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

net/publication/371234314

A Novel Method to Detect Lung Cancer using Deep Learning

Conference Paper · April 2023

DOI: 10.1109/ICICT57646.2023.10134199

CITATIONS READS

0 76

5 authors, including:

Sangeetha Thirumoorthy
Sri Krishna College of Technology
21 PUBLICATIONS 57 CITATIONS

SEE PROFILE

All content following this page was uploaded by Sangeetha Thirumoorthy on 05 June 2023.

The user has requested enhancement of the downloaded file.

 Proceedings of the International Conference on Inventive Computation Technologies (ICICT 2023)
IEEE Xplore Part Number: CFP23F70-ART; ISBN: 979-8-3503-9849-6

A Novel Method to Detect Lung Cancer using Deep

Learning
Sangeetha T Shankar M Sreelekha K V
Department of Information Technology Department of Information Technology Department of Computer Science and
Sri Krishna College of M.P.Nachimuthu M.Jaganathan Engg
Technology Engineering College Hindusthan College Of Engineering And
Coimbatore,India Erode,India Technology
[email protected] [email protected] Coimbatore,India
[email protected]
Indhumathi S Sree Vidhya S
Department of Computer Science and Department of Computer Science and
Engg Engg
Nandha College of Technology Erode Sengunthar Engineering College
Erode,India
2023 International Conference on Inventive Computation Technologies (ICICT) | 979-8-3503-9849-6/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICICT57646.2023.10134199

Erode,India
[email protected] [email protected]

Abstract— Lung cancer is a type of cancer that starts in the formatter will need to create these components, incorporating
cells of the lungs, and it is one of the leading causes of cancer- the applicable criteria that follow. are a type of deep neural
related deaths worldwide. It is often caused by smoking, network that can glancingly learn features from inserted
exposure to radon gas, and exposure to certain toxins and images and classify them with high accuracy. In the case of
pollutants in the environment. Risk factors for lung cancer
include smoking tobacco, exposure to the smoke exhaled by
lung cancer detection, CNNs can analyse CT scan images of
smokers, lung cancer in individual with a family history of the the lungs and point out possibilities cancerous regions. Using
disease, and people have damaged lung, such as COPD. CNNs for lung cancer detection has several advantages.
Diagnosis of lung cancer typically involves a combination of Firstly, CNNs can process large volumes of data quickly and
imaging tests, such as CT scans and X-rays, and biopsy. accurately, making them ideal for inspecting large datasets of
detection of lung cancer allows for more options in terms of medicinal related images. Secondly, CNNs can learn and
treatment and a better chance of curing the cancer. In general, improve their performance over time, as they are trained on
studies on this topic aim to use advanced deep learning more data. In this context, developing a model CNN-based
techniques have the potential to improve the accuracy and lung cancer detection has the potential to revolutionize the
efficiency of lung cancer detection and diagnosis, which can
ultimately lead to better patient outcomes. These techniques
field of lung cancer diagnosis and treatment, by enabling
involves both CNNs and DBNs have shown great promise in the earlier detection and more targeted treatments for patients,
analysis of medical imaging data for the detection and diagnosis which can ultimately lead to better outcomes and improved
of lung cancer, as they are able to automatically identify and quality of life. To develop a CNN-based lung cancer detection
classify patterns in the images that are indicative of lung cancer. model, a dataset of CT scan images of the lungs is required. a
These techniques have the potential to improve the accuracy and dataset should contain both cancerous and non-cancerous
efficiency of lung cancer diagnosis. The recent techniques are images, along with information about the location and size of
GPT-3, Generative adversarial network, Deep reinforcement the cancerous regions.
learning. Here, CNN algorithm is used to detect lung cancer
earlier, they are able to automatically identify and classify The first step in developing a lung cancer detection model
patterns in medical images that are indicative of lung cancer. based on CNN is to pre-process the CT scan images. This
CNN that are particularly well-suited for image analysis involves organise the pixel values of the images and resizing
because they can learn to recognize features in images and gives them to a standard size. The images can then be split into
accuracy that are important for classification and detection. training, validation, and testing sets. Next, the CNN model
architecture needs to be designed. The architecture should
Keywords—component, formatting, style, styling, insert (key include multiple convolutional layers, which are responsible
words) CNN –Convolutional Neural Network, CT – Computed for learning the features of the input images, as well as pooling
Tomography, DBN - Deep Belief Networks, MRI - Magnetic
layers, which reduce the spatial dimensions of the feature
Resonance Imaging
maps. The output of the CNN is then fed into a fully connected
I.INTRODUCTION layer, which classifies the image as cancerous or non-
cancerous. After designing the CNN architecture, the model
However, traditional methods for lung cancer detection, needs to be trained on the training dataset. During training, the
such as X-rays and CT scans are generally considered to be a model learns to classify images by adjusting its weights based
valuable diagnostic tool, they do have some limitations. on the differences between the predicted and actual labels of
Recently, deep learning techniques such as CNNs have the the images. The validation set is used to monitor the
potential to revolutionize the field of medical imaging, by performance of the model during training and to prevent
enabling faster and more accurate diagnoses of lung cancer overfitting. It's worth noting that the accuracy and
and other diseases from imaging data. CNNs parentheses, performance of a CNN-based lung cancer detection model is
following the example. Some components, such as multi- highly dependent on the quality and quantity of the training
levelled equations, graphics, and tables are not prescribed, data, as well as the design of the model architecture and
although the various table text styles are provided. The

979-8-3503-9849-6/23/$31.00 ©2023 IEEE 263

Authorized licensed use limited to: IEEE Customer Ops and Contact Center Staff. Downloaded on June 05,2023 at 11:49:00 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Inventive Computation Technologies (ICICT 2023)
IEEE Xplore Part Number: CFP23F70-ART; ISBN: 979-8-3503-9849-6

training parameters. The model can also be used to predict the staging of lung cancer. The combination of the two modalities
presence of cancerous regions in new CT scan images. allows for more accurate localization of the lesion and also
helps in staging the cancer.
II. RELATED WORKS PET scans can be used in several ways for lung cancer
There are several kind of procedures followed to detect lung detection:
cancer in the literature, including: i. Diagnosis: PET scans can be used in detecting lung
Computed Tomography (CT) scans are commonly used in cancer in individuals who are experiencing symptoms,
detecting lung cancer. CT scans make use of X-ray technology such as a cough or chest pain.
to create thorough images of the lungs, which can be used to ii. Staging: PET scans can be used to determine the stage of
observe lung cancer. [1] CT scans is used to discover small lung cancer, which is important for determining
lung tumors that may not be visible on traditional X-rays. CT treatment options.
scans can also be used to determine the stage of lung cancer, iii. Monitoring treatment: PET scans often used to
which is necessary for determining treatment options. birdwatch the response to treatment and detect
recurrence of the disease.
There are two types of CT scans used for lung cancer It is important to note that PET scans are not recommended
detection: for everyone, especially for pregnant women and patients with
Low-dose CT (LDCT): A low-dose CT scan uses a lower certain medical conditions. Additionally, PET scans may not
amount of radiation than a regular CT scan. This type of CT be available in all areas and may be more expensive than other
scan is typically used for lung cancer screening in highly risk imaging modalities.
individuals, such as current or former smokers. Sputum cytology is a method used for lung cancer
High-resolution CT (HRCT): A high-resolution CT scan uses detection. It involves examining a specimen (mucus and other
a higher amount of radiation than a regular CT scan. This type material upchuck from the lungs) under a microscope to
of CT scan is typically used to diagnose lung cancer in inspect for cancer cells. [4] The sample is collected by having
individuals who are experiencing symptoms, such as a cough the patient cough up sputum into a container. The sputum is
or chest pain. then sent to a laboratory for analysis.
Sputum cytology is a relatively simple and non-
Magnetic Resonance Imaging (MRI) scans are another invasive test that is often used to detect lung cancer in
technology that can be used for lung cancer detection. MRI is individuals who are experiencing symptoms, such as a cough
not commonly used to image the lungs, it can be a useful tool or chest pain. The test can also be used to birdwatch the action
in certain situations and may complement other imaging to treatment and detect recurrence of the disease.
modalities such as CT in the diagnosis and treatment of lung It is important to note that sputum cytology is not always able
cancer and other lung diseases [2]. Unlike X-ray or CT scans, to detect early-stage lung cancer, and a combination of tests
MRI scans do not use ionizing radiation, making them a safer might be used to make a diagnosis. Additionally, the test may
option for certain individuals. not be appropriate for certain individuals, such as those who
There are two types of MRI scans used for lung cancer are unable to produce sputum or those who have had previous
detection: radiation treatment, and alternative tests should be considered.
1) Thoracic MRI: This type of MRI scan focuses on the chest Sputum cytology has a relatively low sensitivity and
and lungs and can be used to detect lung cancer, as well as to specificity, meaning that it may not detect all lung cancers and
evaluate the extent range of the disease and to bird watcher the may produce false positive results. For that reason, it is often
response to treatment. used in combination with other imaging techniques such as
2) Diffusion-weighted imaging (DWI): A subtype of MRI, CT or MRI.
DWI can be used to detect small lesions in the lungs that may Bronchoscopy is a method used for lung cancer
not be visible on regular MRI. It is sensitive to the water detection. It is a procedure a thin, flexible tube called a
diffusion in the tissue and can help to differentiate between bronchoscope is inserted through the nose or mouth and down
malignant and benign tumors. into the lungs. The bronchoscope is usually about the diameter
It is important to note that MRI scans may not be suitable for of a drinking straw, and it is equipped with a light and a
everyone, especially for those with pacemakers, metal camera that allow the physician to see inside the airways and
implants or anaphylaxis to contrast agent. Additionally, MRI lungs and take a sample of any suspicious tissue [5].
scans can be less available or more expensive than other Bronchoscopy can be used in several ways for lung cancer
imaging modalities, and a combination of tests may be used to detection:
make a diagnosis. i.Diagnosis: Bronchoscopy is usually used to detect lung
cancer in individuals who are experiencing symptoms, such
Positron Emission Tomography (PET) scans are a as a cough or chest pain. It can also be used to take samples
technology used for lung cancer detection. PET scans use a of suspicious tissue to be examined in a laboratory.
minimal amount of radioactive material, called a tracer, which ii.Staging: Bronchoscopy can be used to determine the stage
is thrown toward the body. The tracer is taken up by the body's of lung cancer, which is important for determining treatment
cells and emits positrons [3], which are detected by the PET options.
scanner. The PET scanner creates images of the interior of the iii.Biopsy: Bronchoscopy can be used to take a biopsy (a small
body, which is used to detect lung cancer. PET-CT (Positron sample of tissue) from a suspicious area in the lungs. The
Emission Tomography - Computed Tomography) scans are tissue sample can then be inspected under a microscope to
commonly used in clinical practice for the diagnosis and discover if cancer cells are present.

979-8-3503-9849-6/23/$31.00 ©2023 IEEE 264

Authorized licensed use limited to: IEEE Customer Ops and Contact Center Staff. Downloaded on June 05,2023 at 11:49:00 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Inventive Computation Technologies (ICICT 2023)
IEEE Xplore Part Number: CFP23F70-ART; ISBN: 979-8-3503-9849-6

iv.Treatment: In some cases bronchoscopy can be used as a Natural Language Processing: Machine Learning Algorithms
treatment for lung cancer, for example for performing laser could be trained for analyzing the unstructured data, such as
ablation, electrocautery, or brachytherapy. radiology reports, to extract information relevant to lung
Bronchoscopy is a relatively safe procedure, but there are cancer diagnosis.
some risks associated with it, such as bleeding, infection, or It is important to note that machine learning is a relatively new
lung collapse. Additionally, bronchoscopy may not be suitable technology and is still being researched. The accuracy of
for certain individuals, such as those with certain medical machine learning algorithms may vary depending on the
conditions, and alternative tests should be considered. method used, and it may not be appropriate for certain
It is important to note that bronchoscopy is more invasive than individuals. Additionally, machine learning algorithms may
other tests and requires sedation or general anesthesia. not be widely available and may be more expensive than other
Furthermore, the results of the procedure may not be imaging modalities. Furthermore, machine learning models
immediately available and further testing may be needed. need to be validated on multiple independent datasets and
Liquid biopsy is an emerging technology used for should be tested by independent researchers before clinical
lung cancer detection. It is a non-invasive method that use.
involves analyzing a sample of blood, urine, or other bodily X-ray imaging, including chest X-ray, is a widely
fluids to detect cancer cells or biomarkers associated with used imaging modality for lung cancer detection. It involves
cancer. [6] One of the most common forms of liquid biopsy exposing the body to a small amount of ionizing radiation and
used for lung cancer detection is liquid biopsy using blood capturing an image of the internal structures of the body. [8]
samples. This involves analyzing blood for the presence of These are few most common methods for lung cancer
circulating tumor DNA (ctDNA) which is DNA shed by detection in the literature. However, the most appropriate
cancer cells into the bloodstream. This method is known as method for an individual patient will depend on a variety of
liquid biopsy or liquid biopsy using blood samples. factors, including their medical history and the stage of their
Liquid biopsy can be used in several ways for lung cancer cancer.
detection:
i. Diagnosis: Liquid biopsy can be used for detecting lung III.EXISTING SYSTEM
cancer in individuals who are experiencing symptoms, The existing systems for detecting lung cancer, such
such as a cough or chest pain. as chest X-ray, CT scans, PET scans, and sputum cytology,
ii. Monitoring treatment: Liquid biopsy can be used to have several limitations that make them less than ideal for
overview the behavior to treatment and detect recurrence detecting the disease.
of the disease. Chest X-rays may miss small or peripherally located
iii. Early detection: Liquid biopsy may have the potential for tumor , and CT scans have a relatively high risk of false
early detection of lung cancer, especially for individuals positives.
who are at high risk for the disease. PET scans are expensive and not widely available, and
iv. Minimally invasive: Liquid biopsy is a non-invasive sputum cytology may be difficult to obtain a sample from
method that doesn’t require surgery or radiation and has some patients. Additionally, each of these tests has a
a lower risk of complications. relatively high rate of false negatives, meaning that they may
It is important to note that liquid biopsy is a relatively new not detect cancer even when it is present.
technology and is still being researched. The sensitivity and These limitations can make it crucial for doctors to
specificity of liquid biopsy may vary depending on the method accurately diagnose and treat lung cancer in its early stages,
used, and it may not be appropriate for certain individuals. which is essential for improving patient outcomes
Additionally, liquid biopsy may not be widely available and .Furthermore, the high radiation exposure associated with CT
may be more expensive than other imaging modalities. scans and the high cost of tests like PET scans can also create
Machine learning is an emerging technology that is barriers for certain populations to access these methods.
being used for lung cancer detection. It involves training a Additionally, the false positives and false negatives of these
computer algorithm to recognize patterns in medical imaging tests can lead to unnecessary anxiety and further testing for
data and make predictions about the presence of cancer. [7] patients, creating additional burden on the healthcare system
Machine learning can be used in several ways for lung cancer and the patients.
detection: Another issue with the current systems is that they
1) Computer-aided diagnosis (CAD): Machine learning are not tailored for screening highly risk populations, such as
algorithms can be trained to analyze medical images, such as smokers and those with a family history of the disease. These
computed tomography (CT) scans, and identify areas that are populations are more likely to develop the disease and would
suspicious for lung cancer. This can be used as an aid for benefit from more frequent screenings, but the existing
radiologists in detecting lung cancer. systems may not be efficient or cost-effective for this
2) Radiomics: Radiomics is the extraction of quantitative purpose. In conclusion, while the existing systems for
features from medical imaging, which are then used as input detecting lung cancer have some advantages, their limitations
to a machine learning algorithm. Radiomics can be used to make them less than ideal for detecting the disease in its early
classify and predict outcomes of diseases, such as lung cancer. stages. The high cost, radiation exposure, and the high rate of
Image segmentation: Machine learning algorithms could be false negatives and false positives make it crucial for doctors
trained to automatically segment the lung and lung lesions in to accurately determine and treat lung cancer, which is
CT images, which improves the efficiency and accuracy of essential for improving patient outcomes.
lung cancer determination.

979-8-3503-9849-6/23/$31.00 ©2023 IEEE 265

Authorized licensed use limited to: IEEE Customer Ops and Contact Center Staff. Downloaded on June 05,2023 at 11:49:00 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Inventive Computation Technologies (ICICT 2023)
IEEE Xplore Part Number: CFP23F70-ART; ISBN: 979-8-3503-9849-6

treated. This can lead to a reduction in false negatives, which

is particularly important for early detection of lung cancer.
A. Chest X-Ray
Additionally, deep learning algorithms can also be trained to
Advantages: detect lung cancer on low-dose CT scans, which can reduce
1. Relatively inexpensive radiation exposure for patients. This is particularly important
2. Widely available for high-risk populations, such as smokers, who may need to
3. Can detect lung cancer in its early stages. undergo multiple screenings.
To perform a comparative analysis of the proposed model of
Disadvantages: lung cancer detection using deep learning and existing
1. Not as accurate as other imaging techniques techniques:
2. May miss small or peripherally located tumors. 1.Accuracy
2.Speed
B. Computed Tomography (CT) Scans 3.Training Data Requirements
Advantages: 4.Interpretability
1. Can detect small tumors not visible on X-ray. 5.Generalizability
2. Can also detect abnormalities in the lymph nodes and
other organs. Overall, the proposed model of lung cancer detection
using deep learning has the potential to achieve high accuracy
Disadvantages: and generalize well to new data. However, it may require
1. Expensive significant computational resources and labeled training data.
2. High radiation exposure Interpretability and generalizability may also be challenges.
3. Risk of false positives Existing techniques such as CAD systems may be less
resource-intensive but may also suffer from generalization
C. Positron Emission Tomography (PET) Scans and interpretability issues.
Advantages:
1. Can detect early-stage lung cancer.
2. Can also detect metastases to other organs.

Disadvantages:
1. Expensive
2. Not widely available
3. Risk of false positives.

D. Sputum Cytology
Advantages:
1. Inexpensive
2. Non-invasive
3. Can detect cancer at an early stage

Disadvantages:
1. Not as accurate as other techniques
2. May miss small tumor or tumor located in the
peripheral airways
May be difficult to obtain a sample from some patients.
It's worth noting that no single test is perfect for detecting
lung cancer, and a combination of tests is often used to
improve accuracy. A thorough physical examination,
including a chest X-ray or CT scan, is routinely the first step
in detecting lung cancer. If an abnormal area is found,
further testing, such as a biopsy or PET scan, may be
performed to confirm the diagnosis.

IV. PROPOSED SYSTEM

Our proposed system of lung cancer detection using Fig.1. Architecture Diagram for proposed system
deep learning has the potential to be a more accurate and In conclusion, the proposed objective of lung cancer
efficient method for detecting the disease compared to the detection using deep learning is to develop an accurate and
existing systems. efficient system that can identify the presence of lung cancer
Deep learning algorithms, such as convolutional from medical images. This system would use deep learning
neural networks (CNNs), are a promising tool in the fight algorithms, such as convolutional neural networks (CNNs),
against lung cancer, and they have the potential to to analyze medical images and identify any abnormalities that
revolutionize the way that lung cancer is diagnosed and

979-8-3503-9849-6/23/$31.00 ©2023 IEEE 266

Authorized licensed use limited to: IEEE Customer Ops and Contact Center Staff. Downloaded on June 05,2023 at 11:49:00 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Inventive Computation Technologies (ICICT 2023)
IEEE Xplore Part Number: CFP23F70-ART; ISBN: 979-8-3503-9849-6

could indicate the presence of lung cancer. The system would without noise, clears blurness and color conversion. Chosses
be trained on a large dataset of medical images and would use filepath for functions.
transfer learning to leverage pre-trained models for feature Here, x length for feature and y counts for label,
extraction. The challenges faced by such a system include the which appends the training and testing datasets. Length is
need for high-quality medical images, the requirement for a 1097 and y counts for category1 is 561, category2 is 416,
large dataset for training, and the need for interpretability of category3 is 120, which spited as train data and test data. Here
count for train data is 1042 and testing data is 55.
the results to ensure patient safety.

There are two phases in the proposed system namely C. Feature Extraction
1. Training phase
2. Result phase
In Training phase, trains the system. In Result phase,
classifies the output by comparing the given input model and
trained model. We use google collabs for implementation.

A. Important Files
Import library files from googlecollab. By uploading
Kaggle.json file which helps to store large dataset and reads
it. Dataset file name “iqothnccd-lung-cancer-dataset.zip”.
After downloading, it extracts file from dataset to train.
Installing tensor flow to train system, which also helps in data Fig .5. Image Preprocessing
automation, model tracking and performance monitoring. When it comes to append we need to get accurate value,
Installed tensorflow-addons-0.10.0. because of the count difference in testing and trained data
causes inappropriate accuracy. So reshaping occurred.
SMOTE is a popular method for handling class imbalance in
datasets by oversampling the minority class. It creates
synthetic samples by interpolating between existing minority
samples, thereby increasing the size of the minority class.
Using these techniques can help to improve the performance
of machine learning models on imbalanced datasets, and
reduce bias towards the majority class. This is especially
important in medical applications, such as lung cancer
Fig.2. Architecture Layers
detection, where accurate classification of both positive and
negative cases is crucial. Therefore it is used in
implementation to balance the testing database categories,
which replicates the testing samples to be balanced.

E. Sequential Values
Setting sequential to make loop of function. To make loop,
conv2d, maxpooling2d, relu activation function and
normalization is used.
f(x) = max(0,x)
Fig. 3. Data Sizing For output shape flatten, dense, dropout, dense, softmax
activation function is used.
B. Training System With Testcase 𝑒 𝑧𝑖
𝜎(→)𝑖 = 𝐾
Import packages from their respective modules. setting 𝑧 ∑𝑗=1 𝑒 𝑧𝑗
directory for the categories ‘Bengin cases’, ‘Malignant 𝜎 = softmax
cases’, ‘Normal cases’. Sizing data for the cases , Setting file → = feature vector
𝑧
path to reach categories. The total test cases given to train 𝑒 𝑧𝑖 =standard exponential function for input vector
data is 1097.
𝑒 𝑧𝑗 = standard exponential function for output vector
K=number of classes in the multi-class classifier

soft argmax or normalized exponential function, converts a

vector of K real numbers into a probability distribution of K
possible outcomes.

Fig. 4. Three Cases of Lung cancer

It sets categories of 0 for begin cases, 1 for malignant
cases and 2 for normal cases. Pre-processing the image like,
reading, resizing and gaussianblur. Reading the images ,
Resizing the image, and gaussianblur which makes the image Fig .6. SMOTE

979-8-3503-9849-6/23/$31.00 ©2023 IEEE 267

Authorized licensed use limited to: IEEE Customer Ops and Contact Center Staff. Downloaded on June 05,2023 at 11:49:00 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Inventive Computation Technologies (ICICT 2023)
IEEE Xplore Part Number: CFP23F70-ART; ISBN: 979-8-3503-9849-6

In fig .6., shows the count datas before SMOTE and after Saves the model which is appended to trained model. Predicts
SMOTE. the case by classifying using CNN and verifies using
confusion matrix. Predicts case with the accuracy.
𝑇𝑃+𝑇𝑁
Accuracy =
𝑇𝑃+𝑇𝑁+𝐹𝑃+𝐹𝑁
𝑇𝑃
Precision =
𝑇𝑃+𝐹𝑃
𝑇𝑃
Recall =
𝑇𝑃+𝐹𝑁

Fig .9. FINDING ACCURACY

Fig .7. parameters

Total params : 465,775
Trainable params : 464,809
Non – trainable params : 966
Fig. 10. final output
F. Prediction
Setting compiling loss value as The total cases are three cases which is in fig4.
sparse_categorical_crossentropy. When an input given and he image got processed and gives
Here batch size = 25, epochs = 10. the final prediction as Begin cases with the accuracy of
19.30562.

RESULT ANALYSIS
110

60

10
ACCURACY PRECISION RECALL F1-SCORE

CNN LSTM RNN GAN

Fig .8. EPOCHS Fig .11 .Result Analysis

Setting increased epochs gives higher possible accuracy and

In fig .11 show that CNN gives the more accuracy than the
gives loss value.
other algorithms compared. Long Short Term Memory
• True Positive is denoted as TP[Malignant] Networks (LSTMs), Recurrent Neural Networks (RNNs) and
• False Positive is denoted as FP[Malignant] Generative Adversarial Networks (GANs) . other than these
• True Negative is denoted as TN[Malignant] algorithms CNN works well. The
• False Negative is denoted as FN[Malignant] Convolutional Neural Network (CNN or ConvNet) is a sub

979-8-3503-9849-6/23/$31.00 ©2023 IEEE 268

Authorized licensed use limited to: IEEE Customer Ops and Contact Center Staff. Downloaded on June 05,2023 at 11:49:00 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Inventive Computation Technologies (ICICT 2023)
IEEE Xplore Part Number: CFP23F70-ART; ISBN: 979-8-3503-9849-6

type of Neural Networks that is mainly used for applications [9] A Deep Learning Algorithm for Automated Pulmonary Nodule Detection
in image and speech recognition. Its built-in convolutional in CT Scans by X. Chen, Y. Liu, Z. Wang, et al. (2017).
layer reduces the high dimensionality of images without [10] Chest X-Ray and CT Image Classification using Convolutional
losing its information. That is why CNNs are especially Neural Networks: A Review by A. Noailly, J. Desbos, V. Clairand, et
suited for this use case. al. (2018).
[11] Early Detection of Cancer: Progress and Promise" by H. Gilbert Welch,
V. CONCLUSION M.D., M.P.H. and Douglas W. Wilbur, M.D., M.P.H. Published in The
New England Journal of Medicine in 2018.
In conclusion, the use of Convolutional Neural
[12] Artificial Intelligence in Medical Imaging: Past, Present, and Future" by
Network (CNN) algorithms in deep learning has Ge Wang, Ph.D. and Dinggang Shen, Ph.D. Published in IEEE
demonstrated potential in detecting lung cancer. These Transactions on Medical Imaging in 2018.
algorithms can analyze medical images with high accuracy
and detect signs of the disease, leading to early diagnosis and [13] Circulating Tumor DNA for Early Cancer Detection" by Ming-Sound
Tsao, M.D. and Carlo M. Croce, M.D. Published in Cancer Discovery
better patient outcomes. However, it is crucial to
in 2016.
acknowledge that further research and clinical validation is [14] Nanotechnology for Early Cancer Detection" by Xiaoyuan Chen,
required before widespread implementation. These Ph.D. and Weihong Tan, Ph.D. Published in Nature Reviews Cancer in
algorithms should not be solely relied upon for diagnosing 2016.
lung cancer and should be used in conjunction with other [15] Noninvasive diagnosis of cancer by liquid biopsy" by M. Jorge Cardoso,
medical diagnostic tools. M.D. and Christophe Le Tourneau, M.D., Ph.D. Published in Nature
Reviews Clinical Oncology in 2017.
[16] Deep Learning for Lung Cancer Detection in CT Images: A
VI. FUTURE SCOPE Comparative Study" by Chen Chen, Wei-An Lin, and Min Sun.
Deep learning has shown promise in detecting lung Presented at the International Conference on Medical Image Computing
cancer, particularly in the analysis of medical imaging such and Computer-Assisted Intervention (MICCAI) in 2018.
as CT scans. In the future, it is likely that the use of deep [17] A Deep Learning Framework for Automated Prostate Cancer Detection
learning in lung cancer detection will continue to evolve and in Multiparametric MRI" by Yiqiu Shen, Xin Yang, and Guang-Zhong
improve, potentially leading to more accurate and efficient Yang. Presented at the International Conference on Medical Image
Computing and Computer-Assisted Intervention (MICCAI) in 2018.
diagnostic tools. Additionally, the integration of deep [18] Automatic Detection of Pancreatic Cancer in CT Scans using Deep
learning with other technologies such as genomics and Learning" by Yefeng Zheng, Xiaojuan Qi, and Jieping Ye. Presented at
proteomics may lead to a more comprehensive understanding the Conference on Computer Vision and Pattern Recognition (CVPR) in
of lung cancer and improved treatment options. Additionally, 2019.
the use of deep learning models on large scale datasets and [19] A Deep Learning Framework for Automated Detection of Polyps in
with other modalities such as X-ray, MRI, and PET scans will Colonic CT Scans" by Jingting Liang, Fan Wang, and Xiaowei Zhou.
Presented at the Conference on Medical Image Computing and
increase the accuracy and improve early detection.
Computer-Assisted Intervention (MICCAI) in 2019.
Furthermore, the integration of these models with Electronic [20] Multi-Task Deep Learning for Early Prostate Cancer Detection in Multi-
health records and other clinical data will enable better Parametric MRI" by Hao Chen, Yuxing Tang, and Lei Wang. Presented
patient stratification and personalized treatment. However, at the Conference on Medical Image Computing and Computer-Assisted
it's important to note that more research is still needed to Intervention (MICCAI) in 2020.
develop and validate these methods before it could be usually [21] Sangeetha, T. and Mohanapriya, M., 2022. A Novel Exploration of Plant
Disease and Pest Detection Using Machine Learning and Deep Learning
used in clinical practice.
Algorithms. Mathematical Statistician and Engineering Applications,
REFERENCES 71(4), pp.1399-1418.
[22] T Sangeetha, M Kumaraguru, S Akshay, M Kanishka –“Biometric
[1] V. Krishnaiah, G. Narsimha and C. Subhash, "Diagnosis of lung cancer based finger print
prediction system using data mining classification technique", verification system For ATM machines" Journal of Physics:
International Journal of Computer Science and Information Conference Series, 2021.
Technologies, vol. 4, no. 1, pp. 39-45, December 2013. [23] G Lavanya, T Sangeetha, P Alaguvathana, G Prasanna “Kernel-based
[2] C. Anita and S. S. Singh, "Lung cancer detection on CT images by using Attribute-aware Self adaptation and Multi thresholding for Rating
image processing", International Conference on Computing Sciences, Prediction“ IOP Conference Series:Materials Science and ..., 2021
2012. [24] Sangeetha, T., Mohanapriya, M., Pavithra, S., Ragamira, S. and Sneha,
[3] K. Murphy, B. Ginneken, A. M. R. Schilham, B. J. Hoop, H. A. Gietema S., 2022. A Novel Deep LearningApproach for Alzheimer’s Disease
and M. Prokop, "A large-scale evaluation of automatic pulmonary nodule Segmentation and Classification Using RCNN.Mathematical
detection in chest CT using local image features and k-nearest-neighbour Statisticianand Engineering Applications,71(3), pp.1159-1172.45.
classification", Medical Image Analysis, vol. 13, no. 5, pp. 757-770, [25] K.Mythili ,S.Muthilakshmi, Dr.T.Rajesh Kumar, T.Sangeetha
2009. “SimilarityDiseasePredictionSystemforEfficientMedicare”-Publication
[4] T. Atsushi and F. Hiroshi, "Fast lung nodule detection in chest CT images date 2020/4 Test Engineering & Management Volume 83 Issue
using cylindrical nodule-enhancement filter", International journal of ISSN:0193-4120 Pages 3s350 -3354 Publisher the Mattingley
computer assisted radiology and surgery, vol. 8, pp. 193-205, 2013. Publishing Co., Inc.
[5] J. R. Sousa, A. C. Silva, A. C. Paiva AC and A. N. Rodolpho, [26] T Sangeetha, G Lavanya, D Jeyabharathi, T Rajesh Kumar, K.Mythili,
"Methodology for automatic detection of lung nodules in computerized “Detection of Pest and Disease in Banana Leaf Using Convolution
tomography image", 2010, vol. 98, no. 1, pp. 1-14.s. Random Forest” Test Engineering and Management, Volume 83, Page
[6] Lung Nodule Detection in CT Images with Convolutional Neural Number: 3727 -3735,March -April 2020
Networks by J.L. Bieniek, W.E. Lorensen, P.J. Arbelaez, et al. (2016). [27] Sangeetha, T., Miruthula, V., Kavimalar, C., Aakash, V. (2023). Face
[7] Automated lung nodule detection in CT images: A review of deep Mask Detection and Social Distancing Using Machine Learning with
learning algorithms by J. Li, Q. He, D. Wang, et al. (2020). Haar Cascade Algorithm. In: Shakya, S., Du, KL., Ntalianis, K. (eds)
[8] Deep learning for lung cancer screening: Opportunities and challenges Sentiment Analysis and Deep Learning. Advances in Intelligent Systems
by M.F. Gulsun and M.A. Erho (2019). and Computing, vol 1432. Springer, Singapore.
https://doi.org/10.1007/978-981-19-5443-6_72

979-8-3503-9849-6/23/$31.00 ©2023 IEEE 269

Authorized licensed use limited to: IEEE Customer Ops and Contact Center Staff. Downloaded on June 05,2023 at 11:49:00 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Inventive Computation Technologies (ICICT 2023)
IEEE Xplore Part Number: CFP23F70-ART; ISBN: 979-8-3503-9849-6

[28] Sangeetha, T., Sangavi, R., Shinega, P. A., & Shobana, G. (2022, Intelligent Control, Environment, Computing & Communication
December). Developmentof Noninvasive Continuouspatient Glucose Engineering (ICATIECE) (pp. 1-6). IEEE.
Bottle Level and Health Care Monitoring System Using IoT. In 2022
Second International Conference on Advanced Technologies in

979-8-3503-9849-6/23/$31.00 ©2023 IEEE 270

Authorized licensed use limited to: IEEE Customer Ops and Contact Center Staff. Downloaded on June 05,2023 at 11:49:00 UTC from IEEE Xplore. Restrictions apply.

View publication stats