SRM INSTITUTE OF SCIENCE AND TECHNOLOGY

FACULTY OF ENGINEERING AND TECHNOLOGY

DEPARTMENT OF NETWORKING AND COMMUNICATIONS
18CSP107L / 18CSP108L- MINOR PROJECT / INTERNSHIP

Automated Breast Cancer Classification

and Analysis Using CNNs

Student 1 Reg No: RA1911029010067

Student 2 Reg No: RA1911029010090
Batch ID: NWC079067
Guide name and Designation: Mrs. G.
Parimala (Assistant Professor, Dept. of NWC, SRM IST)
 Table of contents

• Introduction – Problem statement

• Objectives
• Literature Survey
• Architecture Diagram
• UML Diagram
• Algorithms used

11/05/2023 2
 Problem Statement

To build an algorithm which automatically

identifies whether a patient is suffering from
breast cancer or not by analyzing biopsy images
and to verify the performance and results when
CNNs are employed for this task (multiple
models).

11/05/2023 3
 Objectives

• To analyze medical biopsy images of cancer

cells (from dataset).
• To classify the cancer cells as malignant or
benign.
• To compare the performance and results
achieved by various CNN models.
• To incorporate a visual layer to this model (UI).

11/05/2023 4
 Research Paper

11/05/2023 5
 Literature Survey
Topic Authors Description
Optimizing the Laith Alzubaidi Nuclei morphological analysis is performed for
classifying a tissue as benign or malignant.
Performance of Breast Omran Al-Shamma Handcrafted features, such as texture, topological and
morphological, are employed by Kowel et al. for
Cancer Classification by Mohammed A. Fadhel training a classifier using 500 images, representing 50
Employing the Same patients. They attained 84%–90% accuracy.

Domain Transfer Learning

from Hybrid Deep
Convolutional Neural A circular Hough transform and Otsu-threshold
techniques were performed by Filipczuk and George
Network Model for extracting the nuclei-related texture, shape and
features. The watershed was used by George for
refining the nuclei segmentation. He obtained an
accuracy of 71.95%–97.15%. In contrast,
the Filipczuk majority voting algorithm obtained
98.51% accuracy based on 11 images.

Belsare extracted spatial–color–texture graphs for

segmenting the cells lumen and the epithelial layers,
as well as utilizing tissue organization architecture to
classify histology images. In addition, he obtained an
accuracy of 70%–100% when he trained the final
classifier on the features of the statistical texture.

11/05/2023 6
 Literature Survey
Topic Authors Description
Optimizing the Laith Farhan Brook et al. introduced an approach of 3-class classification for
classifying the breast histology dataset into three classes:
Performance of Breast Jinglan Zhang invasive, non-invasive carcinoma, and normal. They trained
the Support Vector Machine (SVM) classifier on the biopsy
Cancer Classification by Ye Duan images using connected components. Brook et al. attained
Employing the Same accuracy between 93.4%–96.4%.

Domain Transfer
Learning from Hybrid Zhang trained an SVM on local-binary images and Curvelet
transform, using an approach of cascaded classification. He
Deep Convolutional obtained 97% accuracy. In addition, he included the rejection
technique when dissimilarity has occurred.
Neural Network Model
Spanhol et al. employed a CNN architecture, which was used in
the IMAGENET Classification challenge, for classifying his
dataset, which contained images from 60 patients.
They employed 7000 H&E-stained images, with a diverse range
of magnifications such as 400×, 200×, etc. Random extraction
and sliding window were utilized for extracting patches of 64 ×
64 and 32 × 32 pixels. These patches were employed for training
the classifier. For classification, the sum rule and patch
probabilities with maximum products were utilized. In addition,
they concluded that the attained accuracy decreases with the
increase of magnification.

11/05/2023 7
 Literature Survey
Summary Drawbacks
1. In general, researchers have successfully improved the CNN • 500 images were used to perform the classification in the
architecture to solve several issues related to the breast research, our database contains 2,480 benign and 5,429
histology. The winners of the ICPR 2012 Contest were Ciresan et malignant samples (700X460 pixels, 3-channel RGB, 8-bit
depth in each channel, PNG format.
al. They utilized patches of size 101 × 101 to train the CNN, and
achieved 78% accuracy. Image mirroring and random
rotation were used to increase the complexity and the size of
the training dataset.

• Image pixels were around 64x64, our images

will be around 700x400 for the betterment of the results.

2. Recently, researchers have implemented CNN models on • Single techniques and algorithms were used by every
ICIAR 2018 dataset images to identify four different classes team on the same images, we will be using up to four
of hematoxylin–eosin-stained breast biopsy images, namely, architectures of deep learning in our project.
invasive carcinoma, in-situ carcinoma, benign tumor and normal
tissue, using the fine-tuned deep network fusion and
hybrid Convolutional Neural Networks. These methods • Magnification of 400x and 200x were used on images of
accomplished the highest image-wise accuracies of 92.5% and around 64 patients, we will use microscopic images of
93%, respectively. breast tumor tissue collected from 82 patients using
different magnifying factors (40X, 100X, 200X, and 400X.

11/05/2023 8
 Architecture Diagram
Project Architecture

11/05/2023 9
 Architecture Diagram
Project Architecture

11/05/2023 10
 Architecture Diagram
AlexNet Architecture

11/05/2023 11
 Architecture Diagram
VGG Architecture

11/05/2023 12
 Architecture Diagram
ResNet Architecture

11/05/2023 13
 Architecture Diagram
DenseNet Architecture

11/05/2023 14
 UML Diagram

11/05/2023 15
 Algorithm Used

• Our input is a training dataset that consists of N images,

each labelled with one of 2 different classes.
• Then, we use this training set to train a classifier to
learn what every one of the classes looks like.
• In the end, we evaluate the quality of the classifier by
asking it to predict labels for a new set of images that it
has never seen before. We will then compare the true
labels of these images to the ones predicted by the
classifier.

11/05/2023 16
 For Further Review

11/05/2023 17
 Thank You!

11/05/2023 18