IRJET- A Survey on Soft Computing Techniques for Early Detection of Breast Cancer
0 likes27 views
This document summarizes several papers on using soft computing techniques for early detection of breast cancer. It discusses how techniques like K-nearest neighbors (KNN), support vector machines (SVM), fuzzy C-means, and artificial neural networks (ANN) have been applied to mammogram images and breast cancer data to classify tumors and detect cancer at earlier stages. Feature extraction methods like gray level co-occurrence matrix (GLCM) have also been used to identify textures that can help with classification. The papers find that combinations of these soft computing and image processing approaches can accurately classify cancer and detect it earlier than traditional methods, helping to improve treatment outcomes for breast cancer patients.
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 8137
A Survey on Soft Computing Techniques for Early Detection of Breast
Cancer
Ashwini K C1, Jenifer Jacob2, Sanjana Srinath3, Vignesh Sharma S4
1,2,3,4 Students, Dept. of Computer Science and Engineering, Vidya Vikas Institute of Engineering and
Technology, Mysuru, Karnataka, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract – Breast cancer is most common among women
and is said to be the second major cause of death among
women. For every 19 seconds, somewhere around the world a
case of breast cancer is diagnosed among women. Report says
that for 74 seconds somewhere in the world a women dies
from breast cancer. Most effective way to reduce the death
rate is to detect at an early stage. By detecting at an early
stage proper treatment can be given to save the life of
patients. Accurate classification plays an important role in
medical diagnosis. Soft computing approaches are gaining
importance because of their classification performance in
diagnosing the disease. The goal of this survey paper is to
identify the current state of research in breast cancer and to
summarize the different soft computing techniques that helps
in identification and classification.
Key Words: KNN, SVM, Fuzzy C-means, ANN, GLCM, ROI
1. INTRODUCTION
Breast cancer starts when the cells in the breast begin to
grow out of control and are said to form a tumour. These
tumour can be classified into cancerous or non-cancerous.
According to the worldwide survey that was conducted in
the year 2010 it is estimated that more than 1.5 million
breast cancer cases occurred in women and among the 23%
of breast cancer detected 14% ofdeathisreported[1].There
are 12% of chances that a women might develop breast
cancer during her life time. Regardless of age and their
family history every women is at a risk of developing breast
cancer. Early detection and effective treatment is the only
way to reduce the death rate due to breast cancer. In this
paper various machine learning algorithms and image
processing techniques that are available for detecting and
classifying the tumour cells at an early stage has been
discussed in detail. Different combinations of these will give
different outputs and varying accuracy. Thus making it
easier to select the most accurate algorithms.
2. LITERATURE SURVEY
A novel methodology proposed in the paper [2] to detect
breast asymmetry and calcification cancer cells using
combination of different highly efficienttechnologyofdigital
image processing which are not yet implemented. In this
paper it is basically noted that breast asymmetry is one of
the major method to identify the suspicious region in the
breast and in the segmentation process, the Otsu’s
thresholding algorithm was being applied using its
methodologies, which will segment the micro calcifications
from the image with the highest accuracy. They have used
the K-Nearest neighbor clustering algorithm which will
group the identified micro calcifications into clusters. The
next stage is feature extractions which includes brightness,
contrast, size, shape and textures which can be obtained
from previous micro calcification clusters. The features that
were obtained werethenextractedusingtheSobel algorithm
and the extracted features were then classified using the
Bayes classifier. Thus the major intention of this paper is to
identify the breast asymmetry in the earliest stage as
possible with highest possible accuracy.
The major intention of this paper [3] is to classify the
medical data with efficient and more accurate processing
with simple and faster classification algorithms. The
proposed model in this paper involves the steps where in
initially the breast cancer dataset wastakenasinputandif at
all few values were missing in the same then those values
were handled in the next coming step. Using the K-means
algorithm the clustering of the data set was performed as it
is well known for its simplicity and hidden pattern and data
recognition. Once the clustering was done, the clusters were
classified as cluster1 and cluster2 where in each of it is
carried out with the process of feature reduction using the
FRFS(Fussy Rough Feature Selection) which is efficient in
handling noisy, discrete and continuous data with no loss.
The clustered data was then merged, reduced and classified
using the D-KNN algorithm (Discernibility K-nearest
neighbor) classifier which is known for itshigh accuracyand
better classification of the data set. At the end, after all these
classifications the overall performance of the proposed
system was evaluated.
This paper [4] makes use of K-means and Fuzzy C-means
algorithm for detecting the cancer tumour mass and micro
calcification. K-meansclusteringalgorithmisusedtoidentify
the hidden parts and Fuzzy C-Mean algorithmismainlyused
in pattern recognition and it allows one piece of data to be
present in two or more cluster. Gray level transformation
used in this paper makes use of logarithmic and power law
as a contrast stretching methods which is used to highlight
the detail in dark or washed out images. Gray level
Transformation technique has been used in this paper to
obtain negative of an image with gray levels in the range 0to](https://image.slidesharecdn.com/irjet-v6i31009-190926092206/85/IRJET-A-Survey-on-Soft-Computing-Techniques-for-Early-Detection-of-Breast-Cancer-1-320.jpg)
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 8138
L-1 and this type of processing is well suited for enhancing
the white or gray detail embedded in the dark region of an
image especially when the dark areas are dominant in size.
Log transformation is used in the paper to map a narrow
range of gray level values in the input image into a wider
range of output levels andits transformation.Thispaperalso
uses power law, one of the image enhancement technique
which is useful for general purpose contrast manipulation
and thresholding is used to produce an image with higher
contrast than the original image by darkening the levels
below and brightening the levels above a threshold value.
Robert edge detection, Prewitt edge detection, Sobel edge
detection are some of the different edge detection
techniques used in this paper for identifying the edges from
the original image. Thus, the major intention of this paper
was to identify the presence of breast cancer mass and
calcification in mammograms using image processing
functions, K-means and Fuzzy C-means clustering for clear
identification of clusters.
The technique involved in this paper [5] is the concept of
fractal dimension measurement along with the ultrasound
medicine characteristic and also grouping of the similar
objects together in order to obtain knowledge from those
data. This is done using the K-mean algorithm which
provides the capability to discover new information using
the existing data. The Box Counting method is applied to
establish fractal dimensions for which a software called as
HarFA was built. To calculate fractal dimension on an
ultrasound of breast filtering, 4 types of filters were used
which are: Max filter- this is known to process the low level
image and vision. Min filter- This plays a major part inimage
processing and vision. Geometric filter- This replaces the
gray level by taking thesurroundingdetail andnoise.Median
filter- These contain useful algebraic properties which are
easier to understand with Fourier transform and other
statistical properties. While analysing the tumours, the ROI
(Region of Interest) of an ultrasound was considered and
then the above filters were used to obtain the quality image.
The K-mean algorithm is used to group all these tumours
which are of same dimensions using the ultrasound image.
The advantage of fractal analysis and clustering is majorly
quoted in this paper.
In the referred paper [6], the proposed system for
classification of mammogram image is basedonGLCM(Gray
Level Co-occurrence Matrix). The texture features are
extracted using GLCM of ROI. The system is divided into five
stages to classify mammogram images. First is Dataset
collection where in the mammogram images are obtained
from MIAS dataset which contains normal and abnormal
images. Second stage is ROI extraction process, the original
mammogram images have different types of noise and
artifacts in background, so these unwanted elementsshould
be removed from the image. Third stage is Pre-processing,
which is further divided into twomorestepsnamelyfiltering
where median filter is used for removing noise and second
step is Enhancement where CLAHE (Contrast Limited
Adaptive Histogram Equalization) is used to improve the
appearance of image. In the fourth step we can see Feature
extraction from GLCM, weknowthatprocessingoflargedata
is time consuming and less effective in Digital Image
Processing. So for reducing time, the input data is
transformed into reduced set if feature vector which has
relevant information. This transformation process is called
Feature extraction. Last and the final stageisclassificationof
mammogram image into normal or abnormal images. Here
in this paper classification is done with the help of KNN
algorithm and compared with SVM and ANN. KNN is a
supervised learning algorithm wheredifferentk valuesgives
different results. In this papertheyhaveconsidered1NN and
3NN. The accuracy rate for normal and abnormal
classification as given by 3NN is 96% when compared with
other classifiers.
This paper [7] makes use of KNN algorithm which is a
pattern recognition technique, which is used for
Classification and Regression. In both the cases, the input
will contain K closest training examples in the feature space.
The output that is obtained from this method is dependent
on whether KNN is used for ClassificationorRegression.The
result of this modelling contains accuracy of 0.99. Support-
Vector machines are supervised learning models that are
associated with the learning algorithm, where the data is
used for Classification and Regression analysis. The
clustering algorithm that provides advancement to the
support vector machines is called support vector clustering.
The result of this modelling contains accuracy of 1. Logistic
Regression is a predictive analysis technique. This model is
used to describe the data and toexplaintherelation between
one dependent binary variable and one or more nominal.
The result of this modelling contains accuracy of
0.98.Decision tree learning is a method commonly used in
data mining. The objective of this classification is to develop
a model that predicts the value of a target variable based on
several input variables. The result of this modellingcontains
accuracy of 0.98. Artificial Neural Network isa mathematical
model that defines a function f: X Y. For ANN algorithm,
working with activation and learning rate parameters are
carried out. The result of this model contains an accuracy of
0.98.Greedy Search algorithm uses a heuristic for making
optimal choices at each stage to find a global optimum.
According to the results of all thesemodellingthealgorithms
SVM and KNN are the best for the breast cancer prediction.
In this paper [8] they have discussed about different data
mining modelling techniques and algorithms like Neural
Network, Decision tree, Support vector machines and
Bayesian network. Decisiontreesare easytounderstandand
interpret. A decision tree consists of root, internal and leaf
nodes. With the help of a decision tree unknown data
records can be classified. Decision tree can be applied in
medical field for predicting the death rate, feature selection](https://image.slidesharecdn.com/irjet-v6i31009-190926092206/85/IRJET-A-Survey-on-Soft-Computing-Techniques-for-Early-Detection-of-Breast-Cancer-2-320.jpg)
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 8139
to improve the classification accuracy etc.,decisiontreeisan
effective means of constructing a model to predictthe risk of
mortality. Neural network has a capability to learn a set of
data and construct weight matrixesto representthelearning
patterns. Neural network is a computational representation
that takes a sequence of numbers input as and outputs
another sequence of numbers. These computational nodes
are connected in several layers to achieve high accuracy.
ANN (Artificial Neural network) can be used as a tool to
make decision in diagnosing the various diseases. From this
paper we get to know that RBF neural network has resulted
in 97% succession rate for classifyingthe datasetinto benign
and malignant. Support vector machine is a classification
algorithm and a powerful pattern recognizer.Thefocusison
finding the hyper plane that separates them into cancerous
and non-cancerous cells. SVM and ANN for prediction and
detection of breast cancer is highly accurate than by the
humans. The efficiency of SVM is nearly 97% and the
efficiency of manual detection is around 85%. Bayesian
belief network is well suited for dealing with the incomplete
data. It can be used as both predictive and descriptive
models. In prediction we try to infer tasks such as posterior
probability, diagnostic reasoning, relevance analysis and
classification. As a descriptive tool they represent the
dependence/independence relationships among the
variables. These are some of the data mining model that can
be used for detecting and classifying the tumour cells.
The major intention of this paper [9] is to detect the tumour
in breast from the mammogram images using the various
image processing techniques. In the proposed system,
initially screening and diagnostic techniques are used to get
image of the breast by the digital mammogram. The
proposed detection algorithm is then initiated which
involves image segmentation, image binarization, image
thinning, image triangulation and Euclidean distance
transformation. In the image segmentation process, the
image can be represented in simpler andmeaningful wayfor
which the variance method is used to perform thresholding.
Binarization does the conversion of image into binary form
which is required to know the threshold value. In this paper,
threshold equation is being used tocalculateforegroundand
background information of the binary image. Thinningisthe
next process which is basically used to get a preferredimage
pattern to focus on the region of interest. This is done using
the parallel thinning algorithm which has two iterations to
carry out the same. Next, is the triangulation process which
is the post processing step that is used to clearly identify the
edge of tumour region. A new approach is being used called
the algorithm of Delaunay Triangulation which uses both
iterative and non-iterative methodologies. EDT is the last
step which usually has a lengthy calculation for which this
paper suggests a scan recursive algorithm thatonlyrequires
two scans of images whose coordinates are analyzed. These
processes overall will yield the result where in which the
features obtained are used to find the cancer cell area. The
paper has estimated its overall performance including the
accuracy, sensitivity etc.
3. CONCLUSION
In this survey paper we have discussed the various soft
computing techniques that are available for detecting and
classifying the tumour cell at an early stage. Choosing the
best algorithm for detecting the disease plays a major role.
Combination of algorithms can be used for diseasedetection
and classification. Differentcombinationofmachinelearning
algorithms will give a different output and the accuracy
might be varied. This paper contains variousalgorithmsthat
can be used in the medical field for breast cancer detection
and will be helpful in making the right decision in selecting
the algorithms.
REFERENCES
[1] Chandresh Arya, Ritu Tiwari “Expert System for Breast
Cancer Diagnosis: A Survey” 2016 International Conference
on Computer Communication and Informatics.
[2] Sangeetha R Murthy K andDr.Srikanth“Anovel approach
for Detection of Breast Cancer at an early stage by
identification of breast Asymmetry and Micro Calcification
Cancer cells using Digital Image Processing Technology”
2017 2nd International Conference for Convergence in
Technology.
[3] Ibrahim M. EL-Hasnony, Hazem M. EL-Bakry,andAhmed
A. Saleh, “ClassificationofBreastCancerusingSoftcomputing
Techniques” International Journal of Electronics and
Information Engineering.
[4] Nalini Singh, Ambarish G Mohapatra, Biranchi Narayan
Rath, and Guru Kalyan Kanungo “GUI Based Automatic
Breast Cancer Mass and Calcification Detection in
Mammogram Images Using K- means and Fuzzy C-means
Methods” International Journal Of machine Learning and
Computing.
[5] Simona Moldovanu, Luminita Moraru,” Mass Detection
and ClassificationinBreastUltrasoundImageUsingK-means
Clustering Algorithm”, IEEE Paper.
[6] Ranjit Biswas, Abhijit Nath, Sudipta Roy, “Mammogram
Classification using Gray-Level Co-occurrence Matrix for
Diagnosis of Breast Cancer”, 2016 International Conference
on Micro-Electronics and Telecommunication Engineering.
[7] Meraryslan Meraliyev, Meirambek Zhaparov,Kamalkhan
Artykbayev “Choosing Best Machine Learning Algorithm for](https://image.slidesharecdn.com/irjet-v6i31009-190926092206/85/IRJET-A-Survey-on-Soft-Computing-Techniques-for-Early-Detection-of-Breast-Cancer-3-320.jpg)
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 8140
Breast Cancer Prediction”, International Journal ofAdvances
in Science Engineering and Technology.
[8] S. Kharya, D. Dubey, S. Soni,” Predictive Machinelearning
Techniques for breast Cancer Detection”, International
Journal of Computer science and Information Technologies.
[9] Angayarkanni.N, Kumar.D and Arunachalam.G,”The
Application of Image Processing Techniques for Detection
and Classification of Cancerous Tissue in Digital
Mammograms”, Journal of Pharmaceutical Sciences and
Research.](https://image.slidesharecdn.com/irjet-v6i31009-190926092206/85/IRJET-A-Survey-on-Soft-Computing-Techniques-for-Early-Detection-of-Breast-Cancer-4-320.jpg)
IRJET- A Survey on Soft Computing Techniques for Early Detection of Breast Cancer
- 1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 8137 A Survey on Soft Computing Techniques for Early Detection of Breast Cancer Ashwini K C1, Jenifer Jacob2, Sanjana Srinath3, Vignesh Sharma S4 1,2,3,4 Students, Dept. of Computer Science and Engineering, Vidya Vikas Institute of Engineering and Technology, Mysuru, Karnataka, India ---------------------------------------------------------------------***--------------------------------------------------------------------- Abstract – Breast cancer is most common among women and is said to be the second major cause of death among women. For every 19 seconds, somewhere around the world a case of breast cancer is diagnosed among women. Report says that for 74 seconds somewhere in the world a women dies from breast cancer. Most effective way to reduce the death rate is to detect at an early stage. By detecting at an early stage proper treatment can be given to save the life of patients. Accurate classification plays an important role in medical diagnosis. Soft computing approaches are gaining importance because of their classification performance in diagnosing the disease. The goal of this survey paper is to identify the current state of research in breast cancer and to summarize the different soft computing techniques that helps in identification and classification. Key Words: KNN, SVM, Fuzzy C-means, ANN, GLCM, ROI 1. INTRODUCTION Breast cancer starts when the cells in the breast begin to grow out of control and are said to form a tumour. These tumour can be classified into cancerous or non-cancerous. According to the worldwide survey that was conducted in the year 2010 it is estimated that more than 1.5 million breast cancer cases occurred in women and among the 23% of breast cancer detected 14% ofdeathisreported[1].There are 12% of chances that a women might develop breast cancer during her life time. Regardless of age and their family history every women is at a risk of developing breast cancer. Early detection and effective treatment is the only way to reduce the death rate due to breast cancer. In this paper various machine learning algorithms and image processing techniques that are available for detecting and classifying the tumour cells at an early stage has been discussed in detail. Different combinations of these will give different outputs and varying accuracy. Thus making it easier to select the most accurate algorithms. 2. LITERATURE SURVEY A novel methodology proposed in the paper [2] to detect breast asymmetry and calcification cancer cells using combination of different highly efficienttechnologyofdigital image processing which are not yet implemented. In this paper it is basically noted that breast asymmetry is one of the major method to identify the suspicious region in the breast and in the segmentation process, the Otsu’s thresholding algorithm was being applied using its methodologies, which will segment the micro calcifications from the image with the highest accuracy. They have used the K-Nearest neighbor clustering algorithm which will group the identified micro calcifications into clusters. The next stage is feature extractions which includes brightness, contrast, size, shape and textures which can be obtained from previous micro calcification clusters. The features that were obtained werethenextractedusingtheSobel algorithm and the extracted features were then classified using the Bayes classifier. Thus the major intention of this paper is to identify the breast asymmetry in the earliest stage as possible with highest possible accuracy. The major intention of this paper [3] is to classify the medical data with efficient and more accurate processing with simple and faster classification algorithms. The proposed model in this paper involves the steps where in initially the breast cancer dataset wastakenasinputandif at all few values were missing in the same then those values were handled in the next coming step. Using the K-means algorithm the clustering of the data set was performed as it is well known for its simplicity and hidden pattern and data recognition. Once the clustering was done, the clusters were classified as cluster1 and cluster2 where in each of it is carried out with the process of feature reduction using the FRFS(Fussy Rough Feature Selection) which is efficient in handling noisy, discrete and continuous data with no loss. The clustered data was then merged, reduced and classified using the D-KNN algorithm (Discernibility K-nearest neighbor) classifier which is known for itshigh accuracyand better classification of the data set. At the end, after all these classifications the overall performance of the proposed system was evaluated. This paper [4] makes use of K-means and Fuzzy C-means algorithm for detecting the cancer tumour mass and micro calcification. K-meansclusteringalgorithmisusedtoidentify the hidden parts and Fuzzy C-Mean algorithmismainlyused in pattern recognition and it allows one piece of data to be present in two or more cluster. Gray level transformation used in this paper makes use of logarithmic and power law as a contrast stretching methods which is used to highlight the detail in dark or washed out images. Gray level Transformation technique has been used in this paper to obtain negative of an image with gray levels in the range 0to
- 2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 8138 L-1 and this type of processing is well suited for enhancing the white or gray detail embedded in the dark region of an image especially when the dark areas are dominant in size. Log transformation is used in the paper to map a narrow range of gray level values in the input image into a wider range of output levels andits transformation.Thispaperalso uses power law, one of the image enhancement technique which is useful for general purpose contrast manipulation and thresholding is used to produce an image with higher contrast than the original image by darkening the levels below and brightening the levels above a threshold value. Robert edge detection, Prewitt edge detection, Sobel edge detection are some of the different edge detection techniques used in this paper for identifying the edges from the original image. Thus, the major intention of this paper was to identify the presence of breast cancer mass and calcification in mammograms using image processing functions, K-means and Fuzzy C-means clustering for clear identification of clusters. The technique involved in this paper [5] is the concept of fractal dimension measurement along with the ultrasound medicine characteristic and also grouping of the similar objects together in order to obtain knowledge from those data. This is done using the K-mean algorithm which provides the capability to discover new information using the existing data. The Box Counting method is applied to establish fractal dimensions for which a software called as HarFA was built. To calculate fractal dimension on an ultrasound of breast filtering, 4 types of filters were used which are: Max filter- this is known to process the low level image and vision. Min filter- This plays a major part inimage processing and vision. Geometric filter- This replaces the gray level by taking thesurroundingdetail andnoise.Median filter- These contain useful algebraic properties which are easier to understand with Fourier transform and other statistical properties. While analysing the tumours, the ROI (Region of Interest) of an ultrasound was considered and then the above filters were used to obtain the quality image. The K-mean algorithm is used to group all these tumours which are of same dimensions using the ultrasound image. The advantage of fractal analysis and clustering is majorly quoted in this paper. In the referred paper [6], the proposed system for classification of mammogram image is basedonGLCM(Gray Level Co-occurrence Matrix). The texture features are extracted using GLCM of ROI. The system is divided into five stages to classify mammogram images. First is Dataset collection where in the mammogram images are obtained from MIAS dataset which contains normal and abnormal images. Second stage is ROI extraction process, the original mammogram images have different types of noise and artifacts in background, so these unwanted elementsshould be removed from the image. Third stage is Pre-processing, which is further divided into twomorestepsnamelyfiltering where median filter is used for removing noise and second step is Enhancement where CLAHE (Contrast Limited Adaptive Histogram Equalization) is used to improve the appearance of image. In the fourth step we can see Feature extraction from GLCM, weknowthatprocessingoflargedata is time consuming and less effective in Digital Image Processing. So for reducing time, the input data is transformed into reduced set if feature vector which has relevant information. This transformation process is called Feature extraction. Last and the final stageisclassificationof mammogram image into normal or abnormal images. Here in this paper classification is done with the help of KNN algorithm and compared with SVM and ANN. KNN is a supervised learning algorithm wheredifferentk valuesgives different results. In this papertheyhaveconsidered1NN and 3NN. The accuracy rate for normal and abnormal classification as given by 3NN is 96% when compared with other classifiers. This paper [7] makes use of KNN algorithm which is a pattern recognition technique, which is used for Classification and Regression. In both the cases, the input will contain K closest training examples in the feature space. The output that is obtained from this method is dependent on whether KNN is used for ClassificationorRegression.The result of this modelling contains accuracy of 0.99. Support- Vector machines are supervised learning models that are associated with the learning algorithm, where the data is used for Classification and Regression analysis. The clustering algorithm that provides advancement to the support vector machines is called support vector clustering. The result of this modelling contains accuracy of 1. Logistic Regression is a predictive analysis technique. This model is used to describe the data and toexplaintherelation between one dependent binary variable and one or more nominal. The result of this modelling contains accuracy of 0.98.Decision tree learning is a method commonly used in data mining. The objective of this classification is to develop a model that predicts the value of a target variable based on several input variables. The result of this modellingcontains accuracy of 0.98. Artificial Neural Network isa mathematical model that defines a function f: X Y. For ANN algorithm, working with activation and learning rate parameters are carried out. The result of this model contains an accuracy of 0.98.Greedy Search algorithm uses a heuristic for making optimal choices at each stage to find a global optimum. According to the results of all thesemodellingthealgorithms SVM and KNN are the best for the breast cancer prediction. In this paper [8] they have discussed about different data mining modelling techniques and algorithms like Neural Network, Decision tree, Support vector machines and Bayesian network. Decisiontreesare easytounderstandand interpret. A decision tree consists of root, internal and leaf nodes. With the help of a decision tree unknown data records can be classified. Decision tree can be applied in medical field for predicting the death rate, feature selection
- 3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 8139 to improve the classification accuracy etc.,decisiontreeisan effective means of constructing a model to predictthe risk of mortality. Neural network has a capability to learn a set of data and construct weight matrixesto representthelearning patterns. Neural network is a computational representation that takes a sequence of numbers input as and outputs another sequence of numbers. These computational nodes are connected in several layers to achieve high accuracy. ANN (Artificial Neural network) can be used as a tool to make decision in diagnosing the various diseases. From this paper we get to know that RBF neural network has resulted in 97% succession rate for classifyingthe datasetinto benign and malignant. Support vector machine is a classification algorithm and a powerful pattern recognizer.Thefocusison finding the hyper plane that separates them into cancerous and non-cancerous cells. SVM and ANN for prediction and detection of breast cancer is highly accurate than by the humans. The efficiency of SVM is nearly 97% and the efficiency of manual detection is around 85%. Bayesian belief network is well suited for dealing with the incomplete data. It can be used as both predictive and descriptive models. In prediction we try to infer tasks such as posterior probability, diagnostic reasoning, relevance analysis and classification. As a descriptive tool they represent the dependence/independence relationships among the variables. These are some of the data mining model that can be used for detecting and classifying the tumour cells. The major intention of this paper [9] is to detect the tumour in breast from the mammogram images using the various image processing techniques. In the proposed system, initially screening and diagnostic techniques are used to get image of the breast by the digital mammogram. The proposed detection algorithm is then initiated which involves image segmentation, image binarization, image thinning, image triangulation and Euclidean distance transformation. In the image segmentation process, the image can be represented in simpler andmeaningful wayfor which the variance method is used to perform thresholding. Binarization does the conversion of image into binary form which is required to know the threshold value. In this paper, threshold equation is being used tocalculateforegroundand background information of the binary image. Thinningisthe next process which is basically used to get a preferredimage pattern to focus on the region of interest. This is done using the parallel thinning algorithm which has two iterations to carry out the same. Next, is the triangulation process which is the post processing step that is used to clearly identify the edge of tumour region. A new approach is being used called the algorithm of Delaunay Triangulation which uses both iterative and non-iterative methodologies. EDT is the last step which usually has a lengthy calculation for which this paper suggests a scan recursive algorithm thatonlyrequires two scans of images whose coordinates are analyzed. These processes overall will yield the result where in which the features obtained are used to find the cancer cell area. The paper has estimated its overall performance including the accuracy, sensitivity etc. 3. CONCLUSION In this survey paper we have discussed the various soft computing techniques that are available for detecting and classifying the tumour cell at an early stage. Choosing the best algorithm for detecting the disease plays a major role. Combination of algorithms can be used for diseasedetection and classification. Differentcombinationofmachinelearning algorithms will give a different output and the accuracy might be varied. This paper contains variousalgorithmsthat can be used in the medical field for breast cancer detection and will be helpful in making the right decision in selecting the algorithms. REFERENCES [1] Chandresh Arya, Ritu Tiwari “Expert System for Breast Cancer Diagnosis: A Survey” 2016 International Conference on Computer Communication and Informatics. [2] Sangeetha R Murthy K andDr.Srikanth“Anovel approach for Detection of Breast Cancer at an early stage by identification of breast Asymmetry and Micro Calcification Cancer cells using Digital Image Processing Technology” 2017 2nd International Conference for Convergence in Technology. [3] Ibrahim M. EL-Hasnony, Hazem M. EL-Bakry,andAhmed A. Saleh, “ClassificationofBreastCancerusingSoftcomputing Techniques” International Journal of Electronics and Information Engineering. [4] Nalini Singh, Ambarish G Mohapatra, Biranchi Narayan Rath, and Guru Kalyan Kanungo “GUI Based Automatic Breast Cancer Mass and Calcification Detection in Mammogram Images Using K- means and Fuzzy C-means Methods” International Journal Of machine Learning and Computing. [5] Simona Moldovanu, Luminita Moraru,” Mass Detection and ClassificationinBreastUltrasoundImageUsingK-means Clustering Algorithm”, IEEE Paper. [6] Ranjit Biswas, Abhijit Nath, Sudipta Roy, “Mammogram Classification using Gray-Level Co-occurrence Matrix for Diagnosis of Breast Cancer”, 2016 International Conference on Micro-Electronics and Telecommunication Engineering. [7] Meraryslan Meraliyev, Meirambek Zhaparov,Kamalkhan Artykbayev “Choosing Best Machine Learning Algorithm for
- 4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 8140 Breast Cancer Prediction”, International Journal ofAdvances in Science Engineering and Technology. [8] S. Kharya, D. Dubey, S. Soni,” Predictive Machinelearning Techniques for breast Cancer Detection”, International Journal of Computer science and Information Technologies. [9] Angayarkanni.N, Kumar.D and Arunachalam.G,”The Application of Image Processing Techniques for Detection and Classification of Cancerous Tissue in Digital Mammograms”, Journal of Pharmaceutical Sciences and Research.