See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/380169995

LUNG CANCER DETECTION USING DEEP LEARNING

Article · April 2024

DOI: 10.5281/zenodo.11079202

CITATIONS READS
2 833

6 authors, including:

B.Swarajya Lakshmi
G pullaiah engineering college
5 PUBLICATIONS 23 CITATIONS

SEE PROFILE

All content following this page was uploaded by B.Swarajya Lakshmi on 29 April 2024.

The user has requested enhancement of the downloaded file.

Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024

LUNG CANCER DETECTION USING DEEP LEARNING

B.Swarajya Lakshmi11, G.Bramarambika2, N.MeghanaReddy3, U.GuruKavitha4, T.SudheshnaReddy5
1
Assistant Professor in Department of Computer Science and Engineering, Santhiram Engineering College,
Nandyal, Kurnool, Andhra Pradesh,India.
2,3,4,5
Student, Department of Computer Science and Engineering, Santhiram Engineering College, Nandyal,
Kurnool, Andhra Pradesh, India.
email: [email protected] , [email protected]
DOI : https://doi.org/10.5281/zenodo.11079202

Abstract: Early detection of lung cancer is crucial for effective treatment. Traditional methods like X-ray and CT scans are
widely used but not always accessible globally. This study proposes a comprehensive approach for lung cancer detection using
medical image mining techniques. The process involves lung field segmentation, feature extraction, and classification utilizing
neural networks and support vector machines (SVM). Two datasets were utilized, and various experiments were conducted,
including feature selection and training Capsule Networks (CapsNet) with different parameters. Results indicate promising
accuracy rates: Convolutional Neural Network (CNN) achieved 21%, K-Nearest Neighbors (KNN) achieved 43%, SVM
achieved 90%, and Random Forest achieved 60%. Notably, CapsNet outperformed all other methods with an accuracy of 96%.
This suggests the potential of CapsNet in enhancing lung cancer detection. The findings underscore the importance of
advancing automated techniques for early-stage lung cancer diagnosis, especially in regions with limited access to advanced
screening technologies.

Keywords : Lung Cancer Detection, CNN, SVM, KNN, Random Forest, CapsNet, Deep Learning, Machine Learning.

1. INTRODUCTION Leveraging advanced technologies like Neural Networks

and Machine Learning, this research endeavors to
Lung cancer stands as one of the most formidable streamline the classification process, enabling accurate and
adversaries in the realm of oncology, characterized by swift identification of abnormal lung patterns indicative of
unbridled cellular proliferation within the lung tissues. carcinoma. The evaluation of proposed methodologies
Early detection is paramount to curative interventions, as hinges on the accurate classification of sample data,
treatment efficacy dramatically declines in advanced serving as a litmus test for the efficacy of the developed
stages. While modalities such as X-ray, CT scans, and MRI algorithms.
are pivotal in diagnosis, their widespread availability
remains elusive in many regions, underscoring the Central to this endeavor are fundamental techniques in
significance of simpler yet effective screening methods like medical image mining, encompassing lung field
chest radiography. segmentation, image processing, feature extraction, and
classification. Through the meticulous application of these
The primary objective of this paper is to automate the techniques, digital X-ray chest films are categorized into
classification process for early-stage lung cancer two distinct classes: normal and abnormal. The
prediction, thereby facilitating timely interventions. classification process is underpinned by the deployment of

Page | 100 Copyright @ 2024 Author

Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024
sophisticated algorithms, including Neural Networks and [4] V. Krishnaiah, Dr. G. Narsimha, and Dr. N. Subhash
Support Vector Machines (SVMs), which are trained and Chandra's work focused on the development of a lung
fine-tuned using diverse datasets. cancer prediction system using data mining classification
techniques. Their research aimed to create a predictive
The experimental framework of this study encompasses a
model capable of identifying patterns indicative of lung
multifaceted approach. Two distinct datasets are curated,
cancer, aiding in early diagnosis and treatment.
each subjected to various learning experiments. Feature
selection techniques are employed to optimize model [5] Astha Pathak et al. conducted a survey-based study on
performance, while SVMs are trained with a spectrum of machine learning algorithms for lung cancer prediction.
parameters to ascertain the most effective configurations. Their research aimed to provide insights into the various
The ensuing results are meticulously scrutinized, machine learning techniques employed in predicting lung
compared, and reported, shedding light on the relative cancer, highlighting their strengths and limitations in
efficacy of different methodologies in automating lung clinical applications.
cancer classification.
[6] Kambam Shreya et al. proposed a machine learning
2. LITERATURE SURVEY approach for lung cancer analysis. Their study aimed to
develop a robust model capable of analyzing data related to
lung cancer and providing valuable insights for clinical
[1] Zakaria Suliman Zubi and Rema Asheibani Saad's
decision-making.
research focused on utilizing data mining techniques for
the early diagnosis of lung cancer. They explored the [7] K R Lathakumari, A C Ramachandra, U C Avanthi, C
application of various algorithms to aid in the timely Basil Ronald, and T Bhavatharani explored the
detection of this disease, aiming to improve patient classification of non-small cell lung cancer using deep
outcomes through early intervention. learning techniques. Their research aimed to leverage the
power of deep learning algorithms to accurately classify
[2] In a study by Paola Campadelli, Elena Casiraghi, and
different subtypes of lung cancer, thereby aiding in
Diana Artioli, a fully automated method for lung nodule
personalized treatment strategies.
detection from postero-anterior chest radiographs was
proposed. Their research aimed to develop a robust [8] Asha V and Bhavanishankar K conducted a review on
algorithm capable of accurately identifying nodules in lung cancer detection using CT scans and image processing
chest X-rays, facilitating early detection and treatment through deep learning techniques. Their study aimed to
planning. provide an overview of the current state-of-the-art methods
in lung cancer detection, focusing on the role of deep
[3] Rohit Chitale et al. investigated the utilization of
learning algorithms in analyzing CT scan images.
Convolutional Neural Networks (CNN) and ensemble
methods to support medical practitioners in categorizing [9] P. Chitra et al. investigated lung cancer detection using
lung cancer. Their study delved into the potential of classification algorithms. Their research aimed to develop
advanced machine learning techniques to assist healthcare and evaluate various classification algorithms for
professionals in accurately classifying lung cancer cases. accurately detecting lung cancer, providing insights into

Page | 101 Copyright @ 2024 Author

Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024
the performance of different machine learning techniques surpassing traditional methods such as CNN, KNN, SVM,
in this domain. and Random Forest. Overall, the proposed system
showcases the potential of advanced machine learning
[10] Nitha V. R and Vinod Chandra S. S. explored lung
techniques in enhancing the early detection of lung cancer,
cancer malignancy detection using a voting ensemble
particularly in regions with limited access to advanced
classifier. Their research aimed to develop an ensemble
screening technologies.
learning approach capable of combining multiple
classifiers to improve the accuracy of lung cancer ii) System Architecture:
malignancy detection.
The first image is transformed to gray scale image. After
These studies collectively highlight the growing interest in that, removal of the noises and contrast enhancement is
leveraging advanced technologies, such as machine finished for obtaining the improved images. After image
learning and deep learning, for the early detection and acquisition the system perform preprocessing on image
diagnosis of lung cancer. By harnessing the power of data- understand affected regions and their characteristics in
driven approaches, researchers aim to enhance clinical style of data. This data is classed using
decision-making and improve patient outcomes in the RF,KNN,SVM,CNN & CapsNet. CapsNet classify it as
battle against this deadly disease. normal or diseases lung and identify lung diseases.

3. METHODOLOGY

i) Proposed Work:

The proposed system employs medical image mining

techniques for comprehensive lung cancer detection,
focusing on early-stage diagnosis. Initially, the system Image acquisition
conducts lung field segmentation, followed by image
processing and feature extraction. Utilizing neural
networks and support vector machines (SVM), the Pre processing image
extracted features are then classified into two categories:
normal and abnormal. Training of the system involves
diverse datasets, including computed tomography scan
images with abnormal physiology, enhancing its ability to Feature Extraction
identify subtle abnormalities indicative of lung cancer.
Additionally, the system incorporates Capsule Networks
(CapsNet) for improved classification accuracy. Caps Net
Experimental evaluations demonstrate promising results,
with CapsNet exhibiting a remarkable accuracy of 96%,

Detection
Page | 102 Copyright @ 2024 Author
Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024
ImageDataGenerator module, which provides several
augmentation techniques to enrich the dataset. These
techniques include re-scaling the image to normalize pixel
values, shear transformation to introduce controlled
distortions, zooming to simulate varying perspectives, and
horizontal flipping to increase dataset diversity.
Additionally, reshaping the image ensures uniform
Fig3. 1 Proposed Architecture dimensions for compatibility with neural network
architectures. Another method utilizes Torchvision,
iii) Dataset Collection:
offering pre-defined functions tailored for object detection

The dataset utilized in this study comprises a collection of tasks. This framework enables operations such as resizing,

digital X-ray and computed tomography (CT) scan images normalization, and transformation of images to optimize

of the chest region. These images encompass a diverse their suitability for detection algorithms. By incorporating

range of cases, including both normal and abnormal these image processing techniques, the quality and

physiological conditions associated with lung cancer. The diversity of the dataset are enhanced, leading to improved

dataset is curated to represent various stages and model performance and robustness in detection tasks.

manifestations of the disease, facilitating robust training

v) Training & Testing:
and evaluation of the proposed system. Additionally,
specialized subsets within the dataset focus on specific The dataset is partitioned into training and testing sets
characteristics relevant to lung cancer diagnosis, ensuring using an 80:20 ratio, respectively. This division ensures
comprehensive coverage of relevant image features for that 80% of the data is allocated for training the model,
accurate classification. allowing it to learn patterns and features from a diverse
range of examples. The remaining 20% of the dataset is
reserved for testing the trained model, enabling evaluation
of its performance on unseen data. This split ensures that
the model's generalization ability is effectively assessed,
providing insights into its effectiveness in accurately
classifying lung cancer images while mitigating the risk of
overfitting to the training data.

vi) Algorithms:
Fig 3.2 Sample Dataset Image
CNN (Convolutional Neural Network): A deep learning
iv) Image Processing: algorithm specifically designed for processing and
classifying visual data, such as images. CNNs consist of
Image processing plays a crucial role in preparing and
multiple layers, including convolutional layers for feature
enhancing images for various applications, including
extraction and pooling layers for dimensionality reduction,
detection tasks. One approach involves using the
followed by fully connected layers for classification.

Page | 103 Copyright @ 2024 Author

Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024

Fig 3.5 CNN

CapsNet (Capsule Network): A type of neural network

architecture introduced to overcome the limitations of
traditional CNNs in handling hierarchical relationships and
spatial hierarchies within data. CapsNet incorporates Fig 3.6 KNN
capsules, which are groups of neurons that encode the
instantiation parameters of specific entities in the input SVM (Support Vector Machine): A powerful supervised
data, enabling better generalization and robustness. learning algorithm used for classification and regression
tasks. SVM works by finding the hyperplane that best
separates different classes in the feature space, maximizing
the margin between classes while minimizing classification
errors. SVM can handle linear and non-linear data
separation using different kernel functions.

Fig 3.3 CapsNet

KNN (K-Nearest Neighbors): A simple yet effective

supervised learning algorithm used for classification and
regression tasks. KNN works by assigning a class label to
an input sample based on the majority class label among its
k nearest neighbors in the feature space, where distance
metrics such as Euclidean distance are commonly used to
measure similarity.
Fig 3.7SVM

Random Forest: An ensemble learning algorithm that

operates by constructing a multitude of decision trees
during training and outputting the mode of the classes
(classification) or the mean prediction (regression) of the
individual trees. Random Forest mitigates overfitting and

Page | 104 Copyright @ 2024 Author

Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024
improves robustness by combining predictions from
multiple decision trees trained on different subsets of the
data.

Fig 4.3Login Page

Fig 3.8 Random Forest

4. EXPERIMENTAL RESULTS

Fig 4.4Uploaded Input Image

Fig 4.1 Home Page

Fig 4.5Predict Result for Given Input Image as Patient has

Lung Aca

5. CONCLUSION
Fig 4.2Registration Page
Eventually, this project involved different image
processing procedures of lung modules that were intended
for the most appropriate lung cancer detection. Among the
variety of image processing methods, fuzzy filter,
undoubtedly, showed the best performance in terms of
noise suppression and contrast
improvement. Segmentation by watershed marker-based

Page | 105 Copyright @ 2024 Author

Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024
algorithm provided the means for water areas division, thus From Postero-Anterior Chest Radiographs,” In Proc. of
IEEE TRANSACTIONS ON MEDICAL IMAGING,
facilitating feature extraction. Through this method, not
VOL. 25, NO. 12, DECEMBER 2006.
only accuracy and reliability were enhanced, but also faster
[3] V.Krishnaiah, Dr.G.Narsimha, Dr.N.Subhash Chandra.
result generation was made possible. The final outcome of
2013, “Diagnosis of Lung Cancer Prediction System Using
these efforts was the creation of the CapsNet classifier Data Mining Classification Techniques,” International
Journal of Computer Science and Information
which turned out to be very accurate in differentiating lung
Technologies, Vol. 4 (1), 2013, 39 – 45.
nodules between benign or malignant ones. The CapsNet
classifier exhibit an impressive accuracy rate of 96%, [4] Astha Pathak, Sunil Kumar Dewangan, Mahendra
Kumar Sahu, M Gayatri, Gitanjali Sahu, Prakriti Verma,
which point to its high effectiveness in accurately detecting "A Survey Based on Machine Learning Algorithm for
cancerous tumors in lung images. Such high accuracy, in Lungs Cancer Prediction", 2023 International Conference
on Artificial Intelligence for Innovations in Healthcare
fact, opens new opportunities to improve early-stage lung Industries (ICAIIHI), vol.1, pp.1-6, 2023.
cancer diagnosis and treatment planning. The results of this
[5] Nitha V. R, Vinod Chandra S. S., "Lung Cancer
project demonstrate the significance of using the most Malignancy detection Using Voting Ensemble Classifier",
advanced image processing and machine learning systems 2023 2nd International Conference on Computational
Systems and Communication (ICCSC), pp.1-5, 2023.
in the medical imaging field for improving disease
detection and enhancing patient care. [6] Mahammad, F. S., & Viswanatham, V. M. (2020).
Performance analysis of data compression algorithms for
heterogeneous architecture through parallel approach. The
6. FUTURE SCOPE Journal of Supercomputing, 76(4), 2275-2288.

Future endeavors in this domain could explore integrating [7] Karukula, N. R., & Farooq, S. M. (2013). A route map
for detecting Sybil attacks in urban vehicular
additional advanced image processing techniques and networks. Journal of Information, Knowledge, and
machine learning algorithms to further enhance the Research in Computer Engineering, 2(2), 540-544.

accuracy and efficiency of lung cancer detection. Research [8] Farook, S. M., & NageswaraReddy, K. (2015).
could focus on optimizing computational efficiency and Implementation of Intrusion Detection Systems for High
Performance Computing Environment Applications. Inter
scalability to enable real-time application in clinical national journal of Scientific Engineering and Technology
settings. Additionally, the development of robust Research, 4(0), 41.

automated systems capable of processing large-scale [8] Sunar, M. F., & Viswanatham, V. M. (2018). A fast
datasets could pave the way for personalized medicine approach to encrypt and decrypt of video streams for secure
channel transmission. World Review of Science,
approaches and improved patient outcomes in the field of Technology and Sustainable Development, 14(1), 11-28.
oncology.
[9] Mahammad, F. S., & Viswanatham, V. M. (2017). A
study on h. 26x family of video streaming compression
REFERENCES techniques. International Journal of Pure and Applied
Mathematics, 117(10), 63-66.
[1] ZakariaSulimanZubi and RemaAsheibaniSaad, “Using
Some Data Mining Techniques for Early Diagnosis of [10] Devi,S M. S., Mahammad, F. S., Bhavana, D.,
Lung Cancer,” Recent Researches in Artificial Sukanya, D., Thanusha, T. S., Chandrakala, M., & [11]
Intelligence, Knowledge Engineering and Data Bases, [17] Swathi, P. V. (2022).” Machine Learning Based
Libya, 2007. Classification and Clustering Analysis of Efficiency of
Exercise Against Covid-19 Infection.” Journal of
[2] Paola Campadelli, Elena Casiraghi, and Diana Artioli, Algebraic Statistics, 13(3), 112-117.
“A Fully Automated Method for Lung Nodule Detection

Page | 106 Copyright @ 2024 Author

Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024
[11] Devi, M. M. S., & Gangadhar, M. Y. (2012).” A Roy, S., Parwekar, P. (eds) Proceedings of International
comparative Study of Classification Algorithm for Printed Conference on Recent Trends in Computing. Lecture Notes
Telugu Character Recognition.” International Journal of in Networks and Systems, vol 600. Springer, Singapore.
Electronics Communication and Computer Engineering, https://doi.org/10.1007/978-981-19-8825-7_68
3(3), 633-641.
[23] Murali Kanthi, J. David Sukeerthi Kumar, K.
[12] Devi, M. S., Meghana, A. I., Susmitha, M., Mounika, Venkateshwara Rao, Mohmad Ahmed Ali, Sudha Pavani
G., Vineela, G., & Padmavathi, M. MISSING CHILD K, Nuthanakanti Bhaskar, T. Hitendra Sarma, “A FUSED
IDENTIFICATION SYSTEM USING DEEP 3D-2D CONVOLUTION NEURAL NETWORK FOR
LEARNING. SPATIAL-SPECTRAL FEATURE LEARNING AND
HYPERSPECTRAL IMAGE CLASSIFICATION,” J
[13] V. Lakshmi chaitanya. "Machine Learning Based Theor Appl Inf Technol, vol. 15, no. 5, 2024, Accessed:
Predictive Model for Data Fusion Based Intruder Alert Apr. 03, 2024. [Online]. Available: www.jatit.org
System." journal of algebraic statistics 13, no. 2 (2022):
2477-2483. [24] Prediction Of Covid-19 Infection Based on Lifestyle
Habits Employing Random Forest Algorithm FS
[14] Chaitanya, V. L., & Bhaskar, G. V. (2014). Apriori vs Mahammad, P Bhaskar, A Prudvi, NY Reddy, PJ Reddy
Genetic algorithms for Identifying Frequent Item journal of algebraic statistics 13 (3), 40-45
Sets. International journal of Innovative Research
&Development, 3(6), 249-254. [25] Machine Learning Based Predictive Model for Closed
Loop Air Filtering System P Bhaskar, FS Mahammad, AH
[15] Chaitanya, V. L., Sutraye, N., Praveeena, A. S., Kumar, DR Kumar, SMA Khadar, ...Journal of Algebraic
Niharika, U. N., Ulfath, P., & Rani, D. P. (2023). Statistics 13 (3), 609-616
Experimental Investigation of Machine Learning
Techniques for Predicting Software Quality. [26] Kumar, M. A., Mahammad, F. S., Dhanush, M. N.,
Rahul, D. P., Sreedhara, K. L., Rabi, B. A., & Reddy, A. K.
[16] Lakshmi, B. S., Pranavi, S., Jayalakshmi, C., Gayatri, (2022). Traffic Length Data Based Signal Timing
K., Sireesha, M., & Akhila, A. Detecting Android Malware Calculation for Road Traffic Signals Employing
with an Enhanced Genetic Algorithm for Feature Selection Proportionality Machine Learning. Journal of Algebraic
and Machine Learning. Statistics, 13(3), 25-32.

[17] Lakshmi, B. S., & Kumar, A. S. (2018). Identity- [27] Kumar, M. A., Pullama, K. B., & Reddy, B. S. V. M.
Based Proxy-Oriented Data Uploading and Remote [18] (2013). Energy Efficient Routing In Wireless Sensor
Data Integrity checking in Public Cloud. International Networks. International Journal of Emerging Technology
Journal of Research, 5(22), 744-757. and Advanced Engineering, 9(9), 172-176.

[19] Lakshmi, B. S. (2021). Fire detection using Image [28] Kumar, M. M. A., Sivaraman, G., Charan Sai, P.,
processing. Asian Journal of Computer Science and Dinesh, T., Vivekananda, S. S., Rakesh, G., & Peer, S. D.
Technology, 10(2), 14-19. BUILDING SEARCH ENGINE USING MACHINE
LEARNING TECHNIQUES.
[20] Devi, M. S., Poojitha, M., Sucharitha, R., Keerthi, K.,
Manideepika, P., & Vasudha, C. Extracting and Analyzing [29] “Providing Security in IOT using Watermarking and
Features in Natural Language Processing for Deep Partial Encryption. ISSN No:
Learning with English Language.
[21]Kumar JDS, Subramanyam MV, Kumar APS. Hybrid 2250-1797 Issue 1, Volume 2 (December 2011)
Chameleon Search and Remora Optimization Algorithm-
based Dynamic Heterogeneous load balancing clustering [30] The Dissemination Architecture of Streaming Media
protocol for extending the lifetime of wireless sensor Information on Integrated CDN and P2P, ISSN 2249-6149
networks. Int J Commun Syst. 2023; 36(17):e5609. Issue 2, Vol.2 ( March-2012)
doi:10.1002/dac.5609
[31] Provably Secure and Blind sort of Biometric
[22] David Sukeerthi Kumar, J., Subramanyam, M.V., Siva Authentication Protocol using Kerberos, ISSN: 2249-9954,
Kumar, A.P. (2023). A Hybrid Spotted Hyena and Whale Issue 2, Vol 2 (APRIL 2012)
Optimization Algorithm-Based Load-Balanced Clustering
Technique in WSNs. In: Mahapatra, R.P., Peddoju, S.K.,

Page | 107 Copyright @ 2024 Author

 Juni Khyat (जूनी ख्यात) ISSN: 2278-4632
(UGC CARE Group I Listed Journal) Vol-14, Issue-4, April: 2024
[32]R sumalatha, dr.m.subramanyam, “image denoising
using spatial adaptive mask filter”, ieee international
conference on electrical, electronics, signals,
communication & optimization (eesco-2015),
organized byvignans institute of information technology,
vishakapatnam, 24 th to 26th january 2015. (scopus
indexed)

[33] P.balamurali krishna, dr.m.v.subramanyam, dr.k.satya

prasad, “hybrid genetic optimization to mitigate starvation
in wireless mesh networks”, indian journal of science and
technology,vol.8,no.23,2015. (scopus indexed)

[34] Y.murali mohan babu, dr.m.v.subramanyam,m.n. giri

prasad,” fusion and texure based classification of indian
microwave data – a comparative study”, international
journal of applied engineering research, vol.10 no.1, pp.
1003-1009, 2015. (scopus indexed)

Page | 108 Copyright @ 2024 Author

View publication stats