International Journal of

INTELLIGENT SYSTEMS AND APPLICATIONS IN

ENGINEERING
ISSN:2147-67992147-6799 www.ijisae.org Original Research Paper

Deep Learning Methodology of Lung Cancer Detection and Diagnosis

Using CT Images: A Systematic Approach

Sadhana B.1, Dr. Pramod Kumar Naik 2

,

Submitted: 28/01/2024 Revised: 06/03/2024 Accepted: 14/03/2024

Abstract: In the present era lung nodule is the very dangerous and deadly cancer disease requires initial diagnosis which improves
patient survival probability. Many techniques have played the major and important role in the medical field in detection of lung cancer by
analyzing lung medical image. Here we are carried out a systematic article surveys of different article published from last five years. The
four different databases like (Science Direct, Scopus, web of science, and IEEE), used during last five years and chosen around 15
articles to carry out systematic survey on lung cancer detection. The major goal of the survey work is to consolidate and concise recent
advancement in lung cancer detection medical field, diagnosis considering various detection algorithms and methods. This article concise
and summarizes the deep knowledge by addressing the results and findings of recent research which enhances and provides sufficient
knowledge in the relevant field. We included challenges, applications, and recommendation for further enhancements after analyzing
different research articles in detail. The lung image screening, diagnosis of cancer using CT imaging, in which detailed scanned images
of lung is included. Further enhancements by using of CAS (“computer-assisted systems”) DL concepts were promoted to interpret lung
nodule detection of CT images. The goal of this article is to cover the over-view related to DL techniques, detection of lung cancer using
DL techniques and its related applications and benefits. This paper mainly focuses on two different methods of DL while screening and
diagnosis of lung cancer, like classification and segmentation methods. The shortcomings and benefits, advancement of DL models are
also be elaborated, the analysis picturises and demonstrates there is a potential significance of use of DL techniques to enable and
provide accurate, precise, effective- computer assist for lung cancer diagnosis and screening by considering CT images. The part of this
review article a set of potential recommendations for future scope and work which improves available applications of DL methods to
enhances the better and efficient accurate results in identification of lung cancer.

Keywords: Lung Cancer Detection,Image Pre-Processing, Image Segmentation ,Convo-lution Neural Network.
1. Introduction habits and tobacco consumptions remains the major factor for
cancel developments also nontobacco risks include infections in
As per American cancer society estimations for lung cancer
lung, lifestyle food habits, chronic lung infections etc. Avoidance
during 2023 in US there are 238340 new lung cancer patients out
of tobacco consumptions may reduce the risk of lung cancer
of which 117550 are men and 120790 are women. And the
along with healthy diet plans, physical activities etc. [2].
mortality counts due to lung cancer in men 67,160 and in women
59,910. Despite of covid-19 pandemic, in considerations of other
reasons of death rate, the cancerogenic death- rate has continued
in decline from 2019-2020 by 1.5%, 33% overall reduction from
1991.Because of advancements in treatment techniques there is a
rapid decline in the mortality (2% annually during 2016-2020) for
kidney cancer, melanoma, leukaemia etc., and moderate declines
in the case of lung cancer. In contrast there is rapid decline in -the
mortality ratio because of recent advancements in the clinical
treatments, [1].
In defiance of advances in understanding the risk factor
developments in immunization control, treatments for lung cancer
options, it remained the major cause for human death. Smoking

1 Research Scholar DSU Bangalore, Assistant professor Fig .1. cancer incidence trends (1975–2019) mortality (1975–
Department of ISE, CEC Mangalore, INDIA, 2020wrt sex, US
ORCID ID : 0000-0003-1426-6507
The author has proposed an efficient and novel computed aided
2 Associate Professor, Department of CSE, DSU BANGALORE
automatic nodule detection which can minimize false positive
INDIA
results. To avoid juxta –pleural –nodules and automatic wall
ORCID ID : 0000-0001-5767-256

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 713
mending technique was introduced. The pre-processing method The enhanced DL models doesn’t need any tagged practice of the
adapted in this study was vascular elimination, which can data. The model will analyse data’s inheritance features
enhance and highlights nodule portions by weakening the vessels. considering few pertinent features and characteristics. This type
Four types of CNN structures based on the 4 nodules levels were of models were best suited for feature reduction and clustering
applied. Two distinct group’s forms pairs of images like group 1 methods. The commonly used learning models here are Auto-
and group 2 were the inputs to the CNN model. In comparisons Encoders (AE) and Restricted Boltzmann Machines (RBM)
with the Convention CNN model, the adapted method can Semi-Supervised Deep Learning Models:
simplify the data, and provides better results by considering
These types of learning models can use both labelled and
merits of two group correlations.
unlabelled data, RNN, LSTM, GRU, GAN are the popular
By considering diagnosis information and data which will be learning models.
provided to the radiologist and clinicians, helps them to predict
Reinforced Deep Learning Models:
structure of lung nodule
This learning model focuses more on chooses best
courses of actions to reduce the rewards, and learn and perform
behavioural study by considering the interactions with outside
world. The DL is a branch of which uses multiple layered
complex network structure to extract and learn the features from
enormous amount of dataset. These network models have the
capabilities to detect the linkage and patterns within the complex
data for conventional ML, since they can learn automatically the
complex features from different category of medical images like
2D CT images, 3D CT images, MR images etc.
The CT scanned images need more detailed lung
images as they are most oftenly used in lung cancer
identification. The detailed verification of shape, size, texture,
s.
brightness and intensity of the images, Deep learning methods
can be used to identify and recognize the defected lung nodules.
Deep learning methods uses complete images of the lung
An automated leaning based – nodule detections help the capacity, 3D CT image offers more accuracy in results compared
radiologist to early detect and identify relevance information’s with the 2D CT images.3D CT images exhibits high accuracy to
which are un covered from the big data.[3]. detect the abnormalities. To reduce the exposure of radiations
while screening lung cancer detection, low –dose CT
2.1 Deep Learning Methods Overview
methodologies are more often, which may be coupled with
Currently collecting lung images and interpretations of images for filtering, augmentation of image techniques. But MR images has
diagnosing is a challenging task. The high-resolution image more capabilities to reveal the tissue information’s, blood flow.
capture techniques are available like MRI, XRAYS, CT scans etc.
Based on the observed features DL methods can
After lung image pre-processing step, the model extracts relevant analyse lung images to locate lung nodule including additional
features and information’s from the images and build a training
abnormalities. To increase the image quality, reduction in noise
model there by unknown image can be tested for cancer
filtering methods like Gaussian, Median filters are implemented
detection. The traditional cancer detection technologies were un-
in the pre-processing stages. Region proposal and sliding window
able to detect the lung cancer and to produce reliable results.
techniques were adapted in the case of deep learning approaches.
Deep-learning methods were successfully introduced and
developed which will effectively diagnose the disorders.
The techniques under deep-learning are the subset of ML, that
will enable estimation of outcomes based on provided dataset
using the training model. The DL techniques uses NN with many
layers like input, hidden, output layer. Based on the learning
techniques and methods the outcome of the model can be
validated. The deep learning models may be either i) reinforced
learning ii) unsupervised learning models iii) semi-supervised
learning models iv) supervised- learning models.
Supervised Deep Learning Models:
During the initial training phase for SDL type of learning model,
labelled and specified data is needed. The DL model practices the
training phase by considering every input combination which
covers the target class labels. The list of training models was
developed in order to forecast the labels of unknown samples,
most frequently used DL models may be CNN, LSTM, RNN, and
GRU.
Fig.2. Cancer detection process
Unsupervised Deep Learning Models:

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 714
 Unfortunately, disease detection with patient of stage 4 The author has developed an enhanced model for segmenting the
is associated with less survival rate and high symptoms. Early lung CT images by combining kernel graph cut algorithm and
detection of cancer disease may reduce the mortal ration. The proposed a mathematical model. In the next stage the proposed
initial standard care for stage I and stage II including patients algorithm was compared with K –mean algorithm, cluster
with stage IIIA, NSCLC is a surgical extirpation. The surgical variance algorithm and concluded with the comparative studies
removal of cancer cells follows some systematic procedure and between two approaches [8].
therapies. The advantage of adjuvant cancer therapy varied from In the proposed research article author has developed a CAD
stage to stage includes stage 1B (Tumor size is greater than equal based system which detects cancer from lung CT images. This
to 4cm), with a 3% reduction of risk of death during first 5 years. research has adapted four steps like pre-processing, segmentation,
The usage of chemotherapy for stage III patients increased to classification, and the validation. In the first step filtering process
around 13% comparatively. was carried out to enhance the input CT lung image followed by
This author stated that the CAD model designed is helpful to segmentation of blood vessels and pulmonary nodules by
detect cancer in the early stage using CT lung images. This author applying double level thresholding, morphological operations. In
has adapted four steps to detect cancer cells in the CT images the next step features can be extracted from the segmented lung
taking in to account of medical professional’s challenges to images which are HOG feature VH features, statistical features of
diagnose the lung cancer nodules from CT images. The first step first, second order GLCM etc. Fourth step is to get the better
filters the noisy contents by applying filtering concepts either by accuracy using MFFNN, radial basis NN, SVM.The validation
using Weiner Filter, Median, Gabor filter, Min and Max filter etc. process was carried out and the results obtained was CAR is -
In the preceding step thresholding like multiple and optical 99.06%, S is -100% and SP is -99.2%. As per this article still
thresholding, active contour method, shape-based method etc there is a scope in improvements in early detection and
were used. Segmentation was introduced to segment suspected classification of benign and non-benign tumor. [9]
nodules. The third step is the feature extraction like geometric The proposed article focused on how lung nodules can be
features, shape –size features, gray scale features and statistical diagnosed using skeletonization and Gini coefficient techniques
features need to be extracted from the CT lung images to detect which produces lung images as class malignant or class benign
malignant or non-malignant. To extract the features various from the Data set. The Gini coefficient technique helps to identify
classifiers like vector machine, generic algorithm, ANN. Rule the nodule distribution and the skeletonization helps to analyse
based classifier, Linear Discriminate Analysis methods were the shape of the lung nodules. By considering these two
adapted. As concluded by the research author still there is more parameters research author has proposed a prototype which
of improvements required with respect to sensitivity, accuracy performs discriminant analysis malignant or else benign. The
and specificity in the existing deionisation system. [3] results validation was depending on classification methodology
T Aggrawal et al. (2015) [4], in the proposed research paper the and ROC curve. [10].
classification of cancer nodules, the author has introduced a In the present medical era lung cancer has become the major
model which uses gray scale characteristics and valuable health challenge in human body. Early-stage detection and
thresholding values to perform segmentation process of lung diagnosing of lung cancer is quite difficult with less rate of
nodules by threshold values, Gray scale characteristics to perform survival, risk factor is high if it is not early detected and
segmentation process. The proposed system achieved accuracy, diagnosed. Different levels of cancer treatments are available in
sensitivity, Specificity of 84%, 97.14% and 53.33%. the medical and hospital science like chemotherapy, radiography,
The research author has proposed a model in which CT images surgery depending upon the stages of the lung cancer. The
were pre-processed to remove the noise. In the next step fuzzy k- survival rate is 14% for 5 years since cancer develops with in the
mean algorithm was used to perform segmentation process, later respiratory system epithelium and bronchial trees of the lung and
on the result was improved by k mean approach. In the third step which spread across the human body. [11]
feature extraction was carried out from the CT lung images like Early-stage detection of lung cancer required a powerful
correlation, entropy SSIM, PSNR, homogeneity using statistics technology to assist experts provide a desirable treatment. The
GLCM feature extraction method. Finally, categorization is techniques like ML techniques, artificial intelligence techniques
carried out using supervised NN like BPNN for the detection of and image processing can process the lung cancer images, if
lung cancer. This author has achieved 90.7% of accuracy but this necessary, pre-processing and training of medical CT and X Ray
accuracy can be enhanced using improvised classification images by adapting NN, ML techniques.[12]
technique like support vector machine. [5]
The researcher has developed a prototype to improvise lung
This author has proposed CNN model as the classification cancer detection using Computed-Tomography lung images, are
approach in the CAD system and achieved 84.6%, 82.5% of mainly best suited for detection of pulmonary the nodules in the
accuracy and sensitivity and specificity of 86.7% [6]. Lung cancer. [13, 14].
The author has proposed data mining methods, lung cancer In the present era lung cancer is the major cause which reflects
patient database includes images of human upper half body X- modern lifestyle and contamination and colossally expansion of
rays that classifies like normal or malignant, benign. CAD system lungs issue. The lung illness treatments corresponding to the
uses pattern recognition, feature extraction also classification image processing techniques using CAD frameworks
process. The proposed model used X-ray images and accuracy implementations. In this prototype initially pre-processing was
gained is less comparing with CT images. This study elaborates carried out by highlighting the features from CT lung images.
that to achieve better accuracy implementation work can be This researcher has adapted advanced image processing, machine
enhanced and extended to apply the CT images in superiors’ learning technologies in the implementation step, later on
diagnosis of detection of lung cancer in the human body.[7]
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 715
comparison analysis of diverse classification mechanisms was it is difficult that lung nodule detection using CT scanned images.
carried out which will help to improve % accuracy while Even CAD system is also a crucial for the detection, hence image
detecting lung cancer and suggest the use of advanced processing techniques is basic activity that is to be carried out in a
classification and segmentation technologies. wide range of medical scenario. This method proposed and
To achieve the set objectives the researcher has used an effective demonstrated accurate prediction and classification technique
segmentation techniques like thresholding, ANN, SVM which enables enhanced results with the help of image processing
classification techniques. But the obtained results can be further and machine learning techniques. Lung images were collected
enhanced by improving the accuracy while detecting lung cancer and pre-processed using geometric mean filter to increase the
to avoid false positive results because cancer in lung may be quality of the images. K mean approach was used to segment the
treat-able only it is detected in the initial stages. As per the cancer enhanced images, in the next step ML based categorization
study lung cancer doesn’t exhibits early stage and initial algorithms were used for classification purpose [17].
symptoms and noise CT images doesn’t give exact results in Lung cancer staging and its investigation in the initial stage is the
CAD system. The proposed system eliminates the risk and basic and major predictor of survival since it determines the
challenges by using filtering techniques like Auto encoder treatment options. Using CT images there are much computer-
system, segmentation and classification techniques like OTSU aided diagnostics were developed and this majorly concentrates
algorithm, decision tree, CNN. The researcher has clearly on CNN-based technology to detect various stages of lung cancer.
discussed the LDD and how progressing can fulfil the medical The proposed work focused on the development of a model by
filed using R-CAD (Robust) system to overcome the challenges which an improved rate of accuracy classifications of lung
of future work. [15]. nodules and their analysis can be achieved.
Medical cancer detection tools are more essential in initial stage Datasets
diagnosation and to monitor lung cancer while treatment. For The usage of available datasets plays an important role in
lung cancer detection different medical image modalities like MR automatic lung nodule detection as well as classification. To
images, X-Ray images, computed and position emission attain accurate and correct performance outcomes using any
tomography etc. have been extensively used. The medical computational techniques the availability of related datasets is an
techniques adapted have some major limitations like automatic important parameter. Presently public availability of dataset
classification of lung cancer nodules is the major challenge. Deep usage has high impact fir detection, classification and
learning technique is the quick growing concept in medical identification of lung cancer. The Cancer detections includes
imagining area by considering emerging applications like textual identification and distinguish between presences of nodules in the
and image-based modalities. The author has proposed a deep lung image, classification includes separation of cancer nodules.
learning image-based system for advance detection of the lung CAD system allows timely identification of lung cancer and early
nodules. The recent achievements in the deep learning approaches diagnosis for the same by using CT images using advanced
for lung cancer detection segmentation and classification has techniques like AI approach. This detection system using CAD
proven a wide variety of techniques to achieve enhanced based decision engine, which examines the CT images as an input
accuracy. Most of the system uses CT image datasets to train the image, considering different methodologies for image
networks and clinical and experimental results demonstrates deep segmentation and classification of medical images. The various
learning technique may be superiorly used by the radiologists databases available may be like ANODE09 database, ELCAP
which enhances the effectivity of the segmentation, classification database, LUNA16 dataset, LIDC-IDRI etc. Among available
and detection. Deep learning approach is a potential technique databases, LIDC-IDRI is most commonly standard dataset for
which solves many of the medical challenges in spite of evaluation of lung nodules [18].
limitations like clinical verifications, privacy protection, legal
To detect the shape of the nodules the polygon
accountability etc. [16].
approximation methods can be deployed. The author has adapted
Lung cancer detection is the challenging task for clinical and 3 steps which include region separation using multi-thresholding
medical professionals to identify the true cause for cancer with in method, filtering and smoothing the regions and boundaries
the human body and treatment is the risky and challenging task morphological method are considered. In the next step polygon
because exact treatment have not been invented so far. If cancer approximation methods are used for nodule extraction. Further in
is detected in the initial stage that can be partially treated. In the the following step hybrid feature vector is created in
medical image processing system involves reduction of noise, considerations with the histogram of intensity of the image,
feature extraction, detection of damaged segments and oriented -gradients, geometrical features of enhanced lung nodule
comparison with the of medical history which used to locate candidates. The selected features were classified using SVM
exact region where lung cancer is located. This research has classifications, the image intensity, geometric image features
shown an accurate detection and prediction of lung cancer with vector is fused in to classifier like SVM for lung nodule
the help of image processing techniques and Machine learning identification. The proposed system has achieved accuracy of
techniques. The image filtration steps initiated by the researcher 98.8%, sensitivity of 97.7%, specificity 96.2% by using LIDC
using K-mean method and in the next step images were dataset. In the modern lung cancer technology 3D based
segmented. For classification ML Techniques like KNN, RF and segmentation methods can be adapted to detect lung nodules
ANN were used. Among these ML techniques ANN model accurately [19].
produces better accuracy for prediction of Lung Nodules.
Lung image dataset REVIEW:
Approximately one –million lung cancer patients claiming the
The lung image dataset collection method includes 7 academic
lives because of lung cancer disease since it is a deadliest type of
centers and 8 medical imaging collection companies combinable
cancer disease. According to current affairs in medical challenges

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 716
identified addresses to solve challenges like technical, clinical to 3mm, non-nodule size≥3mm [20],[21]. The LIDC/IDRI Dataset
create a strong database. The LIDC/IDRI Database contains 1018 can be used to equip an important in research of medical imaging
CT images. In the preliminary stage each radiologist separately resource wrt CAD development, validations of the results, and
reviewed CT scanned images and identified nodules belonging to diffusion of clinical applications.
any one of these categories’ nodule size ≥3mm, nodule size <
Table 1: Lung image dataset review
Year of Datasets of CT
Reference Results %
publication images
27 2020 3D-UNet LUNA 16 DSC-95.30
28 2020 DB-ResNet LIDC-IDRI DSC-82.74
END TO END DEEP
29 2021 1916 LUNG TUMOURS IN 1504 PATIENTS SENSITIVITY 93.29
LEARNING
3D ATTENTION
30 2021 COVID 19 DATASET ACCURACY 94.43
UNET

31 2021 IMPROVED U NET LIDC-IDRI PRECISION 84.91

32 2021 U-NET LUNA 16,LIDC-IDRI DSC 89.79,90.35

SENSIIVITY 99.4 +
33 2021 DENSE R2U CNN LUNA
OR – 0.2%
First Affiliated
34 2021 CIRDL SENSIIVITY 87.63
Hospital of Guangzhou University
35 2022 DENSENET201 Seoul St Mary's hospital dataset SENSIIVITY 96.2
The researchers were introduced cascaded 2D and 3D CNN
[22] An automated identification of nodules in lung CT images is model, the results were compared with the previous existing
the active and potential area of research from last two decades. techniques of cancer detection and segmentation methods and
Even though there is limited researches have been carried out they attained a Dice –coefficient metrics for segmentation of lung
which includes comparative analysis of different systems wrt to nodules and sensitivity of nodule detection strategies.
common database. The automatic nodule identification technique The authors developed an image database which is digital,
involves LUNA16 which analyses public dataset of CT images contains 247 chest radiographs which includes with and without
like LIDC-IDRI dataset. nodules. Around 154 chest radiographs with nodules and 93
The LUNA 16, researches can develop and upload predictions on without nodules were designated from 14 medical centres, later
CT scanned images in two different tracks, full or complete these images were digitised in to matrix size of 2048 × 2048
nodule detection technique in which full CAD system can be (0.175-mm pixels) and a gray scale size of 12-bit. The lung
developed. And in the second approach Erroneously reduction in nodules were categorized in to 5 major groups like level 1, very
which a complete set of lung nodule classification be carried out. subtle (detection may be very crucial since this level has low
The proposed system adapted the leading solution in which CNN contrast, small size, or overlap with a normal structure); level 2,
model was deployed, the combined solution exhibits excellent very -subtle (detection itself is crucial (very difficult); level 3,
sensitivity of 95% at lesser than 1.0 false -positive/scan. subtle (detection: difficult); level 4, relatively obvious (detection :
relatively easy); and level 5, obvious (detection : easy). The final
[23] ELCAP is the publicly available database which include set
outcomes were analysed by considering 20 radiologist
of CT images which can be used in different lung diagnosis
observations which is later implemented to ROC analysis
systems. This database contains 50 LDCT lung images &379
(receiver operating characteristic) to detect pulmonary nodules.
unduplicated CT images of lung nodule
[26]
[24] In this proposed study authors have developed new dataset
and conducted testing of computer aided pulmonary nodule
detection and related few strategies in which intended to
complement the existing datasets by adding additional focuses on
radiological variabilities along with clinical and local reality. The
nodule detection, segmentation, characterization techniques and
its outcomes were tested and done comparative study with respect
strategies of radiologist and computer aided systems. This
method adapted Lung Nodule Database (LNDb)with includes 294
CT images (Centro Hospital) - Universitario de São Joãao.
[25] In this proposed system author has adapted maximum
intensity project method while pre-processing and segmentation
method included was segmentation system like SquExUNet and
classification model like 3D-NodNet by considering publicly
accessible database of lung images. Like (LIDC), LNDb
Challenge- Dataset, Indian Lung CT Image Database (ILCID).

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 717
 Table 2 Lung nodule segmentation approaches. automated detection is challenging and key role in spite of
advancements in the Deep learning area. The authors have
assumed that oncology department contours may provide a large
Reference Dataset Sample Number database of 3D segmentation to achieve best models. And they
20 LIDC 1018 CT lung images developed and validated a model using DL to detect and segment
lung tumours using CT images. By considering 1,916(lung
21 LIDC- 1018 CT images of 1010 patients
tumors) of 1,504- patients with radiation treatments in which
IDRI
segmentation of image quality was verified by oncologist with
22 LUNA 888 Lung CT Images :dataset is the help of custom web-based application. While comparing the
16 LIDC-IDRI performance evaluation an external test set of CT images as well
23 ELCAP LDCT images 50 and 379 as segmentation of 59 patients who had single primary /metastatic
unduplicated CT images type of tumor. The sensitivity of detection of tumor with the
definition in which correct prediction at least one voxel with in
24 LNDb 294 CT images (Centro hospitaller
ground truth tumor was 93.2%, Dice coefficient=0.67 with IQR
Universiatario de Sas Joaao
range from 0.53 to 0.85. The researchers have achieved strong
co-relation between physicians predicted tumor size and model
25 ILCID CT Images (400 patients) predicted with r=0.69, P<0.0001.
26 JSRT 154,93 nodules and non nodules with This article helps to develop a systematic model for cancer
lables diagnosis, treatment using a new segmentation approach which
meets the necessity of lung CT- images processed during COVID
-19 pandemic. Initially the selected extracted area in which
The author has proposed a CT image segmentation methodology
implement patch mechanism to fulfil the 3D network
which is based on 3D-UNet, Res2Net and developed CNN model
applicability, eliminated the unwanted background. In the second
referred as 3D-Res2UNet.This new model has identical
step 3-D model was developed to attain a targeted area. The
connection including all the relevant attributes extraction
improvements in the network convergence and loss function is
capacities. This enables the established network to enhance
added to boost gradient while optimizing and training direction.
multiple scaled features with in a excellent granularity as long as
At last augmentation method, conditional random field was used
elaborating the individual filed layers of the model. The proposed
to resample and segment a given data. This method was assessed
technique figures out the depth problems but the modelled
using some comparative study experiments and proposed system
network is not suitable for gradient dis-appearance and
achieves better performance so it has high clinical applications
exploration problem which may enhance the segmentation -
[30].
accuracy. The developed network ensures about feature map size,
repair effectively the lost image features. The data set used was This paper implements DNN established a lung cancer
LUNA16, dice-coefficient indexed value = 95.30% , recall classification using optimizations like hyper parameter. This
rate=99.1% which indicates this method has achieved high optimization of DNN is costly, surrogate –assisted evolution
performance while segmenting Lung image nodules [27]. algorithm introduced to search automatically an optimal hyper
parameter configuration on DNN.This proposed model is distinct
Here the authors have proposed a data driven model called as
from the available surrogate models which adapts covariance
DB-ResNet and integrated two techniques to improve the model
function which is stationary ie kernel to compute difference
performance in which first method can continuously capture
among the hyper parameter points. The non- stationary surrogate
multiple viewed and scaled features of CT lung-nodules. During
models adapt a function whose smoothness changes with respect
second method researchers have combined intensity features and
to spatial- location of the given input. The ML-CNN model was
CNN and adapted pooling methods like intensity of the pooling
built for lung nodule classification, optimization of available
layer to take intensity parameters of voxel (center ) of adapted
hyper parameter configuration is carried out using non-stationary
system model block, in the later stage CNN model was used to
kernel. This system outperforms manual tuning,
get convoluted lung features of voxel of the block. In additions to
optimization(hyperparameter), search (random grid), Gaussian
this weighted sampling strategy was considered which is depends
processes, The Parzen Estimator Approach which is Tree-
on the weighted scores of boundaries of the lung nodules which
structured, Hyper parameter Optimization using RBF, HORD
will improve the results accuracy. The model was evaluated using
(Dynamic coordinate search) [31].
LIDC dataset which contains 986 lung nodules. The carried
experimental results exhibits DB-ResNet is superior segmentation
performances including Dice score=82.74% for LIDC database Performance Evaluation Metrics:
[28]. The performance evaluation measures of segmentation,
The outcomes of proposed study were validated with the classification of DL models like ACC, SN, AUC, TP, SP are
available findings of four radiologists using the same dataset. The basic factors need to be reviewed in the research area. Sensitivity
average dice score of software was higher than the average dice means recalling the probability of recognizing true pixels of
score of human experts by 0.49 percent. Hence, proposed segmented images. SP entitles an ability of identification of
research work is just as better as an experienced medical negative pixels. Acc can be defined as the ratio between truly
radiologist’s.[29] The main purpose of accurate estimation of identified pixels. The total pixels with in the image may be (TP +
lung tumor in the case of biomedical image is an important aspect FP + TN + FN). In the proposed system author has presented an
for real identification and monitoring therapeutic response. An automated lung nodule detection technique by initializing
automated region(seeded) growing (SRG) to perform the
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 718
segmentation technique of lung images. Without considering In the testing phase, a separate dataset (the testing dataset) is
homogeneity condition SRG can segment and CT image used, which the model has not seen during training. This dataset
automatically, it will not focus on similarities between the is also composed of multi-Model medical images with known
neibouring pixels. This system approaches a code technique like ground truth labels.
3D chain for identification of juxta-pleural nodules. It extracts 2.Forward Pass:
segmented lung image voxelized regions and boundary to identify
The testing images are passed through the trained model for
regions and boundaries and detect infected areas. Use of 3D chain
inference. The model predicts whether each image contains
coding is more effective in order to exhibit the better accuracy in
cancerous regions based on the learned fea-tures.
results, it does not need training of data set, classification
methodologies to initialize beginning and endpoint gaps between 3.Performance Metrics:
a nodule. [22]. Various performance metrics are calculated based on the
model's predictions and the ground truth labels in the testing
dataset. These metrics include:
3. Methodology Of Proposed System
• Accuracy: The percentage of correct predictions.
The automatic detection of lung nodule using ML techniques was
adapted in which the technique uses two different phases like • Precision: The ratio of true positives to the sum of true
training and testing phase. During train-ing phase the lung images positives and false pos-itives.
are augmented means preprocessed and next fed in to deep • Recall (Sensitivity): The ratio of true positives to the
learning model. The output of deep learning model produces sum of true positives and false negatives.
features acts as an input to the ML algorithm. Under testing phase • F1-Score: The harmonic mean of precision and recall.
after image screening classification of lung nodules using
• ROC-AUC: The area under the receiver operating
objective or subjective analysis methodologies were adapted.
characteristic curve, which measures the trade-off between true
Training Phase:1.Data Preparation:In the training phase, a diverse positive rate and false positive rate.
dataset of multi-Model medical images, such as X-rays and CT
4.Interpretability:
scans, is prepared. This dataset contains labeled examples
indicating whether each image contains cancerous regions Methods for interpretability, such as generating heatmaps or
(positive class) or not (negative class). saliency maps, may be applied to visualize which regions of the
image influenced the model's predic-tions. This can help in
2.Data Augmentation:
understanding why the model made specific decisions.
Data augmentation techniques are applied to create variations of
5.Fine-Tuning (Optional):
the training im-ages. This includes transformations like rotation,
scaling, and flipping to increase the diversity of the training If the model's performance is found to be unsatisfactory during
dataset. testing, further fi-ne-tuning or optimization of hyperparameters
may be considered to improve ac-curacy.The testing phase
3.Model Initialization:
evaluates the model's real-world performance in accurately
The multi-Model CNN model is initialized with random weights identifying lung cancer regions in multi-Model medical images
and architecture designed to handle different image Modalities. It and provides valua-ble insights into its reliability and
usually includes convolutional layers for feature extraction, effectiveness in clinical practice.
followed by fully connected layers for classification.
4.Loss Function and Optimization:
During training, the model computes a binary cross-entropy loss,
which measures the dissimilarity between the predicted output
and the true labels. The optimiza-tion algorithm, typically Adam
or SGD (Stochastic Gradient Descent), adjusts the model's
parameters to minimize this loss.
5.Forward and Backward Pass: The training dataset is processed
in batches. For each batch, a forward pass com-putes the model's
predictions, while a backward pass updates the model's weights
using gradients obtained from the loss function.
6.Epochs:
The training process occurs over multiple epochs, where each
epoch represents on complete pass through the entire training
dataset. This repetition allows the model to learn iteratively.
7.Validation:
A validation dataset, separate from the training data, is used to
monitor the mod-el's performance during training. This is
important for early detection of overfit-ting.
Testing Phase:
1.Data Preparation:

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 719
 Fig.5. Model accuracy of the proposed system

Fig. 3. Proposed Methodology for Lung Cancer Detection

Model
4. Results And Discussions
Fig .8. lung cancer detection of CT images for different cases
The proposed method uses Deep Learning model for
5. Conclusion
preprocessing segmentation and lung nodule detection. The
Figure. 4 shows lung image of a malignant case using machine The proposed system discusses an automatic detection of lung
learning algorithm. The performance of the model achieved with cancer using Deep learning techniques using CT images. The
accuracy of 99.48%, the accuracy and model loss is show in machine learning algorithms and Deep learning techniques were
Figure 5 and Figure 6 respectively. The presence cancer using CT used for segmentation and validation purpose. The adapted
images like normal or malignant cases are listed in Figure 7. method accurately detects lung nodules with accuracy of 96% .In
the future scope an android based Lung cancer detection
application can be introduced which may help medical experts to
detect lung cancer disease during initial stages
6. Authors’ Note
The author(s) declare(s) that there is no conflict of interest
regarding the publication of this article. Authors confirmed that
the data and the paper are free of plagiarism

References
[1]. D. P. Naidich “Lung Cancer Detection and Characterization:
Challenges and Solutions” SpringerLink DOI: 10.1007/978-3-
642-18758-2_17.
[2]. Xiaodan Chen, Shouting Feng, Daru Pan” An improved
approach of lung image segmentation based on watershed
algorithm” - ICIMCS '15: Proceedings of the 7th International
Fig.4. Malignant case in CT image Conference on Internet Multimedia Computing and Service
August 2015 Article No.: 39Pages 1–
5https://doi.org/10.1145/2808492.2808531.
[3]. Manikandan T “Challenges in lung cancer detection using

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 720
computer-aided diagnosis (CAD) systems – a key for survival of and Engineering nced Trends in Computer Science and
patients. Archives of General Internal Medicine Volume 1 Issue Engineering.
2. [16]. Lulu Wang “Deep Learning Techniques to Diagnose Lung
[4]. Taruna Aggarwal; Asna Furqan; Kunal Kalra” Feature Cancer” Cancers 2022, 14, 5569.
extraction and LDA based classification of lung nodules in chest https://doi.org/10.3390/cancers14225569.
CT scan images” Published in: 2015 International Conference on [17]. Sharmila Nageswaran , G Arunkumar , Anil Kumar Bisht ,
Advances in Computing, Communications and Informatics Shivlal Mewada , J N V R Swarup Kumar , Malik Jawarneh ,
(ICACCI) DOI: 10.1109/ICACCI.2015.7275773. Evans Asenso “Lung Cancer Classification and Prediction Using
[5]. P. B. Sangamithraa, S. Govindaraju “Lung tumour detection Machine Learning and Image Processing” Biomed Res Int. 2022;
and classification using EK-Mean clustering” Published 23 2022: 1755460. Published online 2022 Aug 22. doi:
March 2016 Computer Science, Medicine 2016 International 10.1155/2022/1755460.
Conference on Wireless Communications, Signal Processing and [18]. Setio, A.A.A.; Ciompi, F.; Litjens, G.; Gerke, P.; Jacobs, C.;
Networking (WiSPNET) DOI:10.1109/WISPNET.2016.7566533. Van Riel, S.J.; Wille, M.M.W.; Naqibullah, M.; Sánchez, C.I.;
[6]. Wei Li, Peng Cao, Dazhe Zhao and Junbo Wang Van Ginneken, B. Pulmonary nodule detection in CT images:
“Pulmonary Nodule Classification with Deep Convolutional False positive reduction using multi-view convolutional
Neural Networks on Computed Tomography Images” Comput networks. IEEE Trans. Med. Imaging 2016, 35, 1160–1169.
Math Methods Med. 2016; 2016: 6215085. Published online 2016 [Google Scholar] [CrossRef] [PubMed].
Dec 14. doi: 10.1155/2016/6215085. [19]. Naqi, S.; Sharif, M.; Yasmin, M.; Fernandes, S.L. Lung
[7].Zakaria Suliman Zubi, Rema Asheibani Saad “Improves nodule detection using polygon approximation and hybrid
Treatment Programs of Lung Cancer Using Data Mining features from CT images. Curr. Med. Imaging 2018, 14, 108–
Techniques” Journal of Software Engineering and Applications > 117. [Google Scholar] [CrossRef]
Vol.7 No.2, February 2014 DOI: 10.4236/jsea.2014.72008. [20] Paing, M.P.; Hamamoto, K.; Tungjitkusolmun, S.;
[8]. Xinyan Li,Shouting Feng,Daru Pan” Enhanced lung Visitsattapongse, S.; Pintavirooj, C. Automatic detection of
segmentation in chest CT images based on kernel graph cuts” pulmonary nodules using three-dimensional chain coding and
ICIMCS'16: Proceedings of the International Conference on optimized random forest. Appl. Sci. 2020, 10, 2346. [Google
Internet Multimedia Computing and ServiceAugust 2016 Pages Scholar] [CrossRef][Green Version].
228–233https://doi.org/10.1145/3007669.3007690. [21]. Lung Image Database Consortium (LIDC). Available
[9]. Hanan M. Amer,Fatma E. Z. Abou-Chadi,Sherif S. online: https://imaging.nci.nih.gov/ncia/login.jsf (accessed on 5
Kishk,Marwa I. Obayya” A Computer-Aided Early Detection October 2022).
System of Pulmonary Nodules in CT Scan Images” ICSIE '18: [22]. Armato Samuel, G.; McLennan, G.; Bidaut, L.; McNitt-
Proceedings of the 7th International Conference on Software and Gray, M.F.; Meyer, C.R.; Reeves, A.P.; Zhao, B.; Aberle, D.R.;
Information EngineeringMay 2018 Pages 81– Henschke, C.I.; Hoffman, E.A.; et al. Data from LIDC-IDRI.
86https://doi.org/10.1145/3220267.3220291. 2015. Available online:
[10]. Aristófanes C. Silva,Paulo Cezar P. Carvalho,Marcelo https://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI
Gattass” Diagnosis of lung nodule using Gini coefficient and (accessed on 5 October 2022).
skeletonization in computerized tomography images” SAC '04: [23]. Setio, A.A.A.; Traverso, A.; de Bel, T.; Berens, M.S.N.; van
Proceedings of the 2004 ACM symposium on Applied den Bogaard, C.; Cerello, P.; Chen, H.; Dou, Q.; Fantacci, M.E.;
computingMarch 2004 Pages 243– Geurts, B.; et al. Validation, Comparison, and Combination of
248https://doi.org/10.1145/967900.967954. Algorithms for Automatic Detection of Pulmonary Nodules in
[11] H Witschi “A short history of lung cancer” PMID: 11606795 Computed Tomography Images: The LUNA16 Challenge. Med.
DOI: 10.1093/toxsci/64.1.4. Image Anal. 2017, 42, 1–13. [CrossRef]
[12]. Orkun Furat1,Mingyan Wang, Matthias Neumann, Lukas [24]. ELCAP Public Lung Image Database. 2014. Available
Petrich, Matthias Weber, Carl E. Krill I and Volker Schmidt” online: http://www.via.cornell.edu/lungdb.html (accessed on 5
Machine Learning Techniques for the Segmentation of October 2022).
Tomographic Image Data of Functional Materials” Research [25]. Pedrosa, J.; Aresta, G.; Ferreira, C.; Rodrigues, M.; Leito,
Article | Open Access Volume 2016 | Article ID 6215085 | P.; Carvalho, A.S.; Rebelo, J.; Negrao, E.; Ramos, I.; Cunha, A.;
https://doi.org/10.1155/2016/6215085. et al. LNDb: A Lung Nodule Database on Computed
[13]. E R Heitzman “The role of computed tomography in the Tomography. arXiv 2019, arXiv:1911.08434.
diagnosis and management of lung cancer. An overview” [26]. Prasad, D.; Ujjwal, B.; Sanjay, T. LNCDS: A 2D-3D
MedlinePlus Health Information, The American College of Chest cascaded CNN approach for lung nodule classification, detection
Physicians. Published by Elsevier Inc., 1986. 89(4): p. 5. andsegmentation. Biomed. Signal Process. Control. 2021, 67,
[14]. M A Howe, B H Gross” CT evaluation of the equivocal 102527.
pulmonary nodule” Comput Radiol. 1987 Mar-Apr;11(2):61-7. [27]. Shiraishi, J.; Katsuragawa, S.; Ikezoe, A.; Matsumoto, T.;
doi: 10.1016/0730-4862(87)90012-6. Kobayashi, T.; Komatsu, K.; Matsiu, M.; Fujita, H.; Kodera, Y.;
[15]. Vikul J. Pawar, Kailash D. Kharat, Suraj R. Doi, K.Development of a digital image database for chest
Pardeshi,Prashant D. Pathak “Lung Cancer Detection System radiographs with and without a lung nodule: Receiver operating
Using Image Processing and Machine Learning Techniques” characteristicanalysis of radiologists’ detection of pulmonary
International Journal of Advanced Trends in Computer Science nodules. Am. J. Roentgen. 2000, 174, 71–74. [CrossRef]

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 721
[PubMed] Conference on Advances in Computing and Communication
[28] .Xiao, Z.; Liu, B.; Geng, L.; Zhang, F.; Liu, Y. Segmentation (ICACC) .
of lung nodules using improved 3D-Unet neural network. [42] Mutiullah, M. Bari, A. Ahmed, M. Sabir and S. Neveed,
Symmetry 2020, 12, 1787. [CrossRef]. "Lung Cancer Detection Digital Image Processing Techniques:
[29].Cao, H.C.; Liu, H.; Song, E.; Hung, C.C.; Ma, G.Z.; Xu, A Review", Mehran University Research Journal Engineering &
X.Y.; Jin, R.C.; Jianguo Lu, J.G. Dual-branch residual network Technology, vol. 38, no. 2, pp. 351-360, April 2019.
for lung nodule segmentation. Appl. Soft Comput. 2020, 86, [43] S. Makaju, P. W. C. Prasad, A. Alsadoon, A. K. Singh and
105934. [CrossRef]. A. Elchouemi, "Lung cancer detection using CT scan images",
[30] Kashyap, M.; Panjwani, N.; Hasan, M.; Huang, C.; Bush, K.; Procedia Computer Science, vol. 125, pp. 114.
Dong, P.; Zaky, S.; Chin, A.; Vitzthum, L.; Loo, B.; et al. Deep [44]S. Nageswaran, G. Arunkumar, A. K. Bisht, S. Mewada, J. N.
learning based identification and segmentation of lung tumors on Kumar,M. Jawarneh, and E. Asenso, “Lung Cancer Classification
computed tomography images. Int. J. Radiat. Oncol. Biol. Phys. and prediction using machine learning and image processing,”
2021, 111(3S), E92–E93. [CrossRef]. BioMed Research International, vol. 2022, pp. 1–8, 2022.
[31] Chen, C.; Zhou, K.; Zha, M.; Qu, X.; Xiao, R. An effective [45]J. W. Yang, D. H. Song, H. J. An, and S. B. Seo,
deep neural network for lung lesions segmentation from COVID- “Classification of subtypes including LCNEC in lung cancer
19 CT images. IEEE Trans. Ind. Inform. 2021, 17, 6528–6538. biopsy slides using convolutional neural network from scratch,”
[CrossRef] Scientific Reports, vol. 12, no. 1, 2022.
[32] Zhang, M.; Li, H.; Pan, S.; Lyu, J.; Su, S. Convolutional [46]Q. Wu and W. Zhao, “Small-cell lung cancer detection using
neural networks based lung nodule classification: A surrogate- a supervised machine learning algorithm,” 2017 International
assisted evolutionary algorithm for hyperparameter optimization. Symposium on Computer Science and Intelligent Controls
IEEE Trans. Evol. Comput. 2021, 25, 869–882. [CrossRef]. (ISCSIC), 2017.
[33]. Banu, S.F.; Sarker, M.; Abdel-Nasser, M.; Puig, D.; [47]A. Gupta, Z. Zuha, I. Ahmad, and Z. Ansari, “A study on
Raswan, H.A. AWEU-Net: An attention-aware weight excitation prediction of lung cancer using machine learning algorithms,”
u-net for lung nodule segmentation. arXiv 2021, 2022.
arXiv:2110.05144. [CrossRef] [48]O. Obulesu, S. Kallam, G. Dhiman, R. Patan, R. Kadiyala, Y.
[34]. Dutta, K. Densely connected recurrent residual (DENSE Raparthi, and S. Kautish, “Adaptive diagnosis of lung cancer by
R2UNET) convolutional neural network for segmentation of lung deep learning classification using Wilcoxon gain and generator,”
CT images. arXiv 2021, arXiv:2102.00663. Journal of Healthcare Engineering, vol. 2021, pp. 1–13, 2021.
[35] Lin, X.; Jiao, H.; Pang, Z.; Chen, H.;Wu,W.;Wang, X.; [49]P. Ramesh, R. Karuppasamy, and S. Veerappapillai,
Xiong, L.; Chen, B.; Huang, Y.; Li, S.; et al. Lung cancer and “Machine learning driven drug repurposing strategy for
granuloma identification using a deep learning model to extract identification of potential RET inhibitors against non-small cell
3-dimensional radiomics features in CT imaging. Clin. Lung lung cancer,” Medical Oncology, vol. 40, no. 1, 2022.
Cancer 2021, 22, e756–e766. [CrossRef] [50]M. Saminathan, M. Ramachandran, A. Kumar, K. Rajkumar,
[36]. Kim, H.M.; Ko, T.; Young, C.I.; Myong, J.P. Asbestosis A. Khanna, and P. Singh, “A study on specific learning
diagnosis algorithm combining the lung segmentation method algorithms pertaining to classify lung cancer disease,” Expert
and deep learning model in computed tomography image. Int. J. Systems, vol. 39, no. 3, 2021.
Med. Inform. 2022, 158, 104667. [CrossRef] [51]A. Kumar Dutta, “Detecting lung cancer using Machine
[37] Deep Learning Algorithm for Classification and Prediction Learning Techniques,” Intelligent Automation & Soft
of Lung Cancer using CT Scan Images published in 2019 5th Computing, vol. 31, no. 2, pp. 1007–1023, 2022.
International Conference on Computing Communication Control [52]M. Mamun, A. Farjana, M. Al Mamun, and M. S. Ahammed,
and Automation (ICCUBEA) “Lung cancer prediction model using Ensemble Learning
[38] Random Forest based Classification Model for Lung Cancer Techniques and a systematic review analysis,” 2022 IEEE World
Prediction on Computer Tomography Images by D. Jayaraj AI IoT Congress (AIIoT), 2022.
published in 2019 International Conference on Systems and [53]E. Dritsas and M. Trigka, “Lung cancer risk prediction with
Inventive Technology (ICSSIT). machine learning models,” Big Data and Cognitive Computing,
[39] Predicting Outcomes of Non small Cell Lung Cancer Using vol. 6, no. 4, p. 139, 2022.
CT Image Features by Samuel H. Hawkins; John N. Korecki; [54]N. Banerjee* and S. Das, “Machine learning techniques for
Yoganand Balagurunathan; Yuhua Gu; Virendra Kumar prediction of lung cancer,” International Journal of Recent
published in 2169-3536 2014 IEEE. Translations Technology and Engineering (IJRTE), vol. 8, no. 6, pp. 241–249,
[40] Prediction of Lung Cancer from Low-Resolution Nodules in 2020.
CT-Scan Images by using Deep Features by Anand Gupta; [55]G. M. Alshmrani, Q. Ni, R. Jiang, H. Pervaiz, and N. M.
Sagorika Das; Tarasha Khurana; Kamakshi Suri published in Elshennawy, “A deep learning architecture for multi-class lung
2018 International Conference on Advances in Computing, and diseases classification using chest X-ray (CXR) images,”
Informatics (ICACCI) Alexandria Engineering Journal, vol. 64, pp. 923–935, 2023
[41] Lung Nodule Classification in CT Images Using [56]A. S. U., F. R. P. P., A. Abraham, and D. Stephen, “Deep
Convolutional Neural Network by Gilu K Abraham; Preethi learning- based BoVW–CRNN model for lung tumor detection in
Bhaskaran; V. S Jayanthi Published in: 2019 9th International nano- segmented CT images,” Electronics, vol. 12, no. 1, p. 14,

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 722
2022.
[57]K. Ramana, M. R. Kumar, K. Sreenivasulu, T. R. Gadekallu,
S. Bhatia, P. Agarwal, and S. M. Idrees, “Early prediction of lung
cancers using deep saliency capsule and pre-trained deep learning
frameworks,” Frontiers in Oncology, vol. 12, 2022.
[58]C. Anil Kumar, S. Harish, P. Ravi, M. SVN, B. P. Kumar, V.
Mohanavel, N. M. Alyami, S. S. Priya, and A. K. Asfaw, “Lung
cancer prediction from text datasets using machine learning,”
BioMed Research International, vol. 2022, pp. 1–10, 2022.
[59]M. Humayun, R. Sujatha, S. N. Almuayqil, and N. Z. Jhanjhi,
“A transfer learning approach with a convolutional neural
network for the classification of lung carcinoma,” Healthcare,
vol. 10, no. 6, p. 1058, 2022.
[60]R. P.R., R. A. S. Nair, and V. G., “A comparative study of
lung cancer detection using machine learning algorithms,” 2019
IEEEInternational Conference on Electrical, Computer and
Communication Technologies (ICECCT), 2019.
[61]G. Zaman Khan et al., "An Efficient Deep Learning Model
based Diagnosis System for Lung Cancer Disease," 2023 4th
International Conference on Computing, Mathematics and
Engineering Technologies (iCoMET), Sukkur, Pakistan, 2023,
pp. 1-6, doi: 10.1109/iCoMET57998.2023.10099357.

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(3), 713–723 | 723