Scribd
Upload
10 views5 pages

Zeroth ReviewReport

This document proposes a deep learning system to classify histopathological lung cancer images. It involves collecting images, preprocessing them, using a CNN to extract features, and an SVM to classify the images into cancer types (adenocarcinoma, squamous cell carcinoma, etc.). Prior work showed CNNs achieving 70-99% accuracy on similar tasks. The proposed system aims to automatically diagnose lung cancer types from histopathology slides to assist pathologists.

Uploaded by

Star Boy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
10 views5 pages

Zeroth ReviewReport

This document proposes a deep learning system to classify histopathological lung cancer images. It involves collecting images, preprocessing them, using a CNN to extract features, and an SVM to classify the images into cancer types (adenocarcinoma, squamous cell carcinoma, etc.). Prior work showed CNNs achieving 70-99% accuracy on similar tasks. The proposed system aims to automatically diagnose lung cancer types from histopathology slides to assist pathologists.

Uploaded by

Star Boy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

LUNG CANCER DETECTION USING DEEP LEARNING

TECHNIQUES FROM HISTOPATHOLOGICAL IMAGES

Abstract:

Lung Cancer is a leading cause of death worldwide. Accurate diagnosis


and prognosis are essential in Lung cancer treatment. Histopathological
assessment of tissue is a standard procedure used for early diagnosis. Tissue
analysis is usually performed by pathologist review, even though the process is
time consuming it is considered to be error prone. There are several types of
lung cancer among them adenocarcinoma, benign and squamous cell carcinoma
are more prevailing hence diagnosis and treatment of lung cancer is very much
needed to know about the cancer type. Automated detection of cancer types
would significantly speed up the treatment process and will assist the
pathologist. So we propose a Deep learning based automated classification
system to classify the histopathological images to any of the cancer types.
We also propose to compare our classification result of our deep learning model
with machine learning model. It will be done by extracting the features of the
images using a deep learning model and these features are used to classify the
images using SVM.

Introduction:
Since Lung Cancer is a deadly disease around the world. Prognosis for
patients with this disease is poor with 5-year survival rate of less than 20%. Most
patients have poor prognosis because the diagnosis was made at an advanced
disease stage which leads to higher mortality rate. Patients diagnosed at an early
stage have significantly higher 5-year survival of over 70%. Lung cancer is
caused due to ……...Lung

Therefore, the microscopic examination of tissue slides remains an essential step


in cancer diagnosis. There are different imaging modalities to detect lung cancer,
one among them is histopathological images of cancer tissue.
Types of cancer….
According to each tumor’s histopathological features, lung cancer can be divided
into Non-Small Cell Lung Cancer(NSCLC) and Small Cell Lung Cancer(SCLC).
NSCLC accounts for 85% of Lung cancer cases and can be further separated
into Adenocarcinoma(ADC), Squamous Cell Carcinoma(SCC) lung cancer and
Large Cell Lung Cancer. Each subtype of lung cancer has a distinct biological
origin and mechanism(s).
Identifying the subtypes and recognition of histopathological patterns is required
to find the severity of the diseases. This process is time consuming and
generates considerable inter and intra observer variations. Computer Algorithms
for pathology image analysis have been facilitating cancer diagnosis and
prognosis.

Why DL:
Literature Survey:

Classification Model Accuracy

[1] Normal Vs Tumor CNN


i) VGG16 0.833
ii) ResNet50 0.769

[2] Malignant or CNN 89.8%


Non-malignant

[3] Adenocarcinoma ,
Squamous Cell
Carcinoma and
Small Cell Deep Convolutional Neural 70%
Carcinoma Network

[4] Normal Vs Tumor Inception V3,Fully Trained 0.99


CNN

Adenocarcinoma Vs Inception V3, Transfer 0.84


Squamous Cell Learning
Carcinoma

Adenocarcinoma Vs 0.95
Squamous Cell Inception V3, Fully Trained
Carcinoma CNN

[5] Normal Vs Tumor 0.85


Image Processing Techniques
with ML Algorithms
Adenocarcinoma Vs Image Processing Techniques 0.75
SCC with ML Algorithm

[1] has proposed classification of histopathological images into normal and


tumor class using CNN. Here two CNN architectures (VGG16 and RESNET50)
are trained and their performance was compared, where VGG16 slightly
outperforms ResNet50 by AUC value (VGG16 - 0.833, RESNET50 - 0.796).

[2] has proposed a CNN model to classify each 300*300 pixels image from
H&E stained lung histopathological images as malignant or non-malignant. Here
the overall classification accuracy was 89.8%.

[3] proposed an automated classification of lung cancer types from


cytological images using Deep Convolutional
Neural Network. Here the overall classification accuracy was 70%

[4] Automatic classification of lung cancer tissues, used the inception V3


and the Hematoxylin and Eosin stained histopathological images.

[5] Proposed a image processing techniques with Machine Learning


methods such as Random Forest, SVM, Naive Bayes classifier achieving the
area under curve 0.85,in distinguishing normal from tumor cells, and 0.75 in
distinguishing Adenocarcinoma from Squamous Cell Carcinoma.

Proposed System:

We propose a Deep learning based automated classification system to


classify the histopathological images to any of the cancer types.
We also propose to compare our classification result of our deep learning model
with machine learning model. It will be done by extracting the features of the
images using a deep learning model and these features are used to classify the
images using SVM.
Module Split-up:

Data Collection:
The images in the dataset are collected from the Original sample of
HIPAA complaint and validated sources.It contains 15,000 histopathological
images of 3 classes. These images are of 768*768 pixels in size and are in jpeg
file format.

Data Preprocessing:
Pre-processing methods regulate brightness and contrast variations in the image
and suppress noise. This provides ease of operation for classification algorithms that
are very sensitive to brightness and contrast fluctuations.
Color Normalization:
This task reduces the color and intensity variations present in the stained
images from different laboratories.Consequently, stain Normalization typically increases
the prediction accuracy of the CAD system. Also Normalization algorithms increase the
classification accuracy in the presence of poor quality of input images.
Color Decomposition:
Color appearance of digital histopathological images varies across
different specimen preparation, staining and scanning situations. It digitally separates
stains from color images. This variability may affect the diagnosis and decrease the
accuracy of CAD approaches.
CNN:
The preprocessed images are given to the Convolution Neural Network and the
features are extracted.

SVM:
The Features Extracted from the CNN are used to classify the Lung cancer type
by SVM.

//Conclusion:
We are going to propose a fully automated Deep Learning method to extract
features from the histopathological images and these features are used to classify the
images using SVM. And also possible to diagnose lung cancer types using Cytological
images.This can be done by applying data augmentation in the already existing dataset.

You might also like