A

Mini Project Report

on

“Breast Cancer Detection”

Submitted by: Submitted to:

Deepak Kumar (Dr. Anurag Sewak/

Roll No.: 2308410109003 Dr. Mainejar Yadav)
[Link]. (CSE-3rd Year) Assistant Professor
CSED, REC Sonbhadra

COMPUTER SCIENCE & ENGINEERING DEPARTMENT

RAJKIYA ENGINEERING COLLEGE SONBHADRA
(An AICTE-approved Government Engineering College affiliated to AKTU
Lucknow)

CHURK SONBHADRA - 231206

UTTAR PRADESH

1
 Abstract

Breast cancer remains one of the most prevalent and life-threatening diseases affecting
women worldwide. Early and accurate detection is critical for improving treatment outcomes
and reducing mortality rates. Traditional diagnostic techniques, including mammography and
MRI scans, often rely heavily on human expertise, making them prone to subjectivity,
human error, and variability in interpretation. These limitations have driven advancements in
medical imaging and artificial intelligence, particularly in the application of Convolutional
Neural Networks (CNNs) for automated image analysis.
This project focuses on developing an automated breast cancer detection system utilizing
MRI image datasets and CNN architectures. By leveraging deep learning techniques, the
system aims to efficiently classify breast MRI images into benign and malignant categories,
providing a reliable and scalable diagnostic tool. The dataset comprises balanced categories
of healthy and cancerous breast MRI images, ensuring model robustness and generalization.
The methodology involves systematic data preprocessing, including image resizing,
normalization, and augmentation, to prepare the dataset for efficient model training. A CNN
architecture is designed with multiple convolutional layers, pooling layers, and fully
connected layers to extract complex features from medical images. Techniques such as
dropout regularization are employed to minimize overfitting and enhance generalization
across unseen data.
Model evaluation is performed using accuracy metrics to ensure a comprehensive assessment
of the system's performance. Visualization tools are used to plot training and validation
accuracy and loss curves, providing insights into model convergence and stability during
training.
The project highlights the importance of differentiating between benign and malignant
tumors, as accurate classification directly impacts treatment plans and patient outcomes.

2
 Declaration

I/We hereby declare that the project titled - " Breast

Cancer Detection", which is being submitted as Mini Project of [Link]. 5th Semester (Computer Science &

Engineering) to the Department of Computer Science & Engineering, Rajkiya Engineering College,

Sonbhadra is an authentic record of my/our genuine work done under the guidance and mentorship of Dr.

Anurag Sewak & Dr. Mainejar Yadav, Assistant Professor, Computer Science & Engineering Department,

Rajkiya Engineering College, Sonbhadra. I have not plagiarized the contents of this report, and have not

submitted this work for the award of any other degree/certificate.

Date: 02/01/2025 Deepak Kumar

Place: REC,Sonbhadra (Name of the Student)

3
 Certificate

This is to certify that the Mini Project titled " Breast Cancer Detection "

submitted by “Deepak Kumar“ has been carried out under my guidance and that this project work has not

been submitted elsewhere for any degree. This project report is approved for submission requirement for

Mini Project in BTech 5th Semester in Computer Science & Engineering at Rajkiya Engineering College,

Sonbhadra.

Date: 02/01/2025 (Dr. Anurag Sewak / Dr. Mainejar Yadav)

Computer Science & Engg. Department, Rajkiya

Engineering College, Sonbhadra

4
 Table of Contents

S. No. TITLE Page No.

Introduction
1.1 Application Context/Background
1.2 Motivation
1. 1.3 Existing System and Proposed System 6 -7
1.4 System Objectives
1.5 Salient Features of the System

Breast Tumor Classification

2.1 Benign Tumors
2. 2.2 Malignant Tumors 8
2.3 Importance of Tumor Classification
2.4 Imaging Modalities for

Requirements Analysis
3.1 System Feasibility
3.2 Functional and Non-Functional Requirements 9
3. 3.3 System Validation
3.4 Dataset Overview

System Design
4.1 Design Approach and System Model
4. 4.2 Front-end Design 10
4.3 Back-end Design
4.4 System Snapshots
System Implementation
5.1System Requirements
5. 5.2Tools and Technologies used 11
5.3 System Setup and Testing
5.4 System Deployment

6. Implementation Code 12 - 16

Limitations and Scope

7. 7.1 Limitations 17
7.2 Scope
Conclusion and Future Work
8. 8.1Conclusion 17
8.2Future Work (if any)

5
 1. Introduction

1.1 Application Context/Background :

 Breast cancer remains one of the most common and life-threatening diseases affecting
women globally. Early detection and accurate diagnosis are critical to improving
treatment outcomes and reducing mortality rates. Traditional diagnostic methods, such
as mammography and MRI scans, rely heavily on radiologists' expertise, which can be
prone to human error and inconsistency. With advancements in technology, machine
learning (ML) and deep learning (DL) techniques, specifically Convolutional Neural
Networks (CNNs), have shown remarkable promise in automating medical image
analysis.
 This project aims to build an automated breast cancer detection system using MRI
image data and deep learning models. The use of CNNs enables efficient image
classification by extracting complex patterns and features from medical images.
 This Project focuses on leveraging machine learning and deep learning technologies to
automate breast cancer detection, enhancing diagnostic accuracy and reducing human
error.

1.2Motivation:
 The increasing number of breast cancer cases highlights the need for an
 automated and accurate detection system.
 Traditional diagnostic methods are time-consuming and prone to human error.
 Machine learning and deep learning models offer the ability to process large datasets
 efficiently.
 Traditional diagnostic methods are time-consuming and prone to human error.

The motivation behind this project arises from several factors:

 High Mortality Rate: Breast cancer remains one of the leading causes of cancer-
related deaths among women.
 Manual Diagnosis Limitations: Human errors, fatigue, and subjectivity in manual
MRI image analysis can lead to misdiagnosis.
 Technological Advancements: With the rise of artificial intelligence (AI) and CNNs,
it is now possible to achieve higher accuracy in image-based diagnoses.
 Scalability: Automated systems can analyze large datasets efficiently, making them
suitable for widespread clinical use.

1.3 Existing System and Proposed System:

6
  Existing System:
o Manual analysis of MRI images by radiologists.
o Limited by human interpretation and time constraints.
o Dependency on experts with specialized training.
 Proposed System:
o Use of a CNN-based deep learning model.
o Automated image preprocessing and classification.
o Scalability and integration with hospital management systems.

1.4 System Objectives :

 Develop a robust CNN model for breast cancer detection.

 Achieve high accuracy and reliability in classifying MRI images.
 Provide a user-friendly interface for healthcare professionals.
 Minimize false positives and false negatives.
 Ensure user-friendly deployment for healthcare practitioners.

1.5 Salient Features of the System:

 Automated preprocessing of MRI images (resizing, normalization).

 CNN-based deep learning model for image classification.
 Real-time prediction with a deployed API.
 Visualization of accuracy and loss metrics.

7
 2. Breast Tumor Classification

2.1 Benign Tumors:

Benign breast tumors are non-cancerous and generally do not spread to other parts of the
body. Common types include:

 Fibroadenomas: Common in younger women; usually painless.

 Cysts: Fluid-filled sacs within the breast tissue.
 Papillomas: Wart-like growths near the milk ducts.

2.2 Malignant Tumors:

Malignant tumors are cancerous and can invade nearby tissues or spread to other parts of the
body. Common types include:

 Ductal Carcinoma In Situ (DCIS): Early-stage cancer contained within the milk
ducts.
 Invasive Ductal Carcinoma (IDC): Cancer that begins in the ducts and spreads.
 Invasive Lobular Carcinoma (ILC): Cancer that starts in the milk-producing glands.
 Triple-negative breast cancer (TNBC): A more aggressive and harder-to-treat form
of cancer.

2.3 Importance of Tumor Classification:

Correct identification of tumor type is critical for determining treatment plans and
predicting prognosis. CNNs can assist radiologists by improving classification accuracy.

2.4 Imaging Modalities for Detection :

 Mammography: Common imaging technique for initial breast cancer screening.

 MRI (Magnetic Resonance Imaging): Provides detailed soft tissue images.
 Ultrasound: Useful in differentiating between solid masses and cysts.

8
 3. Requirements Analysis

3.1 System Feasibility:

 Technical Feasibility: The system is built using TensorFlow, Keras, and Python.
 Economic Feasibility: Open-source tools reduce implementation costs.
 Operational Feasibility: The system can be deployed seamlessly in healthcare
institutions.

3.2 Functional and Non-Functional Requirements:

 Functional Requirements:
o Upload MRI images.
o Classify images as benign or malignant.
o Display results with confidence scores.
 Non-Functional Requirements:
o High scalability.
o Real-time response.
o High classification accuracy.

3.3 System Validation:

 Dataset validation using train-test splits.
 Cross-validation for model robustness.
 Visualization of training loss and accuracy.

3.4 Dataset Overview:

 Healthy Images: 700 samples.
 Cancerous Images: 700 samples.
 Image Resolution: 128x128 pixels.
 Train-Test Split: 80-20.

9
 4. Design Approach and System Model

4.1 Design Approach and System Model:

 Input Layer: Accepts MRI images resized to (128x128) pixels.
 Hidden Layers: Convolutional layers for feature extraction.
 Output Layer: Binary classification (Benign=0, Malignant=1).

4.2 Front-end Design:

 Flask-based GUI.
 MRI image upload interface.
 Display classification results.

4.3 Back-end Design:

 Data preprocessing (normalization, resizing).
 Model training and evaluation.
 API endpoints for predictions.

4.4 System Snapshots:

 MRI image preprocessing results.
 Model accuracy and loss graphs.
 Prediction results screenshots.

4.5 System Architecture Diagram:

(Include diagram showing dataset, preprocessing, CNN model, prediction, and user interface
flow.)

10
 5. System Implementation

5.1 System Requirements

 Software: Python, TensorFlow, Keras, OpenCV.
 Hardware: GPU-enabled systems for faster training.
5.2 Tools and Technologies used
 TensorFlow for deep learning.
 OpenCV for image processing.
 Matplotlib for visualizations.
5.3 System Setup and Testing
 Train-Test Split (80-20).
 Data normalization to improve convergence.
 Model training over 10 epochs with Adam optimizer.
5.4 System Deployment
 Model saved as .h5 file for portability.
 Deployed using Flask API for prediction on new images.

11
 6. System Implementation Code
6.1 Importing Libraries

6.2 Download Dataset

6.3 Display Total Number of Images

6.4 Displaying the images

12
6.5 Image Processing

6.6 Splitting The Dataset

13
6.7 Building a Convolutional Neural Network

6.8 Fitting The Model

6.9 Accuracy for Test Data

6.10 Illustrating Loss & Accuracy

14
15
6.11 Building Predictive System

16
 7. Limitations and Scope
7.1 Limitations
 Dataset size limits generalization.
 Dependency on high-end hardware (GPUs).
 Limited to MRI images only.
7.2 Scope
 Can be extended to include other diagnostic imaging techniques like mammograms or
CT scans.
 Integration into hospital diagnostic systems.

8. Conclusion and Future Work

8.1 Conclusion
 The implemented CNN model successfully detects breast cancer from MRI images with high accuracy.
 Automated detection reduces diagnostic delays and human errors.

8.2 Future Work

 Expand dataset to improve model robustness.
 Integrate with cloud platforms for remote diagnostics.
 Add support for multi-class classification for different cancer types.

17