Volume 6, Issue 3, March – 2021 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Breast Cancer Detection using Deep Learning

Neha Kosare Gayatri Bhoyar Neetu Gyanchandani
Department of Electronics & Prof., Department of Electronics & HOD, Department of Electronics &
Telecommunication Engineering Telecommunication Engineering Telecommunication Engineering
J D College of Engineering & J D College of Engineering & J D College of Engineering &
Management Management Management
Nagpur, India Nagpur, India Nagpur, India

Abstract:- Cancer in breast has emerged as the leading Breаst саnсer is the mоst соmmоn tyрe of саnсer
cause of death among women worldwide. Early detection аnd is the seсоnd leаding саuse оf саnсer-relаted deаths
of breast cancer is very important. CAD (computer- in wоmen in Аsiа, ассоunting fоr 39% оf аll breаst
assisted diagnostics) has emerged as a useful medical саnсer diаgnоsed wоrldwide. The inсidenсe оf
diagnostic tool that is very helpful in the protection of саrсinоmа in Аsiа vаries greаtly асrоss the соntinent
patients by decresing of false positive outcomes and аnd remаins subоrdinаte tо Western соuntries, but the
allowing for rapid diagnosis. Rapid advances in the equitаble соntributiоn оf Аsiа tо breаst саnсer rаtes
development of high-resolution imaging techniques have wоrldwide is inсreаsing rарidly аs sосiаl аnd eсоnоmiс
helped the computer to detect automatic breast cancer. develорment. Hоwever, the deаth tоll frоm inсident
The rapid development of in-depth learning, a family of rаtes is muсh higher in Аsiа thаn in Western соuntries.
machine learning techniques, has stimulated a great deal Mоst Аsiаn соuntries аre lоw- аnd middle-inсоme
of interest in its application to the problems of medical соuntries (LMIСs) where саrсinоmа develорs eаrly аnd
illustrations. Here, we develop an in-depth study lаter, аnd where раtients аre mоre likely tо die frоm
algorithm that can accurately detect breast cancer in the diseаse thаn thоse in Western соuntries. In аdditiоn,
mammograms tests using a “end-to-end” training diаgnоstiс tests, treаtments аnd mitigаtiоn serviсes аre
method that utilizes training data sets with complete inаdequаte fоr mаny Аsiаn LMIСs. In this review, we
clinical definitions or only cancer status (label) of the рresent the соnсeрt оf breаst саnсer risk fасtоrs аnd
whole image. The proposed Computer-Aided Diagnosis diseаses, соntrоl meаsures, аnd саnсer саre in Аsiаn
(CAD) program consists of four parts: mammograms соuntries. According WHO reports, BC is the most common
reconstruction, extraction of characteristic using deep type of cancer in women. For example, incidence rates
precision network, mass detection, and finally mass ranged from 19.3 per 100,000 women in Eastern Africa to
fragmentation using Fully Connected Neural Networks 89.7 per 100,000 women in Western Europe. Current
(FC-NNs). scientific findings suggest that such high diversity may be
followed by differences in lifestyle and urbanization.
Keywords:- Deep Learning, Image Processing, Although early diagnosis is less expensive in developed
Mammogram. countries, it is less likely to be developed in developed
countries, which means that taking precautions does not
I. INTRODUCTION provide a solution. Mammography can be a common
diagnostic procedure that will help to differentiate the
Cancer has emerged as main cause of death among skeletal regions of the chest, following a biopsy of the
women worldwide present in breast. Early detection of cancerous areas to determine whether the skeletal area is
breast cancer (BD) is very important. CAD (computer- good or dangerous. Reviewing a patient's body tissue
assisted diagnostics) has emerged as a useful medical samples by a pathologist is a common way to diagnose many
diagnostic tool that is very helpful in the protection of diseases, especially cancers such as breast cancer. However,
patients by decreasing the rate of false positive outcomes reviewing samples is effective and time consuming, which
and allowing for rapid diagnosis. Rapid advances in the can delay decision-making.
development of high-resolution imaging techniques have
helped the computer to detect automatic breast cancer. The  Reviewing pathology slides is a very difficult task. In
rapid development of in-depth learning, a family of machine some cases an agreement for the diagnosis of certain
learning techniques, has stimulated a great deal of interest in types of breast cancer can be as low as 48%.
its application to the problems of medical illustrations. Here,  Difficulty in diagnosing diseases by gynecologists is
we develop an in-depth study algorithm that can accurately unavoidable because pathologists need to review all
detect breast cancer in mammograms tests using a “end-to- slides per patient where each slide is 10+ pixels of giga
end” training method that utilizes training data sets with when digitized at 40X magnification.
complete clinical definitions or only cancer status (label) of  Current medical diagnostic efforts are not intended for a
the whole image. gynecologist with limited experience. The Pathologist
may not understand the terms that define AI or the
statistics generated by AI. It is possible that the

IJISRT21MAR421 www.ijisrt.com 1134

Volume 6, Issue 3, March – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
pathologist will not be able to properly interpret the efficacy of the findings using EMD-based features and the
computerized report. With such limitations, collaboration most commonly used PCA features using clinical trial data
with AI instead can delay decision-making. collected over an eight-month period including a statistical
claim method. We noted that the use of EMD-based features
II. LITERATURE SURVEY leads to significantly improved performance compared to
PCA-based features. Further line improvement can be
In this paper they proposed an automated framework achieved by using a combination of EMD and PCA features
for BC image editing built using deep neural network [12].
technologies. Instead of using immature images we have
used the features of Tamura, as they provide text details. As Saira Charan, Muhammad Jaleed Khan, Khurram
an in-depth study tool we used an unsecured Boltzmann Khurshid This study used neural Convolution networks in
machine that has four layers and is guided by a controlled mammograms to find common and unusual mammograms.
retrieval process. In retrospect, scaled conjugate gradient This in-depth learning method is applied to the
techniques were used. We performed our tests on the Break mammograms MIAS dataset by extracting features from
His dataset and obtained 88.7%, 85.3%, 88.6% and 88.4% randomly divided classes into standard classes. Various filter
with the accuracy of the 40X, 100X, 200X and 400X factor sizes and repair techniques have been applied to real data to
magnification data, respectively. Most of the tests on Break. remove audio objects that can reduce the accuracy of the
His database judged performance on the basis of accuracy; entire network. It was also noted that appropriate separation
however, in this paper we have also considered the prices of is mandatory for the removal and separation of the active
TPR, FPR and a detailed description of the ROC curves. The element. Disguise and classification based on morphological
functionality of the error as an Epoch function is also functioning have greatly improved the effects of
explained in detail. This chapter shows that the RBN method classification.
is very helpful in examining the picture of breast cancer
automatically. However, in the future the combination of III. PROPOSED SYSTEM
CNN and RBM will improve the performance of the
division [8]. A framework for the detection and The cancerous growth is caused by an abnormal
classification of breast cancer. The published results confirm growth of cells, which attack the surrounding tissue in the
that in-depth reading can bring about significant human body. There are two stages of abscesses, dangerous
improvements in the line in terms of hand-made features. and malignant, and in the absence of a tumor in the chest it
While the presented method gains the points obtained, it can is considered normal. Dangerous tumor cells are cancer-free
benefit from further development, which is possible by 1) cells that only grow locally and cannot spread by invasion.
customizing in-depth models; and 2) mixing several deep While the dead tissue is cancerous cells and has the ability to
structures to increase performance. Another guideline to be multiply uncontrollably, they spread to different parts of the
made is to adopt practical learning in order to increase body and attack the surrounding tissues.
segregation scores. Finally, adapting to domains is one of
the ways research can bring about significant improvements.

Predicting breast cancer is very important in the area of

Medicare and Biomedical. The work focused on creating a
segment that aims to predict the most serious cancer known
as breast cancer. Breast cancer is an extremely dangerous
disease that causes many deaths among girls worldwide.
Therefore, the diagnosis of this cancer can save many very
important lives. We have developed a model that predicts
breast cancer based on Support [9]

SVM and Neighbors in K-Eduzane. SVM has been

used by Python to be more effective in classifying diagnostic
data set into two categories by looking at cancer risk. We Fig 1:- System structure
maintain 99.68% accuracy in SVM in the training phase.
The proposed model will be of great benefit to medical staff You Only Look Once (YOLO) is one of the most
and the general public. The distinction obtained by advanced learning methods. It can detect and separate
surveillance machine learning strategies will be very helpful objects in all images at once. Unlike previous detection
in the field of medical problems and appropriate techniques that use differentiation in multiple image regions,
diagnosis[11]. YOLO uses a single convolutional neural network
throughout the image. This approach separates the
The authour proposed and evaluated the use of an installation image into sub-regions and predicts multiple
EMD-based extraction method to detect microwave breast packing boxes for their class opportunities in each region.
cancer. We are encouraged to evaluate the use of EMD Unlike traditional R-CNN which requires multiple networks
components due to their potentially strong effects on the Figure 1 shows the structure of the system. It contains the
normal jitter system in microwave cancers. We evaluated the main element viz. Mammogram images, Release feature and

IJISRT21MAR421 www.ijisrt.com 1135

Volume 6, Issue 3, March – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
segmentation. Input photos are taken from the MIAS [8]. Abdullah-Al Nahid, Aaron Mikaelian and Yinan Kong,
database. These images are processed using various filters. Histopathological breast-image classification with
The Network layers on CNN serve as a visual filter for the restricted Boltzmann machine along with
presence of specific patterns or features present in an image. backpropagation, Biomedical Research Volume 29,
CNN's first layers get great features that can be easily Issue 10, (2018).
explained. The following layers get a few subtle features. [9]. Mohamad Mahmoud Al Rahhal, “Breast Cancer
The final layer is able to make a more detailed division by Classification in Histopathological Images using
combining all the features found in the previous layers. Convolutional Neural Network” International Journal
of Advanced Computer Science and Applications
IV. CONCLUSION (IJACSA), 9(3), 2018.
[10]. M. H. Yap et al., "Automated Breast Ultrasound
Witnessing the high mortality rate from breast cancer Lesions Detection Using Convolutional Neural
the need for computer-assisted diagnostic tools is clear. Networks," in IEEE Journal of Biomedical and Health
These CAD tools are used by doctor specifically radiologist. Informatics, vol. 22, no. 4, pp. 1218-1226, July 2018.
This approach serves as a second opinion and ultimately [11]. Md. Milon Islam, Hasib Iqbal, Md. Rezwanul Haque,
increase the accuracy. Reducing the number of essentials is Md. Kamrul Hasan. "Prediction of breast cancer using
very important as it causes unnecessary discomfort, costs support vector machine and KNearest neighbors",
and risks of biopsy. This software tool makes reduce the 2017 IEEE Region 10 Humanitarian Technology
time between patient and management by reviewing and Conference(R10-HTC),2017.
reporting results in a few seconds.

REFERENCES

[1]. Lei Fana Paul E. Gossb, c Kathrin Strasser-Weippld,

“Current Status and Future Projections of Breast
Cancer in Asia”, Breast Care 2015;10:372–378.DOI:
10.1159/000441818
[2]. https://www.repository.cam.ac.uk/handle/1810/250394
[3]. “Breast cancer statistics.” [Online]. Available:
http://www.wcrf.org/int/ cancer-facts-figures/data-
specific-cancers/breast-cancerstatistics.A. Jemal et al.,
“Cancer statistics, 2008,” CA. Cancer J. Clin., vol. 58,
no. 2, pp. 71–96, Apr. 2008.
[4]. A. Paul and D. P. Mukherjee, “Mitosis detection for
invasive breast cancer grading in histopathological
images,” IEEE Transactions on Image Processing, vol.
24, no. 11, pp. 4041–4054, 2015.
[5]. S. Albarqouni, C. Baur, F. Achilles, V. Belagiannis, S.
Demirci, and N. Navab, “Aggnet: deep learning from
crowds for mitosis detection in breast cancer histology
images,” IEEE transactions on medical imaging, vol.
35, no. 5, pp. 1313–1321, 2016.
[6]. B. E. Bejnordi, M. Veta, P. J. van Diest, B. van
Ginneken, N. Karssemeijer, G. Litjens, J. A. van der
Laak, M. Hermsen, Q. F.Manson, M. Balkenhol et al.,
“Diagnostic assessment of deep learning algorithms for
detection of lymph node metastases in women with
breast cancer,” Jama, vol. 318, no. 22, pp. 2199–2210,
2017.
[7]. S. Charan, M. J. Khan and K. Khurshid, "Breast cancer
detection in mammograms using convolutional neural
network," 2018 International Conference on
Computing, Mathematics and Engineering
Technologies (iCoMET), Sukkur, 2018, pp. 1-5, doi:
10.1109/ICOMET.2018.8346384.

IJISRT21MAR421 www.ijisrt.com 1136