Knowledge Economy (ICCIKE)
December 11–12, 2019, Amity University Dubai, UAE
statistical arguments, kurtosis, and skewedness. The process cancerous). The shape and size of tumors vary, and benign
of classification is accomplished through feed forward and tumors can look very similar to malignant ones. Furthermore,
feedback neural networks. The feed-forward propagation the presence of blood vessels in the lung makes distinguishing
system provides better classification contrasted with feed- tumors a challenging task, especially on 2D image slices.
forward systems..
A. Introduction to Proposed Classifier:
In designed the CAD framework for lung nodules
The proposed Classifier is divided into four separate
classification with artificial neural networks. It gives
stages. They are the: (1) Tumor detector, which accepts a CT
division of knobs on the projections with neural systems
scan and identifies areas that contain tumor; (2) malignancy
grouping among favorable and dangerous knobs with the
detector, which operates similarly to the tumor detector, but
assistance of ANN. It also provided better accuracy,
further classes tumor as benign (non-cancerous) or malignant
sensitivity and specificity and it used an aggregate of 128 CT
(cancerous); (3) tumor classier, which accepts individual tumor
scans acquired from 47 patients. ANN uses three sections in
volumes and similarly classes them as benign or malignant; and
the creator section: input, hidden and output sections. 22
(4) patient classifier, which accepts features from the
neurons in just one hidden layer were used in the
malignancy detector and tumor classifier and yields the
investigation. A tale aspiratory knob recognition strategy
probability of the patient having tumor.
dependent on Progressive Square utilizing Support Vector
Machine (SVM) grouping is proposed in [9] and it comprises
The tumor detector is used to detect regions that
of three stages. The initial step, split information CT pictures
contain tumor. A malignancy detector provides a class
into square pictures and non-enlightening squares are sifted
probability map over grid cells of each cell containing benign,
through. In the subsequent advance, knob applicants are
malignant, or normal (no tumor). Separate code extracts and
identified. Fragment the articles in the chose square pictures
preprocesses tumor volumes and runs the classier on each,
and change their area to identify the knob. Finally classify
yielding the probability of malignancy for each tumor. Finally,
the nodule applicants into knobs and non-knobs. Highlight
the patient classifier pools features from the tumor classifier and
vectors are removed from the item in the knob applicant
the malignancy detector and produces the probability of the
squares. SVM classifiers are applied to the removed element
patient having lung tumor. Figure 1 illustrates the Lung Tumor
vectors to recognize knobs. This methodology has decreased
Classification Process.
the bogus encouraging points in the knob fundamentally and
it accomplished 95.28% affectability. 1) Preprocessing:
An algorithm which efficiently identifies and Noise in medical images is caused by different sources,
classifies tumor in the CT images. Initially the enhancement including many external causes in the transmission system and
technique is applied by means of Wavelet based Adaptive environmental facts. The noise removal methods have become
Histogram Equalization (WAHE) to correct the contrast an essential factor for accurate detection of lung tumor.
which in turn increases the accuracy of lung tumor
segmentation. Next in the segmentation phase, Level set 2) Deep Convolutional Neural Networks Architecture:
based Cellular Automata (LCA) was applied to extract the The Convolutional Neural Networks model by
tumor region from CT images efficiently. A neural network transforming input image with a linear filter, adding a bias term
based approach presented to detect lung cancer from raw and applying a non-linear function, this model can learn lots of
chest images. These extracted highlights are assumed as the feature maps. Three different Convolutional Neural Networks
inputs the neural network. Further, it is verified whether the architecture each with three different patch sizes was proposed
extracted nodule is a non-malignant or malignant. The here to classify the tumors. A detailed result of Convolutional
training dataset of CT images are processed in three stages to Neural Networks with input patch 16 × 16 size of six layers is
attain more quality and accuracy in the processed illustrated in Figure 2, this work included 2 convolutional
examination. For segmentation purpose Fuzzy C-Means layers,2 max-pooling layers, a fully-connected layer and a
(FCM) technique is used. The three kinds of deep neural softmax classifier.
networks are Stacked Auto encoder (SAE). The trial results The convolutional layer C1 ultimate layer contained
show that the CNN system filed the better execution with an two units completely associated with the layer F6, one
exactness, affectability and explicitness which have the best neuron actuated by softmax relapse created tested with three
outcome among the three systems. Computer Aided different patch sizes 16×16, 24×24 and 32×32. The results of
Detection (CAD) framework for feature portrayal and these models were presented in the next section. As the
classification assume basic jobs in false-positive reduction complexity of the network increase, the execution time also
(FPR). Further, structure a profound convolutional neural increases. The key challenge is to balance the accuracy and the
systems strategy for knob arrangement. execution time. We use the limited feature map, if we increase
III. PROPOSED WORK: the number of feature maps the accuracy will be improved.
The presence of pulmonary tumor, solid clumps of These Convolutional Neural Networks architectures are
tissue that appear in and around the lungs, is a key feature of illustrated in Table 1 was used to implement the Convolutional
lung cancer. These tumors are visible in CT scan images and Neural Networks model.
can be malignant (cancerous) in nature, or benign (not
801
2019 International Conference on Computational Intelligence and Knowledge Economy (ICCIKE)
December 11–12, 2019, Amity University Dubai, UAE
YES
Tumor Detected
Benign Malignant
Softmax Classifier
C2: P2: 6*6
Patch: 16×16 C1:32@14*1 32@12*12
P1: 7*7 Normal
4
Benign
P1: Max Pool of 2*2 and stride 2 (with P1:Avg Pool with 2×2 and stride 2 P1:Max Pool with 2×2 and stride 2 (with
Layer 2
padding) (without padding) padding)
C2 : 32 Feature Map, Size of 3×3 and C2 : 32 Feature Map, Size of 3×3 C2 : 32 Feature Map, Size of 3×3 and 1
Layer 3
1 Stride Convolutional Layer and 1 Stride Convolutional Layer Stride Convolutional Layer
Layer 4 P2: Max Pool of 2*2 and stride 2 P2: Avg Pool with 2×2 and stride 2 P2: 2×2 and stride of 2 Max Pool
C3 : 64 Feature Map, Size of 3×3 and C3 : 64 Feature Map, Size of 3×3 and 1
Layer 5 FullyConnected Layer
1 Stride Convolutional Layer Stride Convolutional Layer
802
2019 International Conference on Computational Intelligence and Knowledge Economy (ICCIKE)
December 11–12, 2019, Amity University Dubai, UAE
16×16 74.68
Classification Precision Recall Dice Score
DCNN Architecture 1 24×24 75.76 Algorithms (%) (%) Coefficient (%)
32×32 70.91
SVM 82.66 81.76 78.69
16×16 71.11
ANN 83.33 82.57 79.71
DCNN Architecture 2 24×24 73.25
32×32 73.16 DCNN(Architecture 1)
85.33 84.79 82.57
(24×24)
16×16 82.42
DCNN Architecture 3 DCNN(Architecture 2)
24×24 82.55 84.51 83.45 81.65
(24×24)
32×32 80.02
803
2019 International Conference on Computational Intelligence and Knowledge Economy (ICCIKE)
December 11–12, 2019, Amity University Dubai, UAE
DCNN(Architecture 3) of the figure that the performs better than the existing
85.79 85.24 83.14
(24×24) algorithm.
REFERENCES
804
2019 International Conference on Computational Intelligence and Knowledge Economy (ICCIKE)
December 11–12, 2019, Amity University Dubai, UAE
805