Enhancing Lung Cancer Detection with Deep Learning: A CT Image Classification Approach

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 07 | July 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 754
Enhancing Lung Cancer Detection with Deep Learning: A CT Image
Classification Approach
JEEVIKA K S1, DR. SAVITHA S K2
1PG Student, Dept. of Computer Science Engineering, Bangalore Institute of Technology, Bengaluru, India
2Professor, Dept. of Computer Science Engineering, Bangalore Institute of Technology, Bengaluru, India
---------------------------------------------------------------------***-------------------------------------------------------------------
Abstract - Lung cancer is a highly perilous illness ranking
as one of the primary causes of disease and death, particularly
when diagnosed in its initial stages. It presents significant
challenges, as it is often only discernible after it has already
diffused. This study proposes a lung cancer prognostication
framework that uses deep learning to enhancetheaccuracyof
cancer forecasting and disease determination, thereby
enabling personalized treatmentapproachesbasedon disease
severity. It consists of various steps, including image
preprocessing and segmentation of lung CT image features
extracted from the segmented images. Three differentmodels,
namely a DCNN model, a DCDNN model, and an ANN model,
were employed for image classification, and a deep
convolutional neural network (DCNN)wasemployedtodetect
lung diagnosis based on the extracted feature evaluation
results showing the best accuracy of 99.41% in accurately
discerning the presence or absence of lung cancer. The GAN
model generates realistic lung CT scan images by training a
generator to produce authenticimages, andadiscriminatorto
distinguish between real and fake images. The outcome of the
system depends on the quality of the data, and a well-trained
DCNN through training, validation, and testing on diverse
datasets is crucial to ensure thereliabilityand generalizability
of the model.
Key Words: Lung cancer detection, Deep learning, Deep
convolutional neural network.
1. INTRODUCTION
Lung cancer is a complex and heterogeneous ailment
characterized by a rise in the number of cells in lung tissues.
This stands as the main reason for most cancer-related
deaths worldwide, accounting for a sizable proportion of
cancer-associated morbidity and mortality. Detection
assumes a crucial role in identifying the existence of lung
cancer at an early stage and facilitatingtimelytreatment.This
study focuses on the development and improvement of
detection methods using different techniques.
This study provides an input CT image for image
preprocessing, which includes grayscaleconversionfor noise
removal, histogram calculation, and image quality
enhancement for more clearly visible images after image
enhancement.
The second step is image segmentation, which detects the
edge using Canny edge detection, and lung segmentation
techniques using K-means clustering to separate the
background and foreground. Morphological operations to
refine the lung mask. The original image binary threshold
image eroded, dilated color-labeled image-segmented lung
mask, and segmented lung area in the original image are
displayed. After the segmentation lung feature extraction
process, which is not displayed in the system, the model
directly underwent analysis assessment.
In the third step, the classification model was trained and
evaluated. Collect a dataset of lung images with labels normal
or cancer and split the dataset into three sets. In the
evaluation process, it provides access to the training history
to retrieve the training and validation accuracy, and the loss
values are plotted using Matplotlib. The models used in this
study were the DCNN, DCDNN, and ANN. Every model
exhibited good accuracy and loss percentage and plotted
graphs. Further details are provided in the followingsections.
In the fourth step, the GAN model predicts, the structure of
the image batch, class names, labels, and filenames in this
understood.
During the ultimate step, the Streamlit app for lung cancer
detection allows users to upload images and predict cancer.
User-friendly interface thatinteractswithanapplicationona
web browser.
1.1 Aims and Objectives
1) The aim of this project is to enhance the precision
and effectiveness of lung cancer diagnosis by advanced
deep learning methodologies.
2) To develop an application that detects and properly
classifies Lung cancer in CT scan images using a DCNN.

1.2 Flow Diagram
Figure 1: Proposed system
2. PROPOSED METHODOLOGY
A. Image Preprocessing
Image preprocessing plays a fundamental role in the
accurateanalysis of medical images, particularly indetecting
malignancy using CT scans. This section focuses on the key
preprocessingsteps:greyconversion,histogramanalysis,and
image quality enhancement. Gray conversion simplifies the
image by converting it to grayscale, thereby facilitating the
subsequent processing. Histogramanalysisprovidesinsights
into the pixel intensity distribution, aiding in the threshold
determination for image segmentation. Image quality
enhancement techniques enhance visibilityandreducenoise
and artifacts. The cumulative distribution function is
normalized and scaled to the range [0,25]. By implementing
these preprocessing steps, CT images were optimized for
accurate lung cancer detection, ensuring improved analysis
and more reliable results.
Figure 2: Input image and Quality enhanced image
B. Image Segmentation
Image segmentation is an important step in this study. It
involves precise identification and separation of the
cancerous region within the enhanced tomography images.
To achieve this, the k-means algorithm is typically employed.
The segmentation process begins by examining the pixel
similarity in the lung images. Thepicture waspartitionedinto
multiple subregions, allowing the prediction and localization
of the affected area. This is accomplished by analyzing the
similarity of pixel values and grouping superpixels that
exhibit similar characteristics. The segmentation algorithm
effectively identifies the outline of the malignant region by
detecting abrupt changes in pixel values. Pixel similarity is a
fundamental attribute in image segmentation and critical
analysis of visual data in accurately shaping clusters.
Through quantitative evaluation, the algorithm determines
the level of similarity between pixels, facilitating the correct
division of the cancerous region. This evaluation considered
the enhanced quality of the image pixels, ensuring that the
segmentation process was performed effectively.
An undirected graph was constructed to represent the
abnormal regions in the lung image. This graph provides a
visual representation of the segmented areas, aiding the
identification and analysis of cancerous regions.
Figure 3: Edge-Detected Image
Figure 4: Segmented Images

C. Image Classification
The final stage involves the recognition of malignant.
Detection was performed utilizing algorithms with CT scan
image input. Dataset of lung images with labels indicating
cancer or not. After splitting the dataset into training,
validation, and test sets as a folder. The algorithms used are
deep convolutional neural network, double convolutional
deep neural network, and artificial neural network.
At this stage, The Generative Adversarial Network (GAN)
model acts as a discriminator, tasked to distinguishbetween
real and synthetic images, with the ultimate goal of
producing life-like images capable of misleading
discriminators Through training the GAN model enhances
the fidelity of chest CT scan images which were then saved
and displayed. The saved model is readily deployable for
further analysis, evaluation, or application-specific tasks.
After the user uploads an image of the lung tissue, the image
is preprocessed and passed through a pre-trained model.
The model predicts the probability of an image belonging to
each class: "CANCER" or "NORMAL.” The class with the
highest probability was considered the predicted class for
the input image. The predicted class was then displayed on
the web interface to indicate whether the image was
classified as cancerous or normal lung tissue.
It should be emphasized that the accuracy and reliability of
cancer identification depend on the excellence of the pre-
trained model, the dataset on which it was trained, and the
diversity and quality of the images used for testing and
evaluation. The performance canvarybasedonthesefactors,
and a well-validated and reliable model is recommended for
the accurate identification of lung cancer.
Figure 5: Detection of cancer or not
3. RESULTS
This paper work finds that image preprocessing shows the
results in Figure (2) shows the image quality enhancement,
Figure (3) shows the edge detection using the canny edge
detection, Figure (4) shows the lung segmentation using k-
means clustering, Figure (5) shows the detected cancer or
not, Figure (6) DCNN (7) DCDNN (8) ANN shows the
Accuracy and loss graph for training and validation.
A. Table 1: Accuracy analysis
Methods Final Accuracy Best Accuracy
DCNN 96.6 99.41
DCDNN 93.2 95.15
ANN 57.4 80.0
B. Graph for Accuracy and Loss (training and validation)
A Deep convolutional neural network model consists of
multiple convolutional, pooling, and dense layers. After
training, validation, and testing, the data were obtained.
Training the model saves it, evaluates its performance using
accuracy metrics, and plots graphs.
Figure 6: DCNN graph
A Double convolutional deep neural network it consists of
convolutional and pooling layers, followedbyfullyconnected
layers. After training, validation, and testing, the data were
analyzed. Training the model saves the model, evaluates its
performance using accuracy metrics, and plots the training
and validation accuracy and loss curves.
Figure 7: DCDNN graph
An Artificial neural network in this model consists of a
flattened input layer, a dense hidden layer, and an outputlayer.
After training, validation, and testing data.Trainingthemodel

saved evaluates its performance using the accuracy metrics,
and plots it is shown in below figure 8.
Figure 8: ANN Graph
4. CONCLUSION
In conclusion, this paper developed using deep learning
techniques showed promising results. The main prediction
of the presence or absence of lung cancer is achieving high
accuracy and enabling personalized treatment approaches
based on disease severity. The ability of the system to
classify CT images provides a valuabletool fordetection.The
system development begins with image preprocessing,
segmentation, and image classification using different
algorithms or models, including DCNN, DCDNN,andANN. In
this model, the evaluation results show that the best
accuracy is achieved by the DCNN, which yields 99.41%
accuracy and detects lung cancer.
Future research should prioritize the early detection and
exploration of the stage of cancer.Integrationwithadvanced
medical imaging technologies and clinical data could also be
explored to augment thepredictivecapabilitiesofthesystem
and enable more comprehensive lung cancer analysis.
REFERENCES
[1]. Heng Yu, Zhiqing Zhou, And Qiming Wang: “Deep
Learning Assisted Predict of Lung Cancer on Computed
Tomography Images Using the Adaptive Hierarchical
Heuristic Mathematical Model” In the IEEE Access,
Volume 08 2020, pp 86401-86411, Digital Object
Identifier 10.1109/ACCESS.2020.2992645
[2]. Mohammad A. Alzubaidi, Mwaffaq Otoom ,(Senior
Member, Ieee),AndHamza Jaradat,“Comprehensiveand
Comparative Global and Local Feature Extraction
Framework for Lung Cancer Detection Using CT Scan
Images”, In the IEEE Access, Volume 9, 2021, pp
158140- 158154, Digital Object Identifier
10.1109/ACCESS.2021.3129597
[3]. Wafaa Alakwaa, Mohammad Nassef, Amr Badr, “Lung
Cancer Detection and Classification with 3D
Convolutional Neural Network (3D-CNN)”, in
Proceedings of (IJACSA) International Journal of
Advanced Computer Science and Applications, Volume.
8, 2017, pp 409-417 https://www.ijacsa.thesai.org/
[4]. Prathyusha Chalasani, S Rajesh.: “Lung CT Image
Recognition using Deep Learning Techniques to Detect
Lung Cancer” .in International Journal of Emerging
Trends in Engineering Research., Volume 8. No. 7, July
2020, ISSN 2347–3983,
https://doi.org/10.30534/ijeter/2020/113872020
[5]. Rasika N. Kachore, Kivita Singh: “Lung Nodule
Detection”, in the International Journal of Scientific
Research in Science, Engineering and Technology,
IJSRSET, Volume 3, Issue 2, ISSN: 2395-1990, Online
ISSN : 2394-4099.
[6]. Yaozhi Lu, Shahab Aslani, Mark Emberton, Daniel C.
Alexander, And Joseph Jacob: “Deep Learning-Based
Long Term Mortality Prediction in the National Lung
Screening Trial”, in the proceedings of IEEEAccess,Year
– 2022, VOLUME 10, 2022, pp: 34369-34378, Digital
Object Identifier 10.1109/ACCESS.2022.3161954
[7]. B. Keerthi Samhitha, Suja Cherukullapurath Mana,
Jithina Jose, R.Vignesh, D. Deepa: “Prediction of Lung
Cancer Using Convolutional Neural Network (CNN)”, in
Proceedings of the International Journal of Advanced
Trends in Computer Science and Engineering,Volume9,
No.3, 2020,ISSN 2278-3091,
https://doi.org/10.30534/ijatcse/2020/135932020
[8]. Surbhi Vijh, Prashant Gaurav, Hari Mohan Pandey:
“Hybrid bio-inspiredalgorithmandconvolutional neural
network for automatic lung tumor detection”, in the
springer, Year of publication 19 September 2020,
https://doi.org/10.1007/s00521-020-05362-z
[9]. Dr. M. Sangeetha, P Sangeetha, P. Pavithra:
“Classification of Lung Cancer using Deep Learning
Algorithm”, in the ConferenceproceedingsInternational
Journal of Engineering Research & Technology (IJERT)
ISSN: 2278-0181 Published by, www.ijert.org RTICCT-
2020.
[10]. S. Perumal, T. Velmurugan, “Lung cancer detection
and classification on CT scan images using enhanced
artificial bee colony optimization”, in the International
Journal of Engineering & Technology Website:
www.sciencepubco.com/index.php/IJET, Volume 7,
Year-2018, pp 74-79.
[11]. Onur Ozdemir , Member, IEEE, Rebecca L. Russell,
and Andrew A. Berlin, Member, IEEE “A 3D Probabilistic
Deep Learning System for Detection and Diagnosis of
Lung Cancer Using Low-Dose CT Scans”, in the IEEE

TRANSACTIONS ON MEDICAL IMAGING, Volume. 39,
NO. 5, MAY 2020, pp: 1419-1429
[12]. Khevna Vasani, Ayushi Shah: “Lung Cancer
Detection Using CT Scan Images”, in the International
Research Journal of Engineering and Technology
(IRJET), e-ISSN: 2395-0056 Volume: 08 Issue: 04 Apr
2021 www.irjet.net p-ISSN: 2395-0072
[13]. Longxi Zhou, Zhongxiao Li, Juexiao Zhou, Haoyang
Li, Yupeng Chen, Yuxin Huang, Dexuan Xie, Lintao Zhao,
Ming Fan , Shahrukh Hashmi, Faisal Abdelkareem ,
Riham Eiada , Xigang Xiao, Lihua Li, “Adaptive Local
Ternary Pattern on Parameter Optimized-FasterRegion
Convolutional Neural Network for Pulmonary
Emphysema Diagnosis”, in the IEEE Access, Volume
9,2021, pp:114135-114152, DOI:
10.1109/ACCESS.2021.3105114
[14]. Chanunya Loraksa, Sirima Mongkolsomlit, Nitikarn
Nimsuk, Meenut Uscharapong, and Piya Kiatisevi:
“Development of the Osteosarcoma Lung Nodules
Detection Model Based on SSD-VGG16 and Competency
Comparing with Traditional Method”,intheIEEEAccess,
VOLUME 10, 2022, pp: 65496-65506, Digital Object
Identifier 10.1109/ACCESS.2022.3183604
[15]. Aishwarya Kalra, Brijmohan Singh, Himanshu
Chauhan, “An Approach for LungCancerDetectionusing
Deep Learning”, in the International ResearchJournal of
EngineeringandTechnology (IRJET),e-ISSN:2395-0056
Volume: 07, Issue: 09 Sep 2020 www.irjet.net p-ISSN:
2395-0072
[16]. P. Mohamed Shakeel, M. A. Burhanuddin,
Mohammad Ishak Desa, “Automatic lung cancer
detection from CT image using improved deep neural
network and ensemble classifier”, published in the
springer pp:349579–9592
https://doi.org/10.1007/s00521-020-04842-6
[17]. Zhang Li, Jiehua Zhang, Tao Tan, Xichao Teng,
Xiaoliang Sun, Hong Zhao, Lihong Liu, Yang Xiao, “Deep
Learning Methods for Lung Cancer Segmentation in
Whole-SlideHistopathology Images-TheACDC@LungHP
Challenge 2019”, in the IEEE Journal Of Biomedical And
Health Informatics, Volume. 25, No. 2, February 2021
[18]. Lulu Wang: “Deep Learning TechniquestoDiagnose
Lung Cancer”, in the MDPI, Volume 14 , Issue 22 ,DOI:
10.3390/cancers14225569
[19]. Imdad Ali, Muhammad Muzammil, (Member,IEEE),
IHSAN UL HAQ, (Member, IEEE), AMIR A. Khaliq,
(Member, IEEE), “Efficient Lung Nodule Classification
Using Transferable Texture Convolutional Neural
Network”, In the IEEE Access, Volume 8, 2020, pp:
175859- 175870, Digital Object Identifier
10.1109/ACCESS.2020.3026080, Year of publication –
2020
[20]. Longxi Zhou, Zhongxiao Li, Juexiao Zhou, Haoyang
Li, Yupeng Chen, Yuxin Huang, Dexuan Xie,Lihua Li ,
Member,IEEE, Zhaowen Qiu, Member,IEEE,andXinGao,
Member,IEEE, “A Rapid, AccurateandMachine-Agnostic
Segmentation and Quantification Method for CT-Based
COVID-19 Diagnosis”, in the IEEE Transactions On
Medical Imaging, Volume. 39, NO. 8, AUGUST 2020, pp:
2638- 2652

Enhancing Lung Cancer Detection with Deep Learning: A CT Image Classification Approach

More Related Content

Similar to Enhancing Lung Cancer Detection with Deep Learning: A CT Image Classification Approach (20)

More from IRJET Journal (20)

Recently uploaded (20)

Enhancing Lung Cancer Detection with Deep Learning: A CT Image Classification Approach