Department of computer science and Engineering

Bangladesh University of Business and Technology

Breast Cancer Classification with Deep Learning

Id:16172103257(35)

Mahir Asef

Id: 1617210335

H M Safayat Jakir

Id: 16172103029

Md Rashed
 Id: 17181103026

MD Rashidul Islam

Id : 17181103001

Supervised by

Md. Samsuddin Ahmed

Assistant Professor
Department of Computer Science and Engineering
Bangladesh University of Business & Technology

TABLE OF CONTENTS

• Abstract

• Introduction

• Motivation

• Background

• Breast Cancer

• Types of breast cancer

• Methodology

• Conclusion

ABSTRACTS
Due to the danger that represents breast cancer in the society and also due to the number of death that
causes every year in women's sector, the proposition of solutions that aim to reduce this number of
death has become a primordial need. In this paper we present an approach for breast cancer diagnosis
using machines learning techniques that can be helpful in breast cancer detection in earlier stage and
with the right way without errors to propose the suitable treatment for the patients. We will use
machines learning techniques in our approach due to their performance in domain of medicine.

Introduction
Breast cancer can be defined as an abnormal production of cells in the breast that multiply
uncontrollably. The masses of cells formed in the breast are called tumors. Cancer cells can remain in
the breast or can spread to the body through the blood or lymphatic vessels, that what we named
metastasis. Most of the time, the progression of breast cancer takes several months and even a few
years. In Morocco 40,000 new cases of cancer are diagnosed each year, in which breast cancer came in
the first place in women sector with the percentage of 36% of revealed cases [1]. Also it represents the
second leading cause of death in women after lung cancer [2]. So breast cancer has become a danger for
Moroccan and global health, the number of incidence and mortality increases every year and decision
making for each case is sometimes difficult. So the idea of this paper is to propose an approach that aim
to reduce the number of misdiagnosis persons in order to decrease the number of death using machine
learning.

Motivation
We recognize that the way women decide to attend BSE training is influenced by personal and social
factors. Understanding the different risk assessments women rely on in making their health decisions is
essential. This study will assist researchers and health professionals to gain a better understanding of
alternative ways to deal with breast health, and not to be limited by the recommendations of the health
authorities.

Background
Early diagnosis of breast cancer increases the chance of recovery and life expectancy.

Screening is the primary tool for early diagnosis and timely treatment of breast cancer in early stages.

This qualitative study aimed to explain the motivational factors for breast cancer screening .

BREAST CANCER
Breast cancer as it has already been defined is an abnormal production of cells in the breast that form in
the form of cancerous masses, these masses are called tumors. Cancer cells can stay in the breast these
types of cancer are called non-invasive, they lead to healing and do not produce metastatic cases. The
other type of breast cancer is called invasive. These are dangerous type of cancers that can spread to
the other organs of the body and can lead to metastatic cases.

Types of breast cancer

Ductal carcinoma : This type of cancer is also formed in the lactation ducts, but cancer cells pass
through the canal wall.

Lobular carcinoma : In this type of cancer the cancerous cells appear in the lobules grouped in the
lobes. Afterwards, they cross the wall of the lobules and disperse in the surrounding tissues.

Inflammatory carcinoma: Is a rare type of cancer that is known by a breast that can turn red,
swollen and hot .

Paget's disease : Is also a rare type of cancer that is manifested by a small nipple wound that does
not heal.

Other carcinomas : (medullary, colloidal or mucinous, tubular, papillary). Are the rarest types of
breast cancer.

Methodology
As we said that we want to develop a solution for breast cancer diagnosis using machines learning
techniques, so the architecture of our proposed system will be the following as presented in given figure
. first we will gather the characteristics of the new patient that we want to diagnostic its case, then we
will combine those characteristics with the data of the already exciting patients in a knowledge base of
diagnostics that contain their characteristics and their diagnostics, in order to diagnosis the case of new
patient using machines learning algorithms
 Figure : Proposed System
Supervised learning :In Supervised learning the algorithms work as follows: we have the
input variable (x) and the output variable (y) that means we have the right answer, we have
already classified instances as testing set, so we want to build and develop the predictive model
to classify the new instances.

Unsupervised learning :In unsupervised learning the algorithms work as follows: we have
only the input variable (x) we don’t have the output variable, the grouping and the interpretation
of data is based only on the input. There are also another type of machine learning is semi
supervised this technique is between Supervised and Unsupervised learning techniques, it appear
when we have large amount of data (x) and only some data have the output (y). And because we
have the classes of diagnosis of breast cancer already defined and classified, so we will use the
algorithms of supervised learning and also the results of the experimentation below show that the
algorithms of supervised learning perform well than the algorithms of unsupervised learning.
Conclusion
According to the results of this study, it seems that health systems need to change the attitudes and
beliefs of people to enhance health culture by identifying women’s motivations for breast cancer
screening in different groups and increase social knowledge about screening methods by supporting
and training.